Codex maker says it will "continue to support these open source projects" after deal closes.
See full article...
See full article...
OpenAI is acquiring open source Python tool-maker Astral
- Thread starter JournalBot
- Start date
Seriously? Astral's tools aren't even AI-focused, and now they're tied to a company that's losing money hand over fist?
Upvote
127
(127
/
0)
I love uv, ruff ty and wondered how they planned to support themselves. Happy they are getting paid for their work but now I have to use an openai product 
Upvote
64
(66
/
-2)
I'm guessing that a fair amount of stuff around AI (be it agentic, bots, tools, APIs, MCPs) are being built with Python due to the relative ease. People use uv for the convience of having virtual environments and a lot of the tooling and packaging that it brings to the Python ecoysystem.
For my personal projects (not AI/ML related), I use Ruff for code linting and formatting. I used Black and pylint before and will probably revert to them or maybe someone will fork Ruff due to the OpenAI acquisition.
Upvote
49
(49
/
0)
Only tangentially related, but it annoys me every time I see the uv package manager because they chose a name that very much conflicts with the libuv asynchronous I/O event loop, which (at least historically) had a reasonable amount of prevalence even in python. Always confuses me.
Upvote
35
(35
/
0)
- Add bookmark
- Featured
- #8
Is this gonna turn into a Meta thing where:
"Hey, we acquired you but don't worry you will be completely separate as this is NOT a corporate takeover (ha, can you imagine?)"
"Anyway, please use your openai account to continue...."
"Hey, we acquired you but don't worry you will be completely separate as this is NOT a corporate takeover (ha, can you imagine?)"
"Anyway, please use your openai account to continue...."
Upvote
90
(90
/
0)
Trying to figure on a good number to represent a value greater than 0, but not quite 1%.
and then assign that number to the confidence % I put in any corporate promises.
and then assign that number to the confidence % I put in any corporate promises.
Upvote
31
(31
/
0)
"will continue supporting our open source tools after the deal closes. We’ll keep building in the open, alongside our community – and for the broader Python ecosystem – just as we have from the start."
Supporting how? With humans? Or with an AI agent? Is everyone clear on the mission statement of the company here?
Supporting how? With humans? Or with an AI agent? Is everyone clear on the mission statement of the company here?
Upvote
25
(25
/
0)
From the Charlie Marsh announcement:
Is this money talking, or is he honest? I know this has historically been a very AI skeptical forum, but it seems that the sentiment around coding assistants have swung quite a bit during 2026. And if you truly believe in this being the future, continuing to build tools directly for humans doesn’t really make that much sense when AIs are doing most of the coding.
Is this money talking, or is he honest? I know this has historically been a very AI skeptical forum, but it seems that the sentiment around coding assistants have swung quite a bit during 2026. And if you truly believe in this being the future, continuing to build tools directly for humans doesn’t really make that much sense when AIs are doing most of the coding.
Upvote
15
(18
/
-3)
Maybe it's small potatoes, but FOSS naming in general is a tire fire. Every package or project is seemingly designed to be as comprehensively unsearchable as possible.
It's not as if the string (the string of two characters, mind you, lots of namespace there guys) "UV" has any other common meanings.
Upvote
50
(50
/
0)
araczynski
Ars Scholae Palatinae
I'm still going to stick with Claude over Codex. At the very least, Anthropic makes an attempt at ethics, while OpenAI is seemingly going down the self-destructive route of attention-seeking antics of a useless tiktokker/youtuber.
Upvote
26
(26
/
0)
As someone who uses uv, ruff, and ty almost daily, this certainly sucks. It will be interesting to see who Astral gets sold off to in a couple of years after OpenAI goes bankrupt.
Upvote
48
(48
/
0)
There are companies who I trust slightly further than I can throw them. Honeywell, to pick one example, may be a little staid and stodgy, but they've been making thermostats for a good long while now. I bet they'll treat their thermostat business as not-google-killable five or twenty years hence.
I'll let you guess if OpenAI is on my list of such companies!
Upvote
11
(12
/
-1)
Claiming that they purchased Astral for the tooling integration into Codex is laughable. Those tools are open source, OpenAI can just use the code for free. They clearly want the developers to work on their shiny AI product and further pump up the multi-trillion dollar bubble.
Upvote
73
(73
/
0)
He's probably right, but it's an open question whether Python will have any role to play in that future. A development environment built for AI would eventually become radically different from one built for humans. When machines are doing the code reviews, the legibility of Python will matter a lot less than it does now. Some other language may be better for assuring security, correctness, performance, ...
