> I am still baffled by the fact that we have collectively agreed to use agentic harnesses by the same companies that are selling access to their APIs.
It's because the subscriptions force you to do so. The subscriptions are the most economical way to use e.g. Claude by close to an order of magnitude. If you max out a 20x plan every week, doing the same work with the API would cost you well into the four figures.
Anyone already using the Claude API pricing and using CC over OpenCode is kneecapping themselves.
I switched over to codex with pi last week. Even though I strongly dislike OpenAI and I hope this is a temporary solution, they're the only one of the frontier models that let me use my own harness and after recent CC shenanigans I'm done with proprietary harnesses.
The immediate thing I've noticed: I get way more out of the codex $100 plan than I was getting out of the Anthropic $200. Like, probably 2x at least.
The other think I've noticed: when using strict guardrails, TDD, reviews etc. I cannot notice any quality difference. Not only between Opus and Codex but even between the most recent models - GPT 5.3 code, GPT 5.4, and now GPT 5.5.
Well, 5.5 uses a huge amount of my session limits. 5.3 is very light, 5.4 somewhere in between. So now I use 5.4 for the main session/debugging/planning and then execute with 5.3.
Regarding usage, of course, it's hard to say how much is the model and how much is coming from Claude code and all this ridiculous malware scanning.
But it's nice to use a lightweight harness like pi and see that even with all my personal instructions, a good bunch of skills, custom tools etc., if I start a session and say "hi" I'm starting out with about 15k of context used. I think a closely equivalent setup in CC would start at 30-40k context.
Correct. However, last time I checked enterprise customers are moving to metered billing. GitHub also decided to so. So it seems the subsidy is coming to an end? I don't know.
Strongly agree. And not only that, but time has _already_ shown the continued degradation of the github experience even with users ostensibly sticking around trying to "make it better".
Indeed. Back in 2018 and 2019 I expended a fair amount of time and energy reporting a squash 'n' merge metadata rewriting bug to GitHub and advocating for the behaviour to be changed. [1]
Once or twice someone internal to GitHub got interested... and then drifted away again. Years later the broken behaviour remains. And I'm a lot more cynical about thinking GitHub fundamentals might ever get any better.
I think OP is basically applying "vote with your wallet" strategy and/or some kind of "action speaks louder than words". As I understood from the article, they have been vocal about trying to change things, but they are shouting into the ether since nothing has changed and in fact only getting worse.
There’s a difference between a relationship with a person and an organization. I think the difference is large enough that the analogy doesn’t really hold.
Exactly, only humans should have at least one chance to grow and improve. Orgs are heartless legal entities that deserve no loyalty whatsoever, they are all one acquisition away from turning on you (as a customer or an employee).
> There is no avenue by which you make GitHub better by continuing to use it as it has been.
I feel like in a very mundane sense, I pay GitHub for a service, and they use that money to pay developers, to then make GitHub better.
It's tough to be working somewhere when usage is booming, and your service is crashing all the time. It's also tough to migrate your infrastructure between platforms, which it sounds like GitHub finally has to do in order to scale to the next level, to really take advantage of being part of Microsoft, although that has to feel pretty frustrating in the short term.
So hang in there GitHub team. Just keep fixing things.
I do work at GitHub. I shared the above as a nuanced "yes and" to the pain that Mitchell is feeling.
In the same way that Mastodon didn't replace Twitter even when Twitter went to shit, I don't believe in the various GitHub alternatives becoming a broadly-used thing. Maybe we'll end up with more GitHub-alikes like Codeberg, mabye we'll end up with some communities adopting novel forges like Tangled and Forgejo. But it beggars belief that most of the millions of GitHub's users would switch to something so much more complicated. Has the same energy as "20XX is finally the year of linux on the desktop".
My very personal hot take: the likeliest happy future is _most likely_ to happen through improving GitHub. I vote with my feet to do that from inside, and that's all I wanted to add. Hence "I hope we do the things that make you want to come back one day." I believe in it enough that I choose to work here on exactly that, because like Mitchell, I care very much about the platonic ideal of GitHub. He's ready to move on, and I'm not yet. There's no value judgment hiding inside that.
> But it beggars belief that most of the millions of GitHub's users would switch to something so much more complicated.
