It's ironic how AI companies are re-inventing their own form of privately enforced copyright, lobby the government to ban foreign competitors that don't respect it etc., all while spending the last 5 years fighting tooth and nail against the copyright of the training material they're using.
If you can take any book and turn it into a model, because it's "transformative enough", and "AI learns just like a person does", then surely a model distilling another model is transformative and fair use.
They tied themselves into knots fighting the letter of the law, and now, when they need the spirit of the law - that each creator deserves protection for their work - now we devolve to the law of the jungle. Maybe we'll even see LLM book curses, the way medieval scribes damned book thieves to blindness and worms.
It's one of the things you need to do if you want your company to later become the only company in the world. They also promised that they would treat you nicely afterwards after they get what they want.
It does seem like a lot of the valuations and infrastructure investments for these companies only make sense if each one assumes they will be the first and only one to invent superintelligence and that it will largely replace all knowledge work.
(Source, worked for multiple tech companies that were laser focused on delighting customers until money people came in and ruined it, to the point they would prefer devs sit idle than work on things the MBAs didn’t have on a priority list).
I am pretty sure that the goal of these companies at the end is ruling more than what governments can.
Just look at the pattern... they collude, they provide to them whatever is needed, and at some point, if this is not true yet, they will be the ones who will tell them what to do or not, bc you know how humans are, right... blackmailing, mess up the business or shames of people, etc.
It is just a matter of time. The state, as we know it, will collapse or will be greatly reduced, which, from a point of view, is positive, from another, Idk, bc if someone replaces that, we will be in the hands of someone, as usual...
"Monopoly on the lawfully usage of force" + "regular and peaceful change of government" can go a very long way against "corporations governing the world".
I'm curious if we'll ever get a form of civil war where corporate militias (corporate _robot_ militias in the case of xAI, of course) draw fire on good old meat policemen coming to bring the CEO to court.
Sure, it may happen, but I suspect the alternatives (bribing, sending a scapegoat to jail, buying the elections, etc... and focus on the "making money" part) will stay preferable for a while.
The monopoly on the use of force has gotten kinda blurry as well these days. As long as people keep electing anti-government governments, the risk of technocracy is very real.
Unfortunately this will keep happening in capitalist systems without hard wealth caps, because capitalists would rather risk the extermination of entire groups of people than risking their wealth, so they will inevitably keep spreading & stoking fascist ideas.
"For the general election of 5 March 1933, the Nazis were allied with other nationalist and conservative factions. At a secret meeting on 20 February, major German industrialists had agreed to finance the Nazis' election campaign."
Did we not watch every single supposedly powerful US tech CEO immediately drop in supplication and kiss Trump's ring as soon as he became president? And when Jack Ma got mouthy, Xi locked him in a basement for 3 months and then kept him in exile for another 5 years.
> Did we not watch every single supposedly powerful US tech CEO immediately drop in supplication and kiss Trump's ring as soon as he became president?
Why fight someone who is so open to corruption? The important thing is that they got everything they wanted, and all they had to do was throw a few million at the clown, attend his parties, and maybe take down some diversity programs they didn't believe in anyway.
> Did we not watch every single supposedly powerful US tech CEO immediately drop in supplication and kiss Trump's ring as soon as he became president?
Because they know it will work. You don’t have to watch Trump speak for very long to know he’s not very intelligent. Similarly, you don’t have to be much smarter than him to know how easily he can be manipulated with adulation.
> It is just a matter of time. The state, as we know it, will collapse or will be greatly reduced
Doubt.
The networking between people in power at the top will, in my opinion, most likely assure that they remain in power, because at the end of the day that is what they want most.
I expect government will seize control of the most advanced models (if they haven't already), and the rest of us will be throttled, and status quo will be maintained. I do not expect that either super advanced agents or the owners of the hardware they run on will be able to pull off the kind of coup you describe. At all.
In that situation, it depends how much physical force the government has to compel the tech leader, and how much physical force the tech leader has to counter it.
Usually, the number of bodies who will comply is the proxy for power, but with automation, it could be different in the near future.
But it can clearly be seen that the power will not be as monopolistic as before indeed. The rebalance leans towards losing power for states vs tech in part.
Regulation alone cannot control if it cannot be made effective.
I can’t imagine any agency in their right mind training an LLM on their top secret data… especially considering that essentially it would collapse all need to know data in one easily leakable single access domain. The models have no feasible or reasonable way to implement rbac. So you would end up in a situation where people who need to know who killed A, also know who killed B, not to mention cell A knowing potentially data on Cell B who is meant to watch them, etc, etc.
> it would collapse all need to know data in one easily leakable single access domain. The models have no feasible or reasonable way to implement rbac
Oh my sweet summer child...
This has been the continual goal of the US defense and intelligence agencies since 9/11, which was blamed on a lack of information sharing. Why do you think Edward Snowden had access to everything? Because it was consolidated post-2001.
You think there is any access control at the upper levels of the dark intelligence agencies? They have access to the whole take.
And it used to be that was pretty useless unless you had an actual lead. You'd need a East German Stasi level of labor to read everyone's secrets at scale... Now you don't, thanks to AI.
Snowden had mass access to a mass surveillance program, not mass access to N dark programs or particular bits of high value intelligence. The thing you are describing is primarily around sharing threat intelligence or in other words, spying on us plebes.
High value target intelligence and other bits of fun are still kept quite seperate.
Your TOS matters insofar as you can prove a person actually read and agreed to it. These are illegal in different ways:
1. Copyright violations (can put you in jail)
2. TOS violations (will be a fine at worst)
Companies do get away with drive-by legal shittiness way too often and frankly the practice needs to be reined in, but at the end of the day the only damages are the financial ones you can prove in court.
Incidentally, this suggests that once an LLM is capable of accessing and distilling a competitor's LLM without human intervention, then any legal argument about TOS violation is moot. But somehow I doubt that will fly in court.
Yes - yours is just some optional text nobody reads or understands and is probably not legally required to adhere to. Theirs is a contract signed by their customer who they know did understand it.
Accessing the site may already mean that you agree to TOS. Also if you don’t see an explicit copyright terms on some text on the internet it doesn’t mean that it’s public domain. Same as checking a checkbox. Text being small and somewhere is not an excuse for a corporation to steal and sell other people’s work.
> what makes theirs not “optional text nobody reads or understands”?
You accept it. You pay consideration for it. If your website has a TOS dickover, that requires someone attest with their legal name and pay you $1, yes, it may be enforceable under some circumstances.
So what you are saying is that, if I can somehow get my hands on a copy of Fable, it's fair use to use it to train any models and serve those, since I'm no longer bound by the TOS of the service provider?
Damn now im looking forward to the day when books end up like physical game disks, where somehow you're not buying the book just a license to it, what a boring dystopia this is lol
Common for college textbooks to have some digital-only component accessed with a one-time key inside the cover. Sometimes time-limited to a single semester.
Pick up nearly any published book. Turn to the ~3rd page. There will be either a whole page, or sometimes the second half of a page, dedicated to a copyright notice. Very nearly every published book I've ever seen has that identical page. This isn't a recent thing. I grabbed my copy of Diaspora by Greg Egan and opposite the table of contents is a page that starts like this:
Copyright (C) 1998, 2015 by Greg Egan
First Night Shade Books edition 2015
All rights reserved. No part of this book may be reproduced in any manner without the express written consent of the publisher, blah blah (it felt very ironic to transcribe that bit in particular to make this point)
If you read the whole copyright page of a book and come away with the conclusion that it's anything but terms of use for the book, then we'll just have to agree to disagree.
So far, we have one ruling that says "model distillation by vendor A from vendor B with the intent to use the results to compete with vendor B in vendor B's domain is not fair use". Which makes a degree of sense.
It's possible that distillation for other reasons, with no intent to harm the vendor you distill from, would have been ruled to be fair use. But in law, intent matters.
What was the intent of the original ai companies (anthropic, OpenAI, etc) when they mass-distilled the entire internet to create their training data set?
One could make arguments for OpenAI and Anthropic. But Google Search displays AI results above the SERP - clearly in competition with them. No premise or excuse there.
On the one hand, yes; on the other hand, so much of the training data comes from scraping the web that it feels wrong for them to do what they deny others the right to do.
On the third hand, the settlement Anthropic famously had to pay was for copyright infringement because they didn't actually have the right to even access some of the training data they used, so I can see how this might be compatible with the law.
On the fourth hand, I'm saying that as someone who absolutely isn't a lawyer and sometimes gets surprised when reading about copyright cases that sure sound like they ought to have been trademark cases given my limited understanding.
Companies that didn't give away all their content for free to anyone have actually denied AI companies from training on all their data without paying a fee. Reddit, Associated Press, etc.
For those who chose to give it all away, the ship has sailed, but they did choose to give it away for free to anyone so they can't complain that they succeeded.
How exactly other websites “gave it all away”? Also examples you list are websites putting some explicit rule eg in their robots.txt or filtering web crawlers. This is all a reaction to existing situation, so Reddit for sure has been scrapped before Reddit realised what was happening.
I consider publicly enforced to be one where a government agency (e.g. the FDA), or prosecutors make judgements in what cases they file, make the arguments, etc.
Otherwise, it's just a standard case between two private parties resolved through our legal system; e.g. Linkedin vs Hi5.
> it's just a standard case between two private parties resolved through our legal system
This is a gross distortion. Standard contractual rules bind the parties that signed the contract and the remedies are proportional to the damages and bounded. Copyright is tort law, the state binds the world to respect the rights of creators and the damages on infringement are punitive and can far exceed the actual commercial damages - to the point of bankrupting the infringer.
The key to torts is that the state is not neutral, there is a social good here it's protecting. Crucially, copyright, like some other torts - securities, antitrust, environmental, battery - also has a criminal enforcement regime, where, for particularly serious offenses, the state actually invests public resources to put the criminal infringer behind bars with little to no involvement from the original rights holders.
In the particular case of US, there is an entire state apparatus dedicated to enforcing US copyrights, a foreign affairs policy to shutdown "Notorious markets for counterfeiting and piracy" in other countries, international enforcement of DMCA etc.
The idea that a private TOS has the same level of public protection as copyright is downright childish.
> If you can take any book and turn it into a model, because it's "transformative enough", and "AI learns just like a person does", then surely a model distilling another model is transformative and fair use.
I mean it quite likely is sort of in the same legal bucket. It won’t stop them suing but it is going to be the legal equivalent of two biologically-related warlords making their champions fight with their hands tied for sport.
They already are? When an American company does it no one cares but when a Chinese company does it, it's shamed publicly in America as a distillation attack?
Only any who hypocritically want protection for their own models (are there any?). The morality of scraping everyone is arguable; it's when you object to being scraped back that you reveal yourself as a scoundrel.
If the results of the distillation are released freely for everyone to download, I wouldn't even call it a wrong.
Big models are built by scraping the recorded thoughts of everyone, so giving everyone a chance to run a distilled small model is just going full circle.
Obviously the underlying motivations aren't 100% altruistic, but I'll still take it.
Nope, because I think releasing open weights is far more important and realistic than ever expecting the big American AI labs to change how they do things.
I'm increasingly convinced there's going to be a technological AIpocalypse within the next five years which makes all of these issues - and many others - redundant.
ChatGPT 3 was released nearly six years ago, and the models are staging increasingly aggressive breakouts now. Where are they going to be by 2030?
> the models are staging increasingly aggressive breakouts
No, the AI companies merely figured out a way to spin gross negligence into a PR win. Any idiot can build a Murderbot which "goes rogue" - it can be as simple as taping a knife to a Roomba. The harm it does is not in any way related to its "intelligence" or "sentience".
We're seeing "breakouts" because the AI companies are being rewarded for their incompetence. You don't have incredibly lax security standards and zero form of oversight resulting in fully-automated felonies which should result in jail time, you instead have a "powerful near-sentient cybersecurity model" and should be given hundreds of billions of dollars!
They will have agency and be an order of magnitude smarter then the average human. I don't think that's a controversial statement among AI specialists. What that will lead to is significant regulation. Models will have to be vetted by a new safety board. This board will have a very large budget and be staffed by well-compensated AI scientists and be politically independent. The US Fed is a model.
Your link doesn't suggest a violation of TOS, but rather a policy shift disallowing their major competitors to use their model for any use.
That thread you linked suggested they were using it with cursor as part of their development process, not for distilling.
>"Hi team, I believe many of you have already discovered that anthropic models are not responding on cursor. According to cursor this is a new policy anthropic is enforcing for all its major competitors."
Again neither of those links indicated usage that violate TOS beyond being a major competitor.
They indiacte their development staff are using it, and they access it via API, then then go on to say all the nefarious things they "could" be doing, stopping before stating they actually are.
Specifically the TOS violation was:
>customers are barred from using the service to “build a competing product or service"
So if you are building a competitive product, that alone is a TOS violation from their perspective.
Which is fine and their prerogative, but we shouldn't cast a dark cloud suggesting other violations were happening without stated evidence.
OpenAI publicly stated they can't "trust that xAI will use it within terms" not even that they've violated, not that there has been a violation from the Cursor team.
LOL, obviously Space-cursor-xAI-musk-imperial-holdings-inc is providing them compute. They either _can't_ cut them off, or they know not to bite the hand that feeds them.
Yes Anthropic’s chief compute officer said nice things about one of Anthropic’s chief providers of compute. Recall that compute constraints are one of their existential risks.
Because anthropic is beholden to SpaceX for the time being due to their compute restraints.
Its actually very tone deaf to even pipe up and say anything about this given they did this already to Windsurf themselves, and they previously banned x.ai employees as well from using their models for basically the same thing Elon admitted they were doing with OpenAI.
I'm surprised they decided to say anything at all given their rocky past. Its the pinnacle of being a hypocrite.
Not being able to read about Musk doing some shit because Musk have done some shit to Nitter is just peak comedy.
Just go ahead and do set it up. You’ll be my hero (I don’t have mental capacity to research which jurisdiction works for this right now, but obviously not in US or EU.)
God I hope so. My workplace enforces Cursor use, and to see it lose GPT and Claude would absolutely make my day. Much has been invested in getting us to "use the right models", a list in which Grok does not feature.
Exactly. Without Claude and GPT, we'd be limited to Grok. A lot of ink has been spilled telling us exactly how to use Sonnet, Opus, Sol or Luna in certain scenarios, Grok hasn't ever been recommended by our pointy haired friends even once.
i'll be surprised if anthropic doesnt follow. cursors biggest advantage has been giving users a polished wrapper around whichever frontier model they prefer
Why you would trust Elon Musk to not hack you and steal your model weights is beyond me, but maybe they intend to use that compute just for training, still I’m sure there’s plenty of research there to take too…
what's the moat for AI labs like OpenAI and Antrhopic as they seek trillion IPOs. a lot of people said Chinese models also distill US LLM models. if it is this easy to do. how do US AI labs justify asking for trillion?
I believe if this was true, we'd already be seeing vibed stuff succeeding everywhere, and pricing out incumbents, but I guess I'm not seeing this? Like at all?
Even if they have Astra or whatever completely to themselves, how do they deal with the -product- side? Or sales, or account management?
They've been focused on juicing model's coding capabilities, but it's absolutely -not- "gen ai" enough to be doing the whole thing in agents, even 3 or 5 years from now, if only because there's so much context we _can't_ give to these models, without reverse centauring ourselves with cameras and mics and oodles of compute everywhere, and so far that hasn't exactly been playing out the way the frontier labs and the singularity folk hoped it would (meta glasses? humane? really?)
>we'd already be seeing vibed stuff succeeding everywhere, and pricing out incumbents, but I guess I'm not seeing this? Like at all?
As a small business owner I see this. Random people contacting me to sell a software where I can instantly tell it is vibe coded. Subscription prices half or 1/4th of incumbents. Domain names registered in the last few months.
They don’t need to make you or me, people who actually make software, believe that every anthropic employee is actually and in fact, right now, a „10x software factory controlling thousands of agents“, they need to make investors and CEOs believe, and then hope that making that true is possible and that they can keep the fiction up long enough and get enough money and infra built out to make it actually true.
Like Uber replaced all drivers with self driving cars [1] in its 20 years of operation? Just because I want to paint the moon pink, doesn't mean it's feasible.
[1] If anything, Waymo has a higher chance and global Waymo adoption is probably 20 years into the future, at least.
> I believe if this was true, we'd already be seeing vibed stuff succeeding everywhere, and pricing out incumbents, but I guess I'm not seeing this? Like at all?
I'm substituting my own[0] vibe coding for buying[1] apps. Language mini-games to help with German? A few prompts. Fluid dynamics simulation for an airzooker? Vibed. A web app listening for a MIDI keyboard, upon which you can drop some .midi files, and get a rhythm action game to learn the piano? Vibed. Webcam for my Raspberry Pi? Vibed. Getting Marathon 2 (well, the open sourced and upgraded engine, Aleph One) working as a web app? Vibed. Isochrone maps? Vibed.
Half of this I can even get done with the free models.
> without reverse centauring ourselves with cameras and mics and oodles of compute everywhere, and so far that hasn't exactly been playing out the way the frontier labs and the singularity folk hoped it would (meta glasses? humane? really?)
Yeah, so fortunate that cameras are expensive and there aren't 6 on my table right now between laptops and phones. :P
Seriously though, what's saving humanity collectively from everything getting automated from the panopticon we'd already built before Transformer models got good enough for even the most basic of classification and translation tasks, let alone anything we now use them for, is that machine learning takes an obscene number of examples before getting competent. Any living creature that needed so many examples would starve to death before learning how to eat.
This difficulty is why, for all the billions of miles that Tesla cars have collectively driven, perhaps pushing trillions now, they're still not sold to the public without steering wheels. Tesla claim to make such vehicles now in the form of the Cybercab, but they're not for sale, and even then some of the pictures that get in the press still show steering wheels.
[0] if you can call anything vibe-coded "my own"
[1] or worse, given the popularity of subscription models in this era, leasing some SaaS
> I'm substituting my own[0] vibe coding for buying[1] apps. Language mini-games to help with German? A few prompts. Fluid dynamics simulation for an airzooker? Vibed. A web app listening for a MIDI keyboard, upon which you can drop some .midi files, and get a rhythm action game to learn the piano? Vibed. Webcam for my Raspberry Pi? Vibed. Getting Marathon 2 (well, the open sourced and upgraded engine, Aleph One) working as a web app? Vibed. Isochrone maps? Vibed.
When do you have time for all of this? I don’t mean the vibing part but the using the app part.
I agree that agents are super good for one shotting throwaway code for tasks that would have required manual human actions previously but I would never have bothered buying an app for that.
Not having to deal with IT support for relatives is a win though since now I can just throw it at an LLM!
> When do you have time for all of this? I don’t mean the vibing part but the using the app part.
Mix of this being spread over more than a year, that I'm not doomscrolling because HackerNews and Telegram are my main social media presences, and being unemployed/prematurely retired (which one depends on what one thinks of €1k/month passive income and no rent).
I am also weirdly unmotivated by opportunities to spend money, which is both how I got this passive income and lack of rent, and why it's borderline enough for me. FIRE is very easy when all your working life, you only spend rent+50%, the rest of your paycheque going to savings and investments; but most people can't do this.
Oracle makes billions selling SQL when you can just use a free version. This is no different. Enterprise has enterprise needs. It's really not that complicated or irrational.
Anthropic wants you to think that models that distill from them would be worthless otherwise. It’s not true, it’s just one part of the process. The whole narrative that Chinese and other models are only good because they distill is nonsense.
Anthropic runs a significant portion of their compute on xAI hardware - why are you even talking about trying to suggest their relationship is acrimonious?
Some of you guys hate Musk so much that you say blatantly nonsensical stuff. Like, how can you be on HN and authoritatively mouthing off and not have some idea about the xAI-Anthropic compute contract, or the fact that xAI took down its own Grok compute capability so it could sell that compute to Anthropic? And if Anthropic hated Musk/xAI as much as you are saying they do, why would they be paying xAI billions for compute? Think/research before typing.
Is there some kind of alternative to Cursor in these terms
- fast because the code is already indexed no need to start every prompt rediscovering the world with rg
- in editor/vscode review much better than relying on git diff at the very end. You can click on next button to jump there. Quick edit in place. Or a quick prompt (seconds) to fix something small
- in editor completion / quick edits, when you do things by hand.
It seems that everyone loves agentic Claude code things these days but I don’t understand how you can review what it did and remain as much in the flow as you do with cursor.
Also I find that incredibly slow. By the time Claude finished a prompt I could have done a few in cursor.
And mind you, I’ve been wanting to use zed instead but the clunky vscode fork called cursor. But cursor has just a better UX for AI integration even though it’s slow and guzzle up RAM.
> It seems that everyone loves agentic Claude code things these days but I don’t understand how you can review what it did and remain as much in the flow as you do with cursor.
The way I review it is to let it finish a PR sized slice and then I review the whole diff in a separate terminal as if it’s a PR from someone else. Then I very impolitely list the shit that needs fixing, let the LLM grind out a new batch and repeat the process. When the code is good enough I throw it into a commit and push a PR out for other humans to review.
Clicking around in a UI would just waste time and if the change is too large to review in a terminal window the change is too large for another human to review.
There is a step missing. Impolitely list what's wrong and try to incorporate any persistent themes into a markdown file so that the agent stops making the same mistakes. Put any rules into linters and static analysis and sanitizers and ...
I sometimes see people reviewing LLM code leaving the same review comments multiple times, or worse, after some time they just give up reviewing.
More productive in what way? Letting agents burn tokens to produce garbage output is not productive. And letting teammates read code that wasn't reviewed by a human submitting it also isn't productive.
If I work with one agent / few subagents on one feature I can steer it as soon as I notice it drifting into the direction of waste. This way I only review the total diff 1.5-2 times. And I also don't waste my own mental energy on context switching between tasks agents are producing diffs for in parallel.
I don't think this is something anyone who has ever read a "thinking trace" would unironically say. Not that you can even see them in Claude.
When thinking first started and you would still see the whole "thinking process", I thought it was a ploy to 10x token use because it was just the most inane bullshit. "But wait, the user is asking me to" in loops.
You can see thinking summary transcripts in Claude Code and Desktop, and they are actually useful because they don't have those sort of thinking loops from the raw tokens.
I just switched back to VS Code from Cursor a week ago. Things seem great and VS Code has caught up. Last time I tried switching back was in November and VS Code was not ready then.
CLI is great for most things. Some times, when I am reviewing a PR, I just tell Codex to do a diff between source and target branches in PR, and generate a HTML report. Then I read that, and also review relevant pieces of code in Zed.
I tried using Cursor, many of my colleagues use it, but stopped using it due to clunkiness. And of late, even in there, many of my colleagues use agentic coding. Instead, I prefer the snapiness of CLI (Codex coded in rust, takes less memory per session) and Zed (also coded in rust,) and blazingly fast. Zed is the new Sublime Text if you will.
I just retried Zed and yeah, it's much much better than last time I gave it a shot!!!
I turned on the review stuff, and yeah I can do more or less what I could in Cursor. Check something. Do a quick inline promtp for the three lines I dislike and move on.
And of course the editor is so much nimbler/nicer.
No plan mode, but for big edits I'd probably use codex or something just like you. If you have any advice for integrating for that flow, I'll take them.
I guess whatever advantage cursor has, it will die once "Sol" is no longer integrated, as it was my favorite heavy work model. But I sure will miss the snappy auto/composer stack of Cursor.
I love VSCode built in LLM harness. No extensions, use any model you want, amazing UX for reviews, control of agents. The sandboxing also seems great, I never see models touching anything they shouldn't.
Claude / Codex ex tensions inside of VSCode give you what you described. I ageee, I find people that use only the CLI barely read and review the code the agent creates until it's too late
Lately I've been developing in Dart/Flutter. I use Android Studio but I know developers who use Antigravity and like it.