I've moved my projects over to my own personal Forgejo (when I don't care about collaborating on them) and Codeberg (when I do). I find that ecosystem vastly simpler in the common ways that matter. For instance, viewing large diffs and syntax highlighted files is unbelievably faster, about as fast as GitHub's use to be before it was "improved".
For every way I use those forges as a solo or small-group contributor, the alternatives are as good as or better than GitHub today. Some product manager could become a company legend by figuring out how and why that is, then getting someone to do something about it.
I'm glad you are optimistic. GitHub will need employees with that attitude if they're going to pull out of their current trajectory.
To be clear- from a user perspective, "improving GitHub" means "restoring reliability to what it was 6 years ago". There's no killer feature that makes people stop leaving, if my PRs don't lead every third day and actions never work.
I may have my timelines wrong but I don't remember github being rock solid 5 years ago. I remember multiple outages keeping us from pulling code for go packages that were not using an enterprise dependency cache and killing multiple days of work a year for those systems. It's what I used as a forcing function to move people TO an enterprise dependency cache, and to find the few scofflaws running work code off of github.com versus enterprise.
Speaking of a "year of Linux on desktop", it's mostly not happening because the desktop lock-in has largely eased. I of course love my Linux desktop, but I use relatively few native applications, and every one of them is multiplatform now. Windows desktop becomes less and less relevant in its own way, by degradation of experience, and by being replaced with consoles and the Steam Deck.
Same may happen to GitHub. CI/CD tools and workflows can become more portable and adaptable. Independent code review tools that can use GitHub API along with a few other APIs may become popular. GitHub will become one of, not the one. I won't call it a bad outcome.
Comparing to twitter is astute, as there are some analysis that point to it being mostly bots in 2025.
I can see the same happening for GitHub, in fact it seems to be actively trying to move in that direction: a platform for AI agents to host code, to review code, with little to no human activity.
Just like everyone who didn’t want to deal with bots left twitter, they will soon leave GitHub for similar reasons. I’m sure there is a future for GitHub as the code hosting platform for agents but it should be no surprise then when real people like Mitchell and the rest of us jump ship.
I think a better comparison would be between GitHub and 1Password. Both started out as really excellent things for individuals and both became really awful things for individuals in their pursuit of enterprises.
- They ditched their previous android app for a new one that doesn't get the grandfathered accessibility access so autofill is mostly useless...
- On mac, safari integration is consistently flaky. It regularly keeps getting blocked in a loop telling me to unlock 1password when 1password has already been unlocked.
- Passkeys are unreliable to the point of being unusable
- Autofill frequently doesn't work well where for some reason the site with the same url as saved in 1password is not offered during autofill. When 1password used to work, it helped catch phishing attempts because it wouldn't show autofill on pages that do not match. Nowadays because of the shitty autofill, people get trained to go to the app, copy the password and paste it in the website. This means that it will no longer protect from phishing attempts
- The previous behaviour of saving any newly generated password as a password object (not login) was much better. Now newly generated passwords are only available in the password history of the browser extension you specifically used.
- I can't tell 1password to ignore a specific website
At this point, the only reason I'm not using bitwarden is that search is very slow on it with 2k+ passwords.
When I quit using 1Password, it was when they dumped native apps for electron apps and quit supporting the product I’d been buying upgrades for every couple years, in order to pivot to a cloud model that lets them imposing an enterprise subscription model for enterprise users onto individuals. Dunno what they’re up to these days, but I’d be shocked if they could last six months without enterprise customers, so I know I’m not relevant to them anymore. And that’s the same way I view GitHub — individuals are financially and strategically irrelevant to their bottom line.
It’s their right, certainly, but it means I use GitHub as a Google Site replacement and my only active repo is archived whenever I’m not pushing commits to silence all the unwanted crap that comes with a GitHub repo. I’d be daft to ignore free hosting and I don’t care in the slightest that it’s one nines. Makes me laugh every time, though, to think of all those billion dollar AI-layoffs businesses having to stop AI work for a day because AI proliferation broke the freemium model and GitHub’s too hooked on being home to unfunded, mission-critical infrastructure projects to close the barn doors on free.
I can't globally disable that "autofill" also hits "submit". I want to review what it autofills before I submit. I consider this a security risk. I can disable submit only on a login-by-login basis, and my coworkers are able to reenable it again. I can't globally disable it for myself.