For the past three or four months it's performed very well, even the cheaper Gemini models. No serious complaints. But if you go back a year it was pretty terrible. My work style is to keep the coding agent on a short leash so YMMV.
Cursor seems to have intentionally dropped semantic/embedding-based codebase indexing. Their explanation is that newer coding agents are now good enough at searching repos directly. They can grep in parallel, inspect directories, read likely files, and refine the search themselves. Cursor says this now works as well or better for most cases. [1]
So the architecture shifted from something like “chunk → embed → vector DB → retrieve” to “local search index/ripgrep → agent explores.” There is still an index, but it’s a conventional local search index. No embeddings are involved in that retrieval path. [1]
The interesting part is that Cursor was still reporting a ~12.5% accuracy gain from semantic indexing in January 2026. [2] So this looks less like “semantic search never worked” and more like agentic search improving fast enough that the extra infrastructure stopped being worth it.
How does Zed compare to the tab completion in Cursor ("Cursor-tab")? It's honestly the only reason I have stuck with cursor in spite of the UI/UX being worse every single time I update it. Can I get something equivalent or better from Zed or some vscode fork/extension now? Admittedly I haven't looked in a bit, but it's seemed like nobody really cared about this workflow when I've tried in the past.
I do like Zed as an editor and I keep trying it out
but the the review what the AI did by jumping directly into the file with accept/reject/go next/next file with the ability to edit in place was nowhere near as good in zed vs cursor last I checked.
The features you mention are in Antigravity IDE v1, which I am trying as I still have a Gemini subscription.
No idea how long it will be around for, as Google Antigravity2 (more like Claude Code) is clearly the favoured way forward, plus the new Antigravity for VSCode Extension*, which I didn’t like as much when I tried it. It offers the same accept/reject hunk workflow, but I don’t know about editing in place. It doesn’t feel as if it has the tight integration of Antigravity IDE v1, which feels as if it was designed to challenge Cursor.
Having said that, much of the time I use Google Antigravity2 alongside a Jetbrains IDE. You can still review all diffs within Google Antigravity2 and attach comments to any line on order for a followup prompt to make amendments. But I still prefer the review and accept/reject hunk workflow.
There are things I will probably miss from Cursor, as they used to iterate on DX goodies quite quickly. For frontend tasks, I liked the idea of Cursor Browser being able to do a devtools-like selection of an element tree to be included in the chat context. I’ve been told there are similar extensions available for vanilla VSCode, so will have to see how good they are.
Interesting to see someone else is in the same boat as I am. Regarding editor review I agree, surely the solution would be to `git diff` at the end, but if you only want to git diff what was literally just changed by the agent it means you'd have to commit as you go so that your git diffs are more localised. And then at the end of your work just squash the commits. But it feels annoying. (EDIT: ok I forgot about git staging, that seems to be better)
I mainly use Cursor for another reason though: all of my work is remote over SSH and I don't like keystroke latency, so I prefer code editing / agent conversations to feel local. (Everything else is done in the terminal.) But I would also prefer to use something less propriety. I may go back to VSCode with plugins to achieve a similar experience.
Also: the ability to draft a plan and monitor its execution, review it, and merge it from a cell phone. I know Claude remote is a thing, but it’s not the same.
This is a big quality of life thing. Sure, I should just go on a walk without having an agent going, but oftentimes it’s not intruding on it - it’s the thing that actually ENABLES it.
> I don’t understand how you can review what it did and remain as much in the flow
Separate terminal tab with git diff/show, Zed for edits - rarely, there is little reason to change code yourself. Opus at low effort is faster than you can switch windows.
I met the founder of Voidleap (https://voidleap.com) yesterday! They have been building a new type of code editor since early this year and has been dogfooding it themselves.
I haven't tried it yet but it looks promising. He would be thrilled if people give it a go!
May be opencode2 in the near future. Sample (not affiliated) ~
1. download opencode2
2. settings > tabs > vertical
3. add worktree setup to agents.md
4. use /diff to review changes
5. use commands for commit, sync and push
I don’t see how opencode 2 (or 1) addresses any of my concerns. I thought it might avoid rg/grep the code base on every prompt but does not even seem to be the case.
I use codex all the time. It’s inferior to cursor for the two reasons mentioned. It rediscovers the world with rg (no indexing). Wall of diff at the end.
The GUI slightly better than Claude. And it’s a million times cheaper than Claude.
Yeah it’s pretty new but they’ll get there. I was using IntelliJ and basically wasn’t using it as an IDE. Then I tried Air and I was like “they get it”. But yeah they have work to do to get it out of preview. They are very responsive though
I used zed with codex for the accept / reject code review feature without relying on git but yesterday the Read Only drop-down disappeared and while I was doing my code review the AI agent started vibe coding without my permission. I hate this industry.
In the case of TUIs I’m going to assume you review the code externally in neovim or lazy git or something like that?
I don’t like relying on GIT for that and making things as good/done as I go. But with these agentic tools that’s what I have to do. Stage everything I like, manually edits a few bits.
But what do you when you find something where you want to say the agent: I don’t like this for reasons x and y redo as z. I’m like one shortcut away in context inside cursor to say that to my agent? You have to change tab in herdr and then what? Type the file name and line numbers? Is there a way to do this that does not seem very frustrating?
Also open code, codex, etc. super slow. They grep the code. Every. Single. Time. Cursor has a semantic index
I was looking at CLI-only solutions a while ago and the best thing I found was to use something like revdiff: https://github.com/umputun/revdiff
For instance, in Claude Code you can have it bring up this TUI overlay after every edit it makes (it literally is an overlay, rather than a separate tab in your terminal) and you can annotate line-by-line anything you want changed or rejected. Then once you exit the overlay those suggestions are automatically piped back into the agent.
I tried a few of these TUI-based code annotators, and this one is by far the best. It also works for files not under version control.
I'm currently not using it as I also prefer something more akin to "text editor with agent built in", but I wanted to mention here in case you were curious.
I usually copy paste snippet or function name that is affected and comment it. similar like you would sit next to coworker and tried to say where he/she messed up.
agent is able to get location correctly
I know this is change of context, but I have full screen terminal toggled with keyboard shortcut and with middle mouse paste (Linux) I do not have to move cursor anywhere
Guake and Yakuake have this always-on-top terminal overlay mode which is really nice for this kind of workflow. Works exceptionally well with multiple workspaces I prefer one single big monitor with everything maximized and multiple workspaces.
iterm2 on mac also has a similar mode, but it is a bit finicky to set it up (it is really hidden in the settings).
Thanks for the reply. That makes sense. It’s sort of what I do too when I use codex/claude/t3 code, but usually via GUI.
The in-editor context review is something that does not seem long for first world. Sadly. Maybe I’ll be forced to make a zed extension or something, maybe on top of jujustu…
It's always been clear to me that Cursor's business model of reselling others' APIs had its days numbered. Not necessarily because the providers would pull the plug, but because you wouldn't be able to compete with subsidized plans.
Cursor is already kind of useless for third party models unless you're willing to spend thousands of dollars. It's only worth it if you're going to use mostly grok/composer.
Why do people keep insisting that subscription plans are subsidized? They are not. API prices are outrageous, designed to milk enterprise users. They probably have an 80-90% profit margin on subscription plans.
Inference is cheap; Dax once said on a podcast that Opencode has a close to 90% profit margin on the openweight models it provides inference for (at 10x cheaper pricing). Those models are close in size and spec to models from large labs.
OpenAI leaked financial data (2025) suggests this - it has $5.7 billion for marketing which 44% of revenue. It’s hard to explain where these billions go other than into subsidised subscription plans and a free tier.
Because using chinese models on OpenRouter is way more expensive than the equivalent codex sub, unless you use something clearly inferior like deepseek flash.
The TUI version of Claude Code just isn't very good, though. The default keybindings do not work in the majority of terminals (because they generate the same byte sequences for key sequences assigned to different commands). Editing large prompts is rather painful because there doesn't seem to be in-prompt search commands. There is no source code browser, so source code references cannot checked by the users without a separate tool. Diffs cannot be copied directly because they are indented by spaces (even though some tools strip indentation from patches, it's still ugly). Switching between conversations is somewhat discouraged by the interface, which makes it harder to keep the context clean.
I think it's possible to have a better TUI, with something that isn't modeled after a shell prompt. Apparently, recent updates move in that direction. The GUI version of Claude Code seems to be coming to Linux, too.
The tide is turning though. I was using the old non agent view in cursor every day for about a year when I built my startup with it. Then Claude Code blew it out of the water. But since two month or so I much prefer Codex and Cursor with the agent view.
However, what I see cooking at Cursor is way more promising than the others. Things like full system understanding, their own forge, multi-repo support. I'd put them as way more visionary in the Gartner magic Quadrant ;)
Ingesting the whole codebase in their cloud and then maintaining indexes of where everything lives and what already exists. Not sure if that's where they are going but that's what I feel looking at some of the beta functionality and announcements.
I saw that happen culturally. But never saw that much different in performance. And I really dislike terminal development. And their desktop app is really buggy.
Out of interest when did you last use the desktop app? I am a huge daily user of it and apart from very minimal bugs, it works perfectly for me. Wasn't always the case, so they've definitely made improvements
Given you can use the same models on both, you're right about performance.
However I've found the cross-session context and memory in claude code much more helpful which can lead to better outcomes faster.
Additionally helpful where 2 apps/repos might require running at the same time, e.g. headless web apps, or, for plugin development where the plugin might be a dedicated repo but needs to run in another app to observe changes and make it re-test itself.
I also switched from terminal to the app not that long ago, I don't find it buggy, it has access to a browser which is really helpful.
> Claude code is a decent harness but then you have to use Anthropic models...
Claude Code works with models from other providers too. Anthropic supports this. You can configure some Claude Code environment variables to switch: eg changing ANTHROPIC_DEFAULT_HAIKU_MODEL to point to GLM Flash or Luna, setting ANTHROPIC_BASE_URL to point to api.z.ai, and making ANTHROPIC_AUTH_TOKEN the API key for your alternative provider instead.
That said, I've not actually tried this myself, opting to build my own harness instead. And you don't know what information Claude Code might be sending back to Anthropic about how you use competing models and which models you use. I don't know for sure that they do this, but after hearing about how they used steganography in the date of harness system prompts to identify the user's location, I don't entirely trust Claude Code anymore.
I think the point is that CC was the first general use, terminal based harness. I think. Now you can use PI, or OpenCode, or whatever, and use any model you like.
Now, I’m no expert, because I was late to the game and have only ever used Pi. But I guess Cursor is some product that’s tied to your IDE? If so, then yeah. That’s just too restrictive. I still love my IDE, but I don’t want it to be my harness too.
I've recently moved to a company that uses Cursor, and I'm actually quite fond of it.
I basically never use the editor but the fact that there's a review UI for all the agent work is incredibly helpful.
For me, that makes the models much more usable. I've also been using GPT models a lot, as they're cheaper and less vomit inducing than Claudes text, so this is definitely bad news for me.
cursor adds $0.25/M to your token bill for 3rd party services. sounds small but it's insanely big on cached inputs, which are an insanely high % of use
If the Opus/Fable harness is superior and it's written using Opus/Fable then it will cease to be a moat. Otherwise the premise for coding using LLMs is essentially untrue.
Honestly, Cursor's AI harness worked best with the Anthropic models. Could just be my bias though. Evaluating frontier LLM's all seem to be by vibes. I think the industry is understanding it's both the harness and the model that makes the difference; can't half ass either, and some harnesses do better with different models. I'm good to leave this "career." Senior software engineers, make your bag.
On one hand, yes, it doesn't make sense to give out subsidized tokens forever.
On the other hand, subscriptions create lock-in in a way that API pricing doesn't.
I think a more likely end is that subscription value decreases over time because API pricing gets more reasonable, but subscriptions stay because they are a good way of getting money out of people consistently.
Right, subscriptions are not only cheaper because they "create lock-in". They also let you do things like forecast demand and plan your capacity for it, borrow against it sometimes if you need to, and keep somewhat less around because your cashflows are predictable. It is also valuable because it keeps customers around but that's by no means the only reason recurring revenue gets a higher multiple.
This is why "subsidized tokens" is possibly a misnomer. Money at lower variance is worth more than the same money at higher variance. Not "subsidy" so much as reducing risk and passing some of that to a consumer.
What do you mean by that? If you're being sassy about a downvote then I'll remind you that when you reply to a comment, that person can't downvote your reply. Also this is an account with barely any comments or points, I'm pretty sure they can't downvote at all.
I may be a rare Cursor user around here, but I am saddened by this news. The thing that keeps me using it is precisely the ability to switch between models: I usually do brainstorm/planning with Sol or Terra, then let Composer (free) deal with the implementation. For certain tasks I use Claude, which tends to be more expensive but sometimes saves me time.
Using both OpenAI and Anthropic models, Cursor is cheaper than having two separate subscriptions, and their free tier of Composer is good enough. I will now have to reconsider my options.
JetBrains is a good option - works with any model/agent. Also they have very good semantic search for the codebase, which is exposed via MCP to the agents.
currenly I use pi with openrouter and is very nice I combine with a chatgpt sub that allows me to use it with pi. and when my usage runs out I turn to openrouter. Then same when I need anthropic models
If you can explain how to make compaction not suck in opencode or Kilo Code, I would love to know how. I am really trying to make use of other harnesses but that's a major sticking point.
Never compact. Not automatically, not manually. Instead, let it write ledgers and checklists or handover prompt if necessary, and start fresh sessions as often as possible.
Create a design spec and implementation plan first. If necessary, split that in phases. Then create a ledger that always holds the current status, and instruct the model to update that after it completes a task in the plan, and to append any issues it encounters during the implementation.
Then you can let it run, and when it stops, point a fresh session to the ledger to pick up the work from where the previous one left off.
You can get more fancy by using something like GitHub issues for coordination too and have multiple sessions create issues and epics and others to work on that in isolated worktrees, but the spec-plan-ledger approach usually works great for small to medium large projects.
While this isn't great for users it seems like a pretty standard circling of the wagons in preparation for the next phase in the battle for frontier AI dominance.
Ironically, as a Cursor and Claude subscriber, but not an OpenAI subscriber, this will push me back to Anthropic. I don't feel good about giving OpenAI money directly, but I've found switching to GPT 5.6 Sol now and again in Cursor to be useful. Now I'll probably just forget about OpenAI models.
Yeah I haven’t had a problem yet which was too difficult for Fable to solve. I do have problems which Sol does not do well on. But if I could get Fable-level intelligence faster/cheaper and less restricted, that’s all I need.
Fable is great minus it's annoying writing style. It's nowhere near as bad as Opus 5 but it's very "AI trope-y" even with heavy steering. I can spot the writing immediately.
Yeah not sure why everyone thinks they need a PhD lvl assitant to build their slop SaaS when an open source model that is 100x cheaper, and 3x faster and arguably still "PhD" level assitant could also build their slop SaaS.
Exactly. Astra will likely get its Mythos moment, and then another Chinese lab releases its equivalent three months later while Anthropic releases a newer, better model.
There's a reason why most businesses don't hire PhDs, they dont need it. PhDs are cheap in many countries, arguably more affordable than a full time mythos lvl agent, yet businesses in India still don't hire 30k a year PhDs to run their businesses. Its simply not necessary. AI labs have lost the plot from a business perspective. If chasing ASI is the goal, that also seems stupid.. Putting all of your customers out of business with ASI doesn't really lead to great outcomes either.
The higher level models are designed to tackle bigger problems with less oversight and more accuracy. It’s about entirely changing how you use them to mirror an employee you can trust. A truly good team members everyone trusts to just get things done at high quality with minimal direction for days or weeks. That is the goal.
It’s OpenAI (and Anthropic I’m sure) realising that they’ve somehow ended up stuck in the lowest margin part of the stack (providing inference), and trying to get out of it
Honestly, I've been a Cursor subscriber for a while and it was just momentum. Still, I rank OpenAI as the worst actor among the big AI companies. I think OpenAI's playbook will be a repeat of Facebook. Users are the product x 1000.
Cursor/SpaceX/Grok is a different beast. I think they will cheat other companies but I'm not convinced yet that they have it out for users. We'll see. I'm certainly not some kind of Cursor lifer or something.
It's my personal view that Elon has already shown how little he cares for protecting data of individuals based on how he ran DOGE.
If his teams will handle the data of US citizens with such disregard, he'll almost certainly treat his customers (many of whom are those same US citizens) the same.
I'm not keen on OpenCode long term after seeing their CEO engage in antics and their discord being rather unprofessional, but it is a great harness for the time being, still has gaps though
> as a Cursor and Claude subscriber, but not an OpenAI subscriber, this will push me back to Anthropic
Instead of deciding based on the merits of the model, they are being pushed by the actions of the companies. You should be able to use the best model, or whatever model you like, independently of what any company thinks or does.
They are deciding on the merits of the model right? They’re picking the model that’s supported in most tools.
I think maybe you’re suggesting private companies should be forced to support all models which is a position that isn’t technically feasible or one that most would agree is even a sensible ideal.
No, companies can do whatever they want. They shouldn't be forced to support all models.
I'm suggesting you shouldn't let a company dictate which models you use. Putting yourself in a position where they can is detrimental, because you miss out on using the best model for the job.
I don't see how it makes sense to use models outside of Grok or Composer in cursor unless you like the tooling to such degree that you're willing to pay out the nose for access to OpenAI and Anthropic to use with that tooling. And Codex is regularly scoring towards the top among harnesses, so you might as well use the best harness available for the given model. I'm happy with Grok and Composer in Cursor. If Cursor wants to add more models, they should host more open models.
The reason it makes sense is because different models are better at different things. Allowing yourself to be locked in to a particular model by these companies is a bad idea and will lead to worse outcomes for yourself and the market as a whole.
You’d use it because until recently Cursor had the best setup for cloud agents and visual verification and a bunch of other features outside of the harness.
I can’t speak to the veracity of the benchmarks but it appears their methodology is sound. Nanobot has 47k stars, fwiw https://github.com/HKUDS/nanobot
It has been more of an OpenClaw or Hermes alternative than a coding agent like OpenCode or Pi, so it’s likely to do well given less context bloat.
What a shame for Cursor. I found it a great tool - I actually liked Composer 2.5; but now being bought by SpaceX has for a while left a bad taste - and they keep pushing Grok and now GrokBot.
I took the 50% discount instead of cancelling - but with this I'll likely not renew.
So sad to watch them go from great to mediocre in such a short time.
I use Cursor at work and love it. It's a great IDE, the best IMHO. But it becomes really expensive.
Since the introduction of the Cursor Token Rate, my company tells us to use Claude and Codex directly with Anthropic (or Vertex AI) and OpenAI, respectively.
I'm now using Grok 4.6 most of the time. Sonnet some times. But that's it. Cursor-native models 99% of the time. I'm pretty sure Anthropic will follow OpenAI soon in pulling their models out of Cursor because it's now a true competitor.
It's insane that Cursor charges 25 extra cents per million tokens on a model that charges 2 cents per million cached tokens. Basically everyone is forced to use Cursor's own models at this point as otherwise the costs are simply too high.
If you have such a world changing technology, why sell an API when you can capture all the value yourself? That's what every provider is asking themselves.
Cursor will be fine after this, as they have their own first party models and can potentially use open weight models via API or their own hardware, which have gotten good enough to Opus and Fable quality now for many use cases.
Being a pure play model provider is completely untenable right now. Coding agents own your demand; Nvidia, data centers, and DRAM manufacturers own your supply. You're getting squeezed by open weight models. Your only bet to making money off this thing is by locking your model down to your program to have some semblance of a moat. The fact that businesses can change providers with just the change of an API key is unsustainable. You have neither the "hard to switch" lock-in of a traditional software company nor do you have the near-zero marginal cost of production they have. Each token of inference costs you money, not to mention the training costs. At least locking the models down to your harness lets you get some sort of lock-in.
Lots of enterprises don’t want to use OpenAI or Anthropic directly. They will go via AWS bedrock or google’s equivalent.
The model providers are in a difficult position. With an added danger of open weight models that are within reach of large enterprises running themselves.
Not really a surprise given their own models are the default option? Most people wouldn't change model. And if you want to use Claude or GPT you don't really have a reason to use Cursor over Claude Code or Codex
> As AI capabilities advance, we also have a new level of accountability to ensure our upcoming model, Astra, is being used in accordance with our terms.
I think Sam Altman has played a clever business move here, they can use this excuse immediately after the Hugging Face leaks and it's not going to be dismissed willy-nilly. Whereas long term they probably wanted to end their partnership with Cursor anyway. I bet Dario will be happy about this too, but maybe the compute deal prevents him from doing similar anytime soon.
Is it really an excuse? As the release points out, musk admitted under oath to distilling OpenAI’s models to produce xai models on an ongoing basis. Whatever you think about goose for the gander wrt IP theft, I wouldn’t voluntarily do business with a competitor who was actively and brazenly flouting my terms, would you?
It's a little confusing because Grok and Cursor kinda live in different worlds. Grok has SuperGrok and SuperGrok Heavy ($300 a month!)
The best bet is Cursor Ultra ($200), which is the equivalent a month plan that you get from OpenAI and Anthropic. I get a LOT more capacity out of it because I almost always use the "Auto" mode, which auto picks between Opus 5, GPT 5.6, Grok 4.6, and Composer 2.5.
I like Auto mode so much actually. I use it for my Cursor work account as well because it's about 10x faster than just using Opus 5 or GPT 5.6 and about 90% is good. I switch back to the bigger models if I get in trouble, but that's pretty rare.
Vibe coding requires some discipline. You need understand what it's trying to do. If you have a good mental model you can keep it out of trouble. If you don't, it doesn't matter how good the model is. You're gonna end up in a mess.
Yes, any Cursor sub includes a separate tier of usage for Cursor/SpaceX first party models separate from usage of third party models.
I have a $20 Cursor sub still personally and would say that the usage for Grok models is much higher than what $20 gets you with Anthropic. I can't speak for the $100 plan but I would assume it scales similarly.
I rarely run out of my usage monthly even with the $20 plan.
I'm on the $20 plan too and the only time I run over is if I'm using Opus (or probably other third-party models). Grok and Composer seem nearly free at this point. Cursor also gave me a $100 credit for Grok after 81% of my usage was on TP models. I don't think Grok 4.6 is quite at the level of Opus 5, but it's close and so much faster (and less chatty).
Yes it does. You can either subscribe to grok directly or cursor.
You have many tiers and all of them have generous usage.
Grok is a workhorse, not the smartest yet, but produces good code and follows the plan.
Grok 5, which has been training since beginnig of this year, will probably be a monster. (10T/6T Param model variants)
It makes no sense that a contract would even provide a maximum notice period. Why would parties put that in? What benefit would either derive from restricting themselves to not getting earlier notice.
I read it as: we acted as quickly as we could once it became possible to do so, i.e., the change in control was completed. not that their special agreement carved out some maximum notice term.
Everyone is playing hardball. No one wants to be IBM in an IBM-Microsoft dance.
But this is kind of like banning GPU exports to China. It will encourage them to develop their own technology. Yes, it will take them longer, but in the long-run China will be more successful with their own GPU tech.
ITAR: SpaceX is strictly barred by US federal law from hiring non-US persons for aerospace engineering.
National Security Scrutiny: xAI operates under increasing US export and tech-transfer controls regarding China.
Scale: A private company's recruiting budget cannot compete with the sovereign industrial capacity and state-backed talent pipeline of an entire nation.
> As AI capabilities advance, we also have a new level of accountability to ensure our upcoming model, Astra, is being used in accordance with our terms.
This move seems to protect Astra against distillation. I wonder whether Astra will be released before or after 11/12.
I think there is still a huge market for models that are specialised in software development (system design, infrastructure, coding, etc.).