FYI I recently discovered a 1p browser extension feature named “Password Generator History”. It has a record of all generated passwords, whether their respective items ever ended up saved or not. Live saver.
Same here. I paid for my family's accounts for many years until the app suddenly became much worse. Honestly, Apple's own Passwords app has 95% of the features of, and the ones it does have work far better than the 1P equivalent. I can't imagine paying for a personal account again.
Not the parent, but the only thing I really hate about 1Password is that I can't tell it to never offer to save a specific site's password. I can turn off all offers to save passwords, or I can have the stupid pop-up ask me multiple times a day if I want to save that password. The pop-up chases me across the site until I get rid of it. Aarrgh. Blood boiling. Rage overflowing.
I have the same complaint about lastpass. With lastpass it's doable, but I have to keep looking up how to configure a site to never site and never ask.
It’s getting buggier and buggier, not being able to fill in passwords properly is kind of a glaring omission of a password manager (and that’s on three different computers).
They keep adding features but seem to show little interest in fixing bugs. I submitted debug logs, recorded videos etc but it just trickled out in the sand. And as another poster wrote, it all started going bad with the switch to Electron (might be the rust backend that is the problem, I don’t know and frankly don’t care, it just doesn’t work as well as it did before).
> But it beggars belief that most of the millions of GitHub's users would switch to something so much more complicated. Has the same energy as "20XX is finally the year of linux on the desktop".
This is funny, because 2025-on seems to be starting some couple years of Linux on the desktop/laptop. Valve introduced millions of people to gaming on Linux, bazzite is exploding in popularity, and that popularity is pouring into other projects like Omarchy, Mint, Ubuntu.
GitHub maybe will end up like Twitter - where the people who are there are there because they have to be, while the people actually enjoying their time online are on different platforms.
Maybe you can install homebrew and open source apps to make it more Linux like, but you'll still be stuck with Mac OS's shonky window and task management UI unfortunately.
It is actuly good for the ecosystem to have competition. Githubs quasi monopoly was a bad thing. And will continue to be a bad thing in the future if it remains
The problem is that from the outside it seems like Microsoft no longer cares about the product. So much so that "the product" has become "shareholders"[0].
We've just been moving into a world where metric hacking is the desired outcome, not an outcome to try to avoid. These companies are only surviving because of their monopoly statuses. Because of momentum. It's a powerful force. It's the reason Twitter still is around. The reason Facebook is still around. But them being around doesn't mean they're good. It doesn't mean they're useful. It doesn't mean it is a good product. It doesn't mean the users like it. It just means people are used to the way things are and they aren't angry enough to leave for something else. But these companies are actively creating friction for users, daring them to leave, gouging them for everything they can. FFS Microsoft is the largest contributor (even more than Valve) to creating "the year of linux". Sure, it'll never have M$FT's market share, but it sure is eating into their revenue.
We've all lost sight of what made software so powerful in the first place. Why it became so successful and changed the world. We used to ship good products that help people, make their lives better, and make lots of money in the process. Now, I think all that anyone cares about is the last part. Now we're actively being hostile to those that make the systems better. And that system is fucked up and will destroy itself. That's not a good thing, because it does a lot of damage along the way. It is a system of extreme myopia.
In the last 5 years I'd argue that most software has made my life harder and more complex, not easier. There are definitely exceptions to this (ghostty being a great example), but there is a strong trend. I know I'm not alone in this feeling and I think we're getting to a point where a lot of people are no longer willing to dismiss their own gripes. This is not a good sign...
I'm glad you're optimistic. I do hope things can change. And my frustration is not directed at you. I really do want you to be right and I really do want to see change come from the inside. But I do not think those leading the companies now have any foresight. To be honest, I'm not even sure there's anyone at the wheel. It feels like we've just let the market forces steer the ship. If the currents steer the ship, then there's no captain, regardless of who claims the title. Frankly, I don't want to be on a ship without a captain, but here we are.
Sure it does. Users who continually push for the right features, stress test things (under normal circumstances), demonstrate uses of the platform that could be baked in by default, etc. are all highly valuable to everyone. And the social aspect matters too, even if GitHub really isn't a "social coding" site anymore. If great people doing OSS stuff are all on various GitHub projects, that encourages more good people to do good OSS stuff.