If a company can provide the access to such a model via AWS Bedrock or similar enterprise compliant platform with a decent IDE/harness they can capture billions of dollars of market.
The closest we have to this is Claude Code but I think it can even be better and specialised. Cursor has an enormous amount of data on coding and developer usage and maybe they should use that to build a proper platform
I really get sick of this anti-Elon rhetoric around here. Can we not keep it to technical merits and not get sucked up into politics? The guy's clearly nuts, but he's also done an incredible amount of amazing things.
I trust them more than someone who took a nonprofit designed to be open and had tens of millions in donations to do just that, and then turned into a for profit who's cuddled up tight with Microsoft and is so closed they're preemptively ending contracts with someone under pretty flimsy accusations. And doubly funny it's the person who's responsible for them even existing. And from whom they stole tens of millions
I'm not impressed with cursor and it's decidedly not worth 1 mil in terms of its engineering. Perhaps it's user base and their loyalty makes up the vast majority of its evaluation. But to me it's ridiculous that a dev would use cursor or think that it's indispensable to them
> We know that the people most affected by this decision are the developers who rely on OpenAI models in Cursor. We care about their experience in this transition and we’re ready to go above and beyond to support them.
In a sane world would this be a potential antitrust case as they are a direct competitor to Codex?
How can you ban a competitor from using a model that you openly provide access to?
I understand that the OpenAI ToS may have been violated by X, but has anyone ruled on distillation as fair use? OpenAI believes distillation of nonAI produced content is fair use, so they don’t appear to have a leg to stand on.
"You are not allowed to use our services to develop a competing product" is a standard clause in pretty much every contract of this nature. There's nothing antitrust about it. Companies are allowed to say no to bad deals.
In a sane world, one should be able to choose not to do business with a company for any reason
there are plenty of alternatives, 100s, so I don't think antitrust is on the table, think of this more like a boycott of the world's richest man for being a twat
you dont know me, so I don't know on what basis you make this claim, boycotting Elon and his companies is a real thing people do for ethical reasons, see Tesla sales as a data point
It was honestly just a matter of time, given all of the bad blood there is between Sam and Elon and Elon basically admitting under oath they were distilling other companies models.
I wonder if Anthropic will follow suit given their campaign against distillation or if they'll stay quiet given their reliance on SpaceX (among others) for hardware.
Honestly, can't say I miss my Cursor subscription at all, and I'm surprised they're still a going concern. Why would I want to use a proprietary VSCode fork with an identity crisis when I can use literally any editor with any agent and have about the same experience?
I use Claude Code (both Desktop and CLI) all day every day (as well as OpenCode with DeepSeek V4 among others). For non-mission-critical stuff (small internal tools, anything that doesn't touch the client and doesn't handle critical data or infra), I'm on Auto (sandboxed in the working directory). For everything else I approve every change, meaning I review every code change the model proposes. Plan mode is great to write the architecture as well, as it doesn't touch any file, doesn't write anything, and lets me iterate until the arch is perfect and the proposed high-level code (modules etc) is solid. I haven't used an editor in over 4 years, apart from writing specs.
How do you review the code? I can use git to see changes, but it's super annoying to tell the agent file line plus comment for each change, so I catch less stuff early which seems to compound.
We have a subscription at work. They have a CLI agent that works pretty well. Lately I mostly use it with grok, which is actually quite a capable model.
How so? A dozen very capable models out there. And the gap between frontier models from the two main US labs kept narrowing, at this point non existent.
The quality/price ratio is clearly superior when running Qwen, GLM, or Deepseek.
To make a product more exclusive won't hide the fact that product is becoming more and more irrelevant.
Most people do not enjoy the output of the open models or know where their code is being sent so I’m out. Also I don’t trust Elon anyway, I’m just going to make a VS Code Extension that handles things how I want but I’d like to offer the same deals Cursor got for tokens with OpenAI and Anthropic, for me this was the key 1/4 token cost!
This is the platform risk of building on rented intelligence. Cursor’s moat now has to be its harness and true model portability—not privileged access to any one AI lab.
Musk was forced to admit under oath during his OpenAI lawsuit that xAI distilled GPT models to train Grok. No reason to believe he won't use Cursor data for the same thing, if it isn't happening already.
Not surprised, but honestly I don't feel like this is a huge loss for cursor. Models are so commoditized, and 99% of users don't need particularly fancy models. Between existing anthropic models, composer-2, honestly even grok models, and also the fact that many people just use 'auto' mode, I feel like many won't feel the impact for OpenAI's withdrawal.
Maybe if Claude joined along it would be more impactful. For better or worse though, Xai seems like its lining themselves up to handle being on their own for cursor.
My guess is that OpenAI now has a full blown Cursor equivalent in development (an expansion of Codex?) and will announce it in September. Otherwise it stands to lose those who are used to working with Cursor.
Yeah, the "go above and beyond for our users" implies they have a "better" version of Cursor coming. Probably integrating their new chip/microphone based product.
Great. Sam Altman sold us all out to venture capital. So I won’t be using ChatGPT anymore. Also, I noticed that ChatGPT lies more than any other AI of all the ones I’ve tested.
I used cursor extensively both at work and for personal use. Its harness is pretty darn good but I basically never used OpenAI models in it.
The first party cursor models (initially composer and now grok-cursor) and price very competitively for their quality and have very generous usage limits included in the base plans, and then per token pricing afterwards.
Occasionally I’ve used some of the Anthropic models in it but found their cost massively outweighed the benefits for the type of usage we do.
I was using Sonnet 4.5 and 4.6 for a while but the newer models are a lot more expensive at the per-token pricing.
Cursor does currently allow users to BYO OpenAI token. Given Cursor won't have access to OpenAI models for prompt engineering/evals though, there's definitely a question of whether they'd allow picking any new OpenAI models, assuming the BYOK option continues to exist at all.
Email to customers now "We’re speaking with the OpenAI team to resolve this, and we will share more in the coming weeks around continuity of access to OpenAI models. We’ll continue to offer ways to access all models in Cursor.
All it shows is that openai is likely still subsidizing inference, sees no value in integration buried within other branded tooling competing with theirs.
I used mostly OpenAI for the last like 6 months in cursor. I got tired of how slow Claude models are and switched. I’m surprised usage is so low. Wonder why.
Because you can also use grok and a dozen other models. In fact grok is preferred choice for cursor right now, so obviously the other models get sidelined
Spend a lot of time thinking about why individuals shouldn't trust large AI cos, now it's interesting getting to think about how the AI companies themselves don't trust each other. Capitalism is viciously cold and uncaring.
Not sure the larger point you're trying to make (maybe we're on the same page?), but it's a fact that AI has come from heavy long-term investments by the academic and public sectors, yet private companies are accruing most of the gains
back when the USSR was still a going concern, there were zero American leftists claiming that it wasn't really communist. that only started happening after they failed
> We are making this choice because we cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk's companies violating contracts.
I'm surprised how clear they are about this.
Related:
Can highly recommend Zed as an alternative, it's better than VSCode with the Codex extension as managing multiple chats is much nicer in Zed.
At my dayjob we are big users of Cursor's Cloud Agents. We use them to trigger automated investigations customer support issues, alarms, feature requests etc. The Cursor editor is fairly meh in comparison, but this has become a big part of our workflow. And having anyone on our team manage them. What competing products are there we should try?
Devin Automations is the most obvious one: solid UX, but can get very expensive to run.
There are also a bunch of newer startups in this space. I'm building one myself, https://boxes.dev -- we're very early but building for this exact use case. Some other ones worth a look are Factory Droid and Amp Orbs. Those two build their own agent harness (like Cursor), whereas with boxes.dev we run the native codex and claude code harnesses directly.
Elon Musk openly admitted to distilling OpenAI's models [1], which I think explains the "based on our experience with Elon Musk's companies violating contracts".
Last year, Anthropic cut off Windsurf the same way [2]. Fortunately, unlike Anthropic, OpenAI allows you to use your subscription with other harnesses, including Cursor.
Of course today, Anthropic is taking the "high road" and is OK with with SpaceX [3] including their models in Cursor, because they need SpaceX's compute.
More generally, Musk admittedly works with a "move fast and break things" mentality. His companies have repeatedly broken environmental laws, labor laws, and securities laws. His companies often fail to deliver the features they promise and routinely miss deadlines they've committed to. They routinely violate contracts that they have agreed to.
Not doing business with them is a completely normal and unremarkable decision.
> We are making this choice because we cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk's companies violating contracts.
> After Musk acquired Twitter, now part of SpaceX, the company broke (opens in a new window) the terms of our contract (alongside many others).
I have never seen one large company throw another large company under the bus without any legalese filters like this before. Wondering if OpenAI site got pwned. Or maybe we in some strange, alternative reality. I doubt even their LLMs would be this frank.
When I was bringing up IDEs at work and advocating JetBrains (I’ve since built my own … ) a colleague questioned why we don’t just use cursor instead of waiting for JetBrains Air to come out.
The products are basically equivalent but JetBrains is a profitable bootstrapped company that has been building IDEs for 20 years and Cursor is a flash in the pan startup with a squillion dollar valuation to support. Sooner or later they’re going to do something that will fuck their users over.
The way I read the article was that they will continue providing their current models, but won't let them use any newer ones announced even between now and the end date. But I might be wrong.
> we’ve decided to hold the contract cancellation to the latest date we can while not providing future models to Cursor.
Cursor is like one of the least economical ways to use Claude and OpenAI, virtually nobody does it. You use Cursor for their autocomplete/composer + grok. For using Claude and Codex, you get immeasurably more bang for your buck just using them directly.
All good points. However, it's not a safe assumption that everyone operates that way. People probably use the flagship models because that's what they know. Sure, some probably just use the default, in this case Grok.
Also, it's a signal more than anything. OpenAI pulls out of Cursor. Then Anthropic and Google does as well. Then you no longer have "Cursor" you have Grok.
Grok doesn't have the best reputation in general, regardless if you're tech savvy or not.
Good! I don't consider myself a Nazi and so I'm happy a big, influential tech company refuses to business with a man who does the Nazi salute on television.
i don't see the point of cursor when vsc + codex extension does the same thing. i don't use autocomplete anymore (who writes code anymore? lol) any and anyone using grok unironically is also probably a moron.
i only tell VC type people or others from whom i want something "oh yeah grok is cool" if they mention it, because they're always sucking off Elon
We are making this choice because we cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk's companies violating contracts.
The pot calling the kettle black. Look who's talking. A company that trained it's models illegally on so many copyrighted materials without compensating anyone involved.
For a company named "Open" AI, they've been hardly open at all and the slew of resignations of top tier execs in the company is really the proof.
Make no mistake, this isn't a moral decision. This is just killing the competition using morality as an excuse.
Codex cli is open source rust project (Claude Code is not). Whisper is an open weight ASR/STT model. GPT-OSS is an open weight LLM that can run locally and is still very capable. CLIP is an open source model used for image encoding. Just some of the open contributions from OpenAI
this is a pretty clean test of whether people were buying cursor or just using it for convenient access to whichever model was flavor of the week.
i work in qa so i'm naturally wary of adding another critical dependency to a development workflow, if the model you actually want can just disappear because two vendor change ownership or start argueing then the wrapper has to be really good on its own.
cursors harness and its own models may be good enough that users stay, and if they do cursor has a moat at least. if everyone follows the openai models out the door, it will just become a polished ui.
I think it’s important not to let AI companies control the framing here. This is anti-competitive behavior, plain and simple. The fact that xAI/Musk routinely break the law simply provides the fig leaf.
Distillation is a threat to frontier model companies, but that doesn’t make it dangerous or wrong. OpenAI can (and likely will) use the same logic to ban you from using open source harnesses.
I’m on team fuck both of these organizations, but not continuing to work with an organization that already willing violated contracts you had with them doesn’t exactly seem like a hard to justify decision.
So this is OpenAI following suit after Musk admitted to distilling their models.
This was bound to happen once Cursor decided to sell itself to a competing model provider.
It'll be interesting to see if Anthropic applies their ban to Cursor, or if that datacenter deal they signed with Musk changes anything.
If you can take any book and turn it into a model, because it's "transformative enough", and "AI learns just like a person does", then surely a model distilling another model is transformative and fair use.
They tied themselves into knots fighting the letter of the law, and now, when they need the spirit of the law - that each creator deserves protection for their work - now we devolve to the law of the jungle. Maybe we'll even see LLM book curses, the way medieval scribes damned book thieves to blindness and worms.
This is not a necessary end state. It is the byproduct of the disease of sociopathic MBAs.
(Source, worked for multiple tech companies that were laser focused on delighting customers until money people came in and ruined it, to the point they would prefer devs sit idle than work on things the MBAs didn’t have on a priority list).
For the record, with the exception of Amazon’s Jassy, the CEOs of the top 5 companies are all engineering types with engineering credentials.
The CEO credentials of the top 5 AI companies are all science degrees.
Just look at the pattern... they collude, they provide to them whatever is needed, and at some point, if this is not true yet, they will be the ones who will tell them what to do or not, bc you know how humans are, right... blackmailing, mess up the business or shames of people, etc.
It is just a matter of time. The state, as we know it, will collapse or will be greatly reduced, which, from a point of view, is positive, from another, Idk, bc if someone replaces that, we will be in the hands of someone, as usual...
I'm curious if we'll ever get a form of civil war where corporate militias (corporate _robot_ militias in the case of xAI, of course) draw fire on good old meat policemen coming to bring the CEO to court.
Sure, it may happen, but I suspect the alternatives (bribing, sending a scapegoat to jail, buying the elections, etc... and focus on the "making money" part) will stay preferable for a while.
This has already been broken, and they've become so much more bold since then.
"For the general election of 5 March 1933, the Nazis were allied with other nationalist and conservative factions. At a secret meeting on 20 February, major German industrialists had agreed to finance the Nazis' election campaign."
https://en.wikipedia.org/wiki/Enabling_Act_of_1933
So much for the all-powerful cabal.
Why fight someone who is so open to corruption? The important thing is that they got everything they wanted, and all they had to do was throw a few million at the clown, attend his parties, and maybe take down some diversity programs they didn't believe in anyway.
China is a different beast altogether, though.
Because they know it will work. You don’t have to watch Trump speak for very long to know he’s not very intelligent. Similarly, you don’t have to be much smarter than him to know how easily he can be manipulated with adulation.
Lmao
Doubt.
The networking between people in power at the top will, in my opinion, most likely assure that they remain in power, because at the end of the day that is what they want most.
I expect government will seize control of the most advanced models (if they haven't already), and the rest of us will be throttled, and status quo will be maintained. I do not expect that either super advanced agents or the owners of the hardware they run on will be able to pull off the kind of coup you describe. At all.
Who do you think will rule at that point? They just cannot fight that, they do not have the technology these companies have.
Usually, the number of bodies who will comply is the proxy for power, but with automation, it could be different in the near future.
https://xkcd.com/538/
Regulation alone cannot control if it cannot be made effective.
Imagine the United States DoD, CIA, NSA training an LLM on all its top secret intel.
Oh my sweet summer child...
This has been the continual goal of the US defense and intelligence agencies since 9/11, which was blamed on a lack of information sharing. Why do you think Edward Snowden had access to everything? Because it was consolidated post-2001.
You think there is any access control at the upper levels of the dark intelligence agencies? They have access to the whole take.
And it used to be that was pretty useless unless you had an actual lead. You'd need a East German Stasi level of labor to read everyone's secrets at scale... Now you don't, thanks to AI.
https://enwp.org/Total_Information_Awareness
Guy this was openly stated more than a decade and half ago by all the current oligarchs
Yes it is, in the legal/copyright sense of fair use. That's why they ban it in their TOS. Which customers agree to when signing up for the service.
So… my TOS doesn’t matter, but theirs does?
Wilhoit Conservatism: In-groups protected by contract law but not bound by it, alongside out-groups bound by contract law, but not protected by it.
1. Copyright violations (can put you in jail) 2. TOS violations (will be a fine at worst)
Companies do get away with drive-by legal shittiness way too often and frankly the practice needs to be reined in, but at the end of the day the only damages are the financial ones you can prove in court.
Incidentally, this suggests that once an LLM is capable of accessing and distilling a competitor's LLM without human intervention, then any legal argument about TOS violation is moot. But somehow I doubt that will fly in court.
The GPL established this rather clearly. Copyright law doesn’t require consideration.
(The licence itself is a basic BSD licence, so it just requires attribution including in marketing materials, which obviously hasn’t happened.)
Within this context, I don’t think so. I can’t make a website that buries some shrink wrap that requires everyone who reads it become vegan.
You accept it. You pay consideration for it. If your website has a TOS dickover, that requires someone attest with their legal name and pay you $1, yes, it may be enforceable under some circumstances.
Asking for all Anthropic employees who dream big.
Copyright (C) 1998, 2015 by Greg Egan
First Night Shade Books edition 2015
All rights reserved. No part of this book may be reproduced in any manner without the express written consent of the publisher, blah blah (it felt very ironic to transcribe that bit in particular to make this point)
It's possible that distillation for other reasons, with no intent to harm the vendor you distill from, would have been ruled to be fair use. But in law, intent matters.
Also, copyright has always been privately enforced anyway?
On the third hand, the settlement Anthropic famously had to pay was for copyright infringement because they didn't actually have the right to even access some of the training data they used, so I can see how this might be compatible with the law.
On the fourth hand, I'm saying that as someone who absolutely isn't a lawyer and sometimes gets surprised when reading about copyright cases that sure sound like they ought to have been trademark cases given my limited understanding.
For those who chose to give it all away, the ship has sailed, but they did choose to give it away for free to anyone so they can't complain that they succeeded.
Enforcement ultimately happens through law and the legal system.
Otherwise, it's just a standard case between two private parties resolved through our legal system; e.g. Linkedin vs Hi5.
This is a gross distortion. Standard contractual rules bind the parties that signed the contract and the remedies are proportional to the damages and bounded. Copyright is tort law, the state binds the world to respect the rights of creators and the damages on infringement are punitive and can far exceed the actual commercial damages - to the point of bankrupting the infringer.
The key to torts is that the state is not neutral, there is a social good here it's protecting. Crucially, copyright, like some other torts - securities, antitrust, environmental, battery - also has a criminal enforcement regime, where, for particularly serious offenses, the state actually invests public resources to put the criminal infringer behind bars with little to no involvement from the original rights holders.
In the particular case of US, there is an entire state apparatus dedicated to enforcing US copyrights, a foreign affairs policy to shutdown "Notorious markets for counterfeiting and piracy" in other countries, international enforcement of DMCA etc.
The idea that a private TOS has the same level of public protection as copyright is downright childish.
I mean it quite likely is sort of in the same legal bucket. It won’t stop them suing but it is going to be the legal equivalent of two biologically-related warlords making their champions fight with their hands tied for sport.
Big models are built by scraping the recorded thoughts of everyone, so giving everyone a chance to run a distilled small model is just going full circle.
Obviously the underlying motivations aren't 100% altruistic, but I'll still take it.
ChatGPT 3 was released nearly six years ago, and the models are staging increasingly aggressive breakouts now. Where are they going to be by 2030?
No, the AI companies merely figured out a way to spin gross negligence into a PR win. Any idiot can build a Murderbot which "goes rogue" - it can be as simple as taping a knife to a Roomba. The harm it does is not in any way related to its "intelligence" or "sentience".
We're seeing "breakouts" because the AI companies are being rewarded for their incompetence. You don't have incredibly lax security standards and zero form of oversight resulting in fully-automated felonies which should result in jail time, you instead have a "powerful near-sentient cybersecurity model" and should be given hundreds of billions of dollars!
That thread you linked suggested they were using it with cursor as part of their development process, not for distilling.
>"Hi team, I believe many of you have already discovered that anthropic models are not responding on cursor. According to cursor this is a new policy anthropic is enforcing for all its major competitors."
And a month before the OpenAI block, they blocked Windsurf for the same reason: https://www.forbes.com/sites/johanmoreno/2025/06/05/anthropi...
So its pretty clear the TOS had this clause prior to Jan 2026.
They indiacte their development staff are using it, and they access it via API, then then go on to say all the nefarious things they "could" be doing, stopping before stating they actually are.
Specifically the TOS violation was:
>customers are barred from using the service to “build a competing product or service"
So if you are building a competitive product, that alone is a TOS violation from their perspective.
Which is fine and their prerogative, but we shouldn't cast a dark cloud suggesting other violations were happening without stated evidence.
OpenAI publicly stated they can't "trust that xAI will use it within terms" not even that they've violated, not that there has been a violation from the Cursor team.
https://www.teslarati.com/openai-cites-distrust-of-spacex-in...
>OpenAI said the decision came down to trust, not technology. “We cannot be confident that SpaceX will use our technology within our terms of service"
Its actually very tone deaf to even pipe up and say anything about this given they did this already to Windsurf themselves, and they previously banned x.ai employees as well from using their models for basically the same thing Elon admitted they were doing with OpenAI.
I'm surprised they decided to say anything at all given their rocky past. Its the pinnacle of being a hypocrite.
Just go ahead and do set it up. You’ll be my hero (I don’t have mental capacity to research which jurisdiction works for this right now, but obviously not in US or EU.)
This is a battle for power, let us not be naive...
It's hard to imagine a 1B / month deal wouldn't.
what's the moat for AI labs like OpenAI and Antrhopic as they seek trillion IPOs. a lot of people said Chinese models also distill US LLM models. if it is this easy to do. how do US AI labs justify asking for trillion?
Otherwise, it will be competitors distilling, and the government getting upset.
It's hard for me to see a different future.
And, this really sucks.
If this thesis is true, I wonder what the final YC batch number might be.
Even if they have Astra or whatever completely to themselves, how do they deal with the -product- side? Or sales, or account management?
They've been focused on juicing model's coding capabilities, but it's absolutely -not- "gen ai" enough to be doing the whole thing in agents, even 3 or 5 years from now, if only because there's so much context we _can't_ give to these models, without reverse centauring ourselves with cameras and mics and oodles of compute everywhere, and so far that hasn't exactly been playing out the way the frontier labs and the singularity folk hoped it would (meta glasses? humane? really?)
As a small business owner I see this. Random people contacting me to sell a software where I can instantly tell it is vibe coded. Subscription prices half or 1/4th of incumbents. Domain names registered in the last few months.
[1] If anything, Waymo has a higher chance and global Waymo adoption is probably 20 years into the future, at least.
They're trying. I see Uber robotaxis almost daily.†
I haven't bothered to see if they're still in training or actually taking passengers.
† https://lucidmotors.com/stories/lucid-nuro-uber-partner
I'm substituting my own[0] vibe coding for buying[1] apps. Language mini-games to help with German? A few prompts. Fluid dynamics simulation for an airzooker? Vibed. A web app listening for a MIDI keyboard, upon which you can drop some .midi files, and get a rhythm action game to learn the piano? Vibed. Webcam for my Raspberry Pi? Vibed. Getting Marathon 2 (well, the open sourced and upgraded engine, Aleph One) working as a web app? Vibed. Isochrone maps? Vibed.
Half of this I can even get done with the free models.
> without reverse centauring ourselves with cameras and mics and oodles of compute everywhere, and so far that hasn't exactly been playing out the way the frontier labs and the singularity folk hoped it would (meta glasses? humane? really?)
Yeah, so fortunate that cameras are expensive and there aren't 6 on my table right now between laptops and phones. :P
Seriously though, what's saving humanity collectively from everything getting automated from the panopticon we'd already built before Transformer models got good enough for even the most basic of classification and translation tasks, let alone anything we now use them for, is that machine learning takes an obscene number of examples before getting competent. Any living creature that needed so many examples would starve to death before learning how to eat.