This doesn't apply to current Github issues, where rather than a lack of the "right" new features, it's just an escalating degradation of existing services that is the complaint.
The attitude of "stay to support the product" can prevent a better replacement. When Digg torpedoed themselves back in 2012 or whenever, that exodus was a big part of Reddit growing from niche to dominant.
The only users who can push for features now are those who can somehow directly influence people working on GitHub (a small number of users) or those with massive purchasing accounts that can shake Microsoft itself to its core (governments, fortune 100 companies).
I suppose us "normals" can push by making it easy to replace GitHub with something else, so that they start risking losing it all.
> Users who continually push for the right features, stress test things (under normal circumstances), demonstrate uses of the platform that could be baked in by default, etc. are all highly valuable to everyone
That's the job of GitHub's product and engineering teams, not the users.
To add on, GitHub has made it explicitly clear that they are both not working on features to focus on their Azure adoption and many core projects are in stasis even from community contributions.
No. Products don't magically get good because people conjured up features from thin air or just copied a competitor. It is very much a two-way street, especially when the product acts as a platform that tries to support heterogeneous use cases.
That sounds like a reading comprehension skill issue? In which case I don't see why me summarizing would move the needle.
But if it helps, no, the data being discussed is surrounding the economics of running inference and R&D, nothing to do with the utility of models for coding.
This is also something of a non issue because as context grows and attention gets diluted, the models perform worse. It'll cost Anthropic more to run your 900k context session, yes, but it's in your interest not to have a 900k session in the first place.
Copy/pasting at scale is how tons of software has been written for a long time, or have we all forgotten the jokes people used to make about StackOverflow?
If the other project were equally aligned with the value OpenAI places on consolidating power and wealth onto Sam Altman, I don't see why OpenAI wouldn't do what they say.
> There is the famous lawsuit of Mattel suing Bratz, on the basis that the Bratz creator started to work on his new dolls while being employed by Mattel.
That's at least reasonable considering Bratz is a competitor.
If the Bratz creator started working on them while working for a company that made water filters, that would not be reasonable.
> have a clause that says that the employer owns everything you do while you’re working for them
Good companies will have a clause that says the employer owns everything you do in the relevant field of the company while working for them, explicitly naming that field.
If you work for a logistics company, you would be able to write your own video editor without any worry. If you work for a not-shitty company.
I’ve worked at 12 companies in Silicon Valley and they all had variations of such language, whether huge Fortune 50 corporations or 4-person startups with the ink still wet on their incorporation filing. Yes, the better ones have a clause that restricts the claim to competitive work, but not all do. Smaller companies will be more flexible with you, particularly if you’re a key hire. At larger companies, it’s typically take it or leave it. HR has forms that must be signed and they sure aren’t going to involve the legal department for you. And to be clear, nobody is going to come after you for a completely orthogonal work product. There’s no point because they can’t monetize whatever orthogonal thing you created and it would cost them lawyers fees for no return. But I have seen companies try to enforce those contracts before. I have not ever seen one go to court. They typically settle before they get that far, but rumors were that IP was given to the company. The employee left in any case and redid the work outside.
> Yes, the better ones have a clause that restricts the claim to competitive work, but not all do.
Yeah, this is my whole point. The rest of your comment, I agree with. In practice you're unlikely to have an issue for something orthogonal and especially something financially worthless.
But you might. There are a lot of people out there that have predatory attitudes towards IP, especially among C-level or would-be-C-level.
And companies that restrict to competing IP do exist.
It’s certainly shifted toward restricting the scope to competing IP, but that’s only in the last 15 years or so. Prior to that it was definitely more all encompassing. But yea, companies are becoming more enlightened as key employees push back and the companies figure out that they don’t really want your orthogonal ideas anyway. But back in the day, particularly when many employees might only work for a maximum of three to five companies anyway, it was extremely common to have “we own everything you create” clauses.
It's because the subscriptions force you to do so. The subscriptions are the most economical way to use e.g. Claude by close to an order of magnitude. If you max out a 20x plan every week, doing the same work with the API would cost you well into the four figures.
Anyone already using the Claude API pricing and using CC over OpenCode is kneecapping themselves.
reply