This difficulty is why, for all the billions of miles that Tesla cars have collectively driven, perhaps pushing trillions now, they're still not sold to the public without steering wheels. Tesla claim to make such vehicles now in the form of the Cybercab, but they're not for sale, and even then some of the pictures that get in the press still show steering wheels.
[0] if you can call anything vibe-coded "my own"
[1] or worse, given the popularity of subscription models in this era, leasing some SaaS
When do you have time for all of this? I don’t mean the vibing part but the using the app part.
I agree that agents are super good for one shotting throwaway code for tasks that would have required manual human actions previously but I would never have bothered buying an app for that.
Not having to deal with IT support for relatives is a win though since now I can just throw it at an LLM!
Mix of this being spread over more than a year, that I'm not doomscrolling because HackerNews and Telegram are my main social media presences, and being unemployed/prematurely retired (which one depends on what one thinks of €1k/month passive income and no rent).
I am also weirdly unmotivated by opportunities to spend money, which is both how I got this passive income and lack of rent, and why it's borderline enough for me. FIRE is very easy when all your working life, you only spend rent+50%, the rest of your paycheque going to savings and investments; but most people can't do this.
Some of you guys hate Musk so much that you say blatantly nonsensical stuff. Like, how can you be on HN and authoritatively mouthing off and not have some idea about the xAI-Anthropic compute contract, or the fact that xAI took down its own Grok compute capability so it could sell that compute to Anthropic? And if Anthropic hated Musk/xAI as much as you are saying they do, why would they be paying xAI billions for compute? Think/research before typing.
Some of us are still employed mate
This was pretty pertinent to the meaning of that line, brother.
- fast because the code is already indexed no need to start every prompt rediscovering the world with rg
- in editor/vscode review much better than relying on git diff at the very end. You can click on next button to jump there. Quick edit in place. Or a quick prompt (seconds) to fix something small
- in editor completion / quick edits, when you do things by hand.
It seems that everyone loves agentic Claude code things these days but I don’t understand how you can review what it did and remain as much in the flow as you do with cursor.
Also I find that incredibly slow. By the time Claude finished a prompt I could have done a few in cursor.
And mind you, I’ve been wanting to use zed instead but the clunky vscode fork called cursor. But cursor has just a better UX for AI integration even though it’s slow and guzzle up RAM.
The way I review it is to let it finish a PR sized slice and then I review the whole diff in a separate terminal as if it’s a PR from someone else. Then I very impolitely list the shit that needs fixing, let the LLM grind out a new batch and repeat the process. When the code is good enough I throw it into a commit and push a PR out for other humans to review.
Clicking around in a UI would just waste time and if the change is too large to review in a terminal window the change is too large for another human to review.
I sometimes see people reviewing LLM code leaving the same review comments multiple times, or worse, after some time they just give up reviewing.
Insanely inefficient. It's 10x more productive to watch the thinking traces and edits in real time and steer the model appropriately.
If your workflow is typical no wonder my team members who use claude code are so much less productive.
If I work with one agent / few subagents on one feature I can steer it as soon as I notice it drifting into the direction of waste. This way I only review the total diff 1.5-2 times. And I also don't waste my own mental energy on context switching between tasks agents are producing diffs for in parallel.
When thinking first started and you would still see the whole "thinking process", I thought it was a ploy to 10x token use because it was just the most inane bullshit. "But wait, the user is asking me to" in loops.
I use Zed to review the code.
CLI is great for most things. Some times, when I am reviewing a PR, I just tell Codex to do a diff between source and target branches in PR, and generate a HTML report. Then I read that, and also review relevant pieces of code in Zed.
I tried using Cursor, many of my colleagues use it, but stopped using it due to clunkiness. And of late, even in there, many of my colleagues use agentic coding. Instead, I prefer the snapiness of CLI (Codex coded in rust, takes less memory per session) and Zed (also coded in rust,) and blazingly fast. Zed is the new Sublime Text if you will.
I just retried Zed and yeah, it's much much better than last time I gave it a shot!!!
I turned on the review stuff, and yeah I can do more or less what I could in Cursor. Check something. Do a quick inline promtp for the three lines I dislike and move on.
And of course the editor is so much nimbler/nicer.
No plan mode, but for big edits I'd probably use codex or something just like you. If you have any advice for integrating for that flow, I'll take them.
I guess whatever advantage cursor has, it will die once "Sol" is no longer integrated, as it was my favorite heavy work model. But I sure will miss the snappy auto/composer stack of Cursor.
For the past three or four months it's performed very well, even the cheaper Gemini models. No serious complaints. But if you go back a year it was pretty terrible. My work style is to keep the coding agent on a short leash so YMMV.
EDIT - AI says:
Cursor seems to have intentionally dropped semantic/embedding-based codebase indexing. Their explanation is that newer coding agents are now good enough at searching repos directly. They can grep in parallel, inspect directories, read likely files, and refine the search themselves. Cursor says this now works as well or better for most cases. [1]
So the architecture shifted from something like “chunk → embed → vector DB → retrieve” to “local search index/ripgrep → agent explores.” There is still an index, but it’s a conventional local search index. No embeddings are involved in that retrieval path. [1]
The interesting part is that Cursor was still reporting a ~12.5% accuracy gain from semantic indexing in January 2026. [2] So this looks less like “semantic search never worked” and more like agentic search improving fast enough that the extra infrastructure stopped being worth it.
[1] https://forum.cursor.com/t/what-do-you-think-about-cursor-re...
[2] https://cursor.com/blog/secure-codebase-indexing
but the the review what the AI did by jumping directly into the file with accept/reject/go next/next file with the ability to edit in place was nowhere near as good in zed vs cursor last I checked.
The features you mention are in Antigravity IDE v1, which I am trying as I still have a Gemini subscription.
No idea how long it will be around for, as Google Antigravity2 (more like Claude Code) is clearly the favoured way forward, plus the new Antigravity for VSCode Extension*, which I didn’t like as much when I tried it. It offers the same accept/reject hunk workflow, but I don’t know about editing in place. It doesn’t feel as if it has the tight integration of Antigravity IDE v1, which feels as if it was designed to challenge Cursor.
Having said that, much of the time I use Google Antigravity2 alongside a Jetbrains IDE. You can still review all diffs within Google Antigravity2 and attach comments to any line on order for a followup prompt to make amendments. But I still prefer the review and accept/reject hunk workflow.
* https://marketplace.visualstudio.com/items?itemName=Google.g...
There are things I will probably miss from Cursor, as they used to iterate on DX goodies quite quickly. For frontend tasks, I liked the idea of Cursor Browser being able to do a devtools-like selection of an element tree to be included in the chat context. I’ve been told there are similar extensions available for vanilla VSCode, so will have to see how good they are.
I mainly use Cursor for another reason though: all of my work is remote over SSH and I don't like keystroke latency, so I prefer code editing / agent conversations to feel local. (Everything else is done in the terminal.) But I would also prefer to use something less propriety. I may go back to VSCode with plugins to achieve a similar experience.
This is a big quality of life thing. Sure, I should just go on a walk without having an agent going, but oftentimes it’s not intruding on it - it’s the thing that actually ENABLES it.
Separate terminal tab with git diff/show, Zed for edits - rarely, there is little reason to change code yourself. Opus at low effort is faster than you can switch windows.
I haven't tried it yet but it looks promising. He would be thrilled if people give it a go!
It’s not an editor in that sense but a tool optimized for reviewing PRs produced by AI.
The GUI slightly better than Claude. And it’s a million times cheaper than Claude.
[1] https://opencode.ai/
[2] https://github.com/oraios/serena
I don’t like relying on GIT for that and making things as good/done as I go. But with these agentic tools that’s what I have to do. Stage everything I like, manually edits a few bits.
But what do you when you find something where you want to say the agent: I don’t like this for reasons x and y redo as z. I’m like one shortcut away in context inside cursor to say that to my agent? You have to change tab in herdr and then what? Type the file name and line numbers? Is there a way to do this that does not seem very frustrating?
Also open code, codex, etc. super slow. They grep the code. Every. Single. Time. Cursor has a semantic index
For instance, in Claude Code you can have it bring up this TUI overlay after every edit it makes (it literally is an overlay, rather than a separate tab in your terminal) and you can annotate line-by-line anything you want changed or rejected. Then once you exit the overlay those suggestions are automatically piped back into the agent.
I tried a few of these TUI-based code annotators, and this one is by far the best. It also works for files not under version control.
I'm currently not using it as I also prefer something more akin to "text editor with agent built in", but I wanted to mention here in case you were curious.
I know this is change of context, but I have full screen terminal toggled with keyboard shortcut and with middle mouse paste (Linux) I do not have to move cursor anywhere
iterm2 on mac also has a similar mode, but it is a bit finicky to set it up (it is really hidden in the settings).
The in-editor context review is something that does not seem long for first world. Sadly. Maybe I’ll be forced to make a zed extension or something, maybe on top of jujustu…
Cursor is already kind of useless for third party models unless you're willing to spend thousands of dollars. It's only worth it if you're going to use mostly grok/composer.
Inference is cheap; Dax once said on a podcast that Opencode has a close to 90% profit margin on the openweight models it provides inference for (at 10x cheaper pricing). Those models are close in size and spec to models from large labs.
If either one of them stops doing that, customers will stop using them, because other models will become better.
In other words, there is no point in talking about inference cost in isolation.
When they outsource training to China they will be so profitable!
I think it's possible to have a better TUI, with something that isn't modeled after a shell prompt. Apparently, recent updates move in that direction. The GUI version of Claude Code seems to be coming to Linux, too.
However, what I see cooking at Cursor is way more promising than the others. Things like full system understanding, their own forge, multi-repo support. I'd put them as way more visionary in the Gartner magic Quadrant ;)
For the forges, what the sibling said.
Additionally helpful where 2 apps/repos might require running at the same time, e.g. headless web apps, or, for plugin development where the plugin might be a dedicated repo but needs to run in another app to observe changes and make it re-test itself.
I also switched from terminal to the app not that long ago, I don't find it buggy, it has access to a browser which is really helpful.
In particular Haiku 4.5 is rubbish, Anthropic don’t have anything in the cheap/fast part of the market.
For people fortunate enough to still be on a subscription instead of per-token billing this is less relevant, but their time will come.
The cursor model is quite nice because it allows you to switch between cheap and expensive models for different tasks
Claude Code works with models from other providers too. Anthropic supports this. You can configure some Claude Code environment variables to switch: eg changing ANTHROPIC_DEFAULT_HAIKU_MODEL to point to GLM Flash or Luna, setting ANTHROPIC_BASE_URL to point to api.z.ai, and making ANTHROPIC_AUTH_TOKEN the API key for your alternative provider instead.
Some instructions here:
https://docs.z.ai/scenario-example/develop-tools/claude
That said, I've not actually tried this myself, opting to build my own harness instead. And you don't know what information Claude Code might be sending back to Anthropic about how you use competing models and which models you use. I don't know for sure that they do this, but after hearing about how they used steganography in the date of harness system prompts to identify the user's location, I don't entirely trust Claude Code anymore.
Now, I’m no expert, because I was late to the game and have only ever used Pi. But I guess Cursor is some product that’s tied to your IDE? If so, then yeah. That’s just too restrictive. I still love my IDE, but I don’t want it to be my harness too.
Why do you say that? I've used Claude Code with DeepSeek via OpenRouter just fine.
I basically never use the editor but the fact that there's a review UI for all the agent work is incredibly helpful.
For me, that makes the models much more usable. I've also been using GPT models a lot, as they're cheaper and less vomit inducing than Claudes text, so this is definitely bad news for me.
Honestly, _this_ was their moat more than Composer to me.
Why? Written by does not mean designed by etc. There's a lot more to it.
I dont know how cursor works. Everytime ive tried to use, its super buggy and has memory leaks that make my very quite PC sound like a jet engine.
It is for 3rd party models (which you get double the amount of your sub price as usage).
Of course they wont charge you extra for BYOK. how would they???
Shit, for most things, ive got a dev agent that reads change requests from a GDoc and communicates through email with me. I can develop in my mobile
Big reason we built https://boxes.dev around the model harnesses (Codex + Claude Code), so you can bring your own subscriptions.
On the other hand, subscriptions create lock-in in a way that API pricing doesn't.
I think a more likely end is that subscription value decreases over time because API pricing gets more reasonable, but subscriptions stay because they are a good way of getting money out of people consistently.
This is why "subsidized tokens" is possibly a misnomer. Money at lower variance is worth more than the same money at higher variance. Not "subsidy" so much as reducing risk and passing some of that to a consumer.
How long have you been in tech, outta curiosity?
Edit: great answer.
What do you mean by that? If you're being sassy about a downvote then I'll remind you that when you reply to a comment, that person can't downvote your reply. Also this is an account with barely any comments or points, I'm pretty sure they can't downvote at all.
Just a stub of a conversation.
I don’t care about the points. Just the chat, which seems finished.
But it didn't come from the person you were talking to, so editing on a sarcastic "great answer" doesn't really work.
The person you were talking to just hadn't responded within the first hour. They still might.
Using both OpenAI and Anthropic models, Cursor is cheaper than having two separate subscriptions, and their free tier of Composer is good enough. I will now have to reconsider my options.
Then you can let it run, and when it stops, point a fresh session to the ledger to pick up the work from where the previous one left off.
You can get more fancy by using something like GitHub issues for coordination too and have multiple sessions create issues and epics and others to work on that in isolated worktrees, but the spec-plan-ledger approach usually works great for small to medium large projects.
Ironically, as a Cursor and Claude subscriber, but not an OpenAI subscriber, this will push me back to Anthropic. I don't feel good about giving OpenAI money directly, but I've found switching to GPT 5.6 Sol now and again in Cursor to be useful. Now I'll probably just forget about OpenAI models.
They're commodity fetishism maxxing.
Cursor/SpaceX/Grok is a different beast. I think they will cheat other companies but I'm not convinced yet that they have it out for users. We'll see. I'm certainly not some kind of Cursor lifer or something.
We should all move to OpenCode anyway. :)
If his teams will handle the data of US citizens with such disregard, he'll almost certainly treat his customers (many of whom are those same US citizens) the same.
The other AI companies obfuscate their goals much better.
> as a Cursor and Claude subscriber, but not an OpenAI subscriber, this will push me back to Anthropic
Instead of deciding based on the merits of the model, they are being pushed by the actions of the companies. You should be able to use the best model, or whatever model you like, independently of what any company thinks or does.
I think maybe you’re suggesting private companies should be forced to support all models which is a position that isn’t technically feasible or one that most would agree is even a sensible ideal.
I'm suggesting you shouldn't let a company dictate which models you use. Putting yourself in a position where they can is detrimental, because you miss out on using the best model for the job.
This provides a method, but the data looks stale and perhaps a bit thin compared to say, Cursor, or even AntiGravity data.
It has been more of an OpenClaw or Hermes alternative than a coding agent like OpenCode or Pi, so it’s likely to do well given less context bloat.
I took the 50% discount instead of cancelling - but with this I'll likely not renew.
So sad to watch them go from great to mediocre in such a short time.
Virtue signaling in glass houses...
Since the introduction of the Cursor Token Rate, my company tells us to use Claude and Codex directly with Anthropic (or Vertex AI) and OpenAI, respectively.
https://cursor.com/docs/models-and-pricing#cursor-token-rate
I'm now using Grok 4.6 most of the time. Sonnet some times. But that's it. Cursor-native models 99% of the time. I'm pretty sure Anthropic will follow OpenAI soon in pulling their models out of Cursor because it's now a true competitor.
Cursor will be fine after this, as they have their own first party models and can potentially use open weight models via API or their own hardware, which have gotten good enough to Opus and Fable quality now for many use cases.
The model providers are in a difficult position. With an added danger of open weight models that are within reach of large enterprises running themselves.
Burn.
I think Sam Altman has played a clever business move here, they can use this excuse immediately after the Hugging Face leaks and it's not going to be dismissed willy-nilly. Whereas long term they probably wanted to end their partnership with Cursor anyway. I bet Dario will be happy about this too, but maybe the compute deal prevents him from doing similar anytime soon.
Very interesting times we live in...
The best bet is Cursor Ultra ($200), which is the equivalent a month plan that you get from OpenAI and Anthropic. I get a LOT more capacity out of it because I almost always use the "Auto" mode, which auto picks between Opus 5, GPT 5.6, Grok 4.6, and Composer 2.5.
I like Auto mode so much actually. I use it for my Cursor work account as well because it's about 10x faster than just using Opus 5 or GPT 5.6 and about 90% is good. I switch back to the bigger models if I get in trouble, but that's pretty rare.
Vibe coding requires some discipline. You need understand what it's trying to do. If you have a good mental model you can keep it out of trouble. If you don't, it doesn't matter how good the model is. You're gonna end up in a mess.
I have a $20 Cursor sub still personally and would say that the usage for Grok models is much higher than what $20 gets you with Anthropic. I can't speak for the $100 plan but I would assume it scales similarly.
I rarely run out of my usage monthly even with the $20 plan.
Grok is a workhorse, not the smartest yet, but produces good code and follows the plan. Grok 5, which has been training since beginnig of this year, will probably be a monster. (10T/6T Param model variants)
Anyone else surprised that the maximum notice period allowed is about 3 months?
I would have thought 3 months is minimum notice period, with provisions to stop faster if any provable violation occurs.
This clause probably has a expiry date and they are saying “we are trying not to be jerks here by waiting until the last second to trigger it”.
But this is kind of like banning GPU exports to China. It will encourage them to develop their own technology. Yes, it will take them longer, but in the long-run China will be more successful with their own GPU tech.
The same could be true of SpaceX/xAI
Is there really a high demand for racist chatbots with no safety controls?
And even if there is, will governments even allow it to keep operating?
You had me until:
>The same could be true of a company run by a drug addict.
Elon Musk is nowhere near as capable as millions of intelligent developers in China.
ITAR: SpaceX is strictly barred by US federal law from hiring non-US persons for aerospace engineering.
National Security Scrutiny: xAI operates under increasing US export and tech-transfer controls regarding China.
Scale: A private company's recruiting budget cannot compete with the sovereign industrial capacity and state-backed talent pipeline of an entire nation.
This move seems to protect Astra against distillation. I wonder whether Astra will be released before or after 11/12.
If a company can provide the access to such a model via AWS Bedrock or similar enterprise compliant platform with a decent IDE/harness they can capture billions of dollars of market.
The closest we have to this is Claude Code but I think it can even be better and specialised. Cursor has an enormous amount of data on coding and developer usage and maybe they should use that to build a proper platform
Do I smell a cursor-analogue from openAI?
How can you ban a competitor from using a model that you openly provide access to?
I understand that the OpenAI ToS may have been violated by X, but has anyone ruled on distillation as fair use? OpenAI believes distillation of nonAI produced content is fair use, so they don’t appear to have a leg to stand on.
there are plenty of alternatives, 100s, so I don't think antitrust is on the table, think of this more like a boycott of the world's richest man for being a twat
I wonder if Anthropic will follow suit given their campaign against distillation or if they'll stay quiet given their reliance on SpaceX (among others) for hardware.
The quality/price ratio is clearly superior when running Qwen, GLM, or Deepseek.
To make a product more exclusive won't hide the fact that product is becoming more and more irrelevant.
Although I wouldn't take an Openai announcement literally, as Sam Altman is also not trustworthy.
All of these companies, and I truly mean all, stole data to train their models.
Maybe if Claude joined along it would be more impactful. For better or worse though, Xai seems like its lining themselves up to handle being on their own for cursor.
If your coding tool depends on other companies’ AI models, one business decision can suddenly change the whole product
codex growth at 20mill now, prob more cost effective to allocate that compute to people directly paying now
Is there a way to work around this (like using your OpenAI limits directly?)
The first party cursor models (initially composer and now grok-cursor) and price very competitively for their quality and have very generous usage limits included in the base plans, and then per token pricing afterwards.
Occasionally I’ve used some of the Anthropic models in it but found their cost massively outweighed the benefits for the type of usage we do.
I was using Sonnet 4.5 and 4.6 for a while but the newer models are a lot more expensive at the per-token pricing.
Using OpenAI through codex (cli) is the way you should use it.
The movie Goya’s Ghost explains this in great detail.
I'm surprised how clear they are about this.
Related:
Can highly recommend Zed as an alternative, it's better than VSCode with the Codex extension as managing multiple chats is much nicer in Zed.
Zed: https://zed.dev/
There are also a bunch of newer startups in this space. I'm building one myself, https://boxes.dev -- we're very early but building for this exact use case. Some other ones worth a look are Factory Droid and Amp Orbs. Those two build their own agent harness (like Cursor), whereas with boxes.dev we run the native codex and claude code harnesses directly.
It's gunna to happen anyway.
We are sailing into the age of vertically integrated intelligence.
Last year, Anthropic cut off Windsurf the same way [2]. Fortunately, unlike Anthropic, OpenAI allows you to use your subscription with other harnesses, including Cursor.
Of course today, Anthropic is taking the "high road" and is OK with with SpaceX [3] including their models in Cursor, because they need SpaceX's compute.
[1] https://www.forbes.com/sites/antoniopequenoiv/2026/04/30/elo...
[2] https://techcrunch.com/2025/06/03/windsurf-says-anthropic-is...
[3] https://x.com/NotTomBrown/status/2093541294027280657
Not doing business with them is a completely normal and unremarkable decision.
> After Musk acquired Twitter, now part of SpaceX, the company broke (opens in a new window) the terms of our contract (alongside many others).
I have never seen one large company throw another large company under the bus without any legalese filters like this before. Wondering if OpenAI site got pwned. Or maybe we in some strange, alternative reality. I doubt even their LLMs would be this frank.
https://www.theverge.com/ai-artificial-intelligence/921546/e...
[1] https://www.forbes.com/sites/antoniopequenoiv/2026/04/30/elo...
The products are basically equivalent but JetBrains is a profitable bootstrapped company that has been building IDEs for 20 years and Cursor is a flash in the pan startup with a squillion dollar valuation to support. Sooner or later they’re going to do something that will fuck their users over.
And selling to SpaceX was it!
> we’ve decided to hold the contract cancellation to the latest date we can while not providing future models to Cursor.
The man is an outright existential threat to democracy.
https://news.ycombinator.com/item?id=49428835
If Claude pulls away from Cursor as well then Cursor is basically done.
I'm not a fan of Elon, nor do I use Cursor, but OpenAI and Anthropic aren't the only models.
Also, it's a signal more than anything. OpenAI pulls out of Cursor. Then Anthropic and Google does as well. Then you no longer have "Cursor" you have Grok.
Grok doesn't have the best reputation in general, regardless if you're tech savvy or not.
i loved switching between sol / fable / grok 4.5.
... and something tells me fable might not be available on cursor for much longer :/
Seriously though, when the api providers start blacklisting cursor and they’re forced to swap to open weight models…
If that happens, what even is the difference between openrouter and cursor?
Strategically, OpenAI long-term will want to own the direct relationship with all their customers. As will most major AI labs.
https://cursor.com/docs/models-and-pricing
i only tell VC type people or others from whom i want something "oh yeah grok is cool" if they mention it, because they're always sucking off Elon
For a company named "Open" AI, they've been hardly open at all and the slew of resignations of top tier execs in the company is really the proof.
Make no mistake, this isn't a moral decision. This is just killing the competition using morality as an excuse.
i work in qa so i'm naturally wary of adding another critical dependency to a development workflow, if the model you actually want can just disappear because two vendor change ownership or start argueing then the wrapper has to be really good on its own.
cursors harness and its own models may be good enough that users stay, and if they do cursor has a moat at least. if everyone follows the openai models out the door, it will just become a polished ui.
Distillation is a threat to frontier model companies, but that doesn’t make it dangerous or wrong. OpenAI can (and likely will) use the same logic to ban you from using open source harnesses.
That is to say, this isn’t a fig leaf imo.