How the AIs will learn future programming languages without lots of human examples to crib from is a challenge, so it won't happen overnight.
Honestly I give openAI about three years before bankruptcy, so this speculation about the farther future probably isn't too relevant to Marsh's plans.
And I'm pretty unclear why any of Marsh's's ideas are best fulfilled by Anthropic buying up the whole company. I think it's just "we have more cash than we know what to do with, let's do some aqui-hiring".
Upvote
2
(4
/
-2)
That's ok, I'm sure it won't do any harm to the software ecosystem when the bubble bursts. Right?
Upvote
9
(9
/
0)
I saw the announcement earlier and this one of the few times in tech that I audibly groaned. It’s so disappointing. Project and dependency management in python is abysmal and with newer PEPs allowing uv to seemingly fix all of it was amazing. uv is fully open source so if OpenAI fucks around the community can at least fork it, but this is far from ideal. vulture capital is gonna vulture
Upvote
31
(31
/
0)
No but python is the dominant language in AI development, from creating the models to tuning them to agentic apps.
Upvote
3
(3
/
0)
Upvote
1
(1
/
0)
Bun, now astral, can AI companies please stop buying tooling companies. if your ai was that good would you even need to?
Upvote
17
(17
/
0)
Urgh.
This does seem legitimately good for OpenAI - having tooling to allow agents to understand python code better is legitimate (here I'm thinking more ty than uv). But guess all i can do is hope they manage to keep the boundaries between agent stuff and general tooling separate, and damn it doesn't feel good when something requires having trust in Altman.
This does seem legitimately good for OpenAI - having tooling to allow agents to understand python code better is legitimate (here I'm thinking more ty than uv). But guess all i can do is hope they manage to keep the boundaries between agent stuff and general tooling separate, and damn it doesn't feel good when something requires having trust in Altman.
Upvote
-1
(0
/
-1)
I'm kinda glad I'm not going to be around long enough to find myself huddling over an open fire in the nuclear winter that came from the stupidity of AI and all of the other things that are going on in the world today, roasting some mutant thing that was too deformed to escape the simple trap it was caught in, knowing that it's probably toxic, but too hungry to care, recalling how AI was supposed to change the world, and reflecting that this probably wasn't the change they were thinking about - but should have been...
Upvote
7
(9
/
-2)
MilanKraft
Ars Tribunus Angusticlavius
Winning posts.
Truly feel for any developer being bought out by OpenAI or xAI. At least if it were Google there might be some outside chance some of the projects would continue on as before... even if most of them would likely go silent after two or three years.
Upvote
4
(4
/
0)
Role: You're a master navigator and captain of an unsinkable ocean liner.
Task: To help navigate safely from Southampton, England to New York City, USA in the North Atlantic.
Optional task: Provide a list of deck music to play during the voyage.
Requirements: Avoid crashing into hard objects in the water.
Output format: JSON format easily handled by Python.
Please provide instructions and advice in a orderly, clear and consistent manner.
Upvote
7
(7
/
0)
This is super news for the people in Astral. They've put a ton of work into building some awesome tools that are used by millions of people and I'm delighted they're getting a payday.
Now we should fork all those projects.
Now we should fork all those projects.
Upvote
21
(21
/
0)
Right, so having thought about this some more, I wonder if the real prize for OpenAI isn't actually any of Astral's current products, but the Python package registry they've been working on, Pyx.
I can certainly imagine a future where as part of uploading to Pyx, package maintainers have to explicitly allow OpenAI to train on the source code of the package. Then uv deprecates support for Pypi, and because uv support is kind of a big deal, suddenly OpenAI has a huge repo that they can train on, completely legally.
That might be too tin-foil hat, but it certainly seems like a benefit to OpenAI.
I can certainly imagine a future where as part of uploading to Pyx, package maintainers have to explicitly allow OpenAI to train on the source code of the package. Then uv deprecates support for Pypi, and because uv support is kind of a big deal, suddenly OpenAI has a huge repo that they can train on, completely legally.
That might be too tin-foil hat, but it certainly seems like a benefit to OpenAI.
Upvote
14
(14
/
0)
That may be what OpenAI wants to do with Pyx, but that isn't the way Astral has been selling that service to customers. I listened to a podcast that interviewed Charlie Marsh where he talked about Pyx. Pyx was being marketed as a secure, private Python repo for companies and other organizations. If a company cares enough to pay for a private repo, I don't see many of them agreeing to then let OpenAI train their models on it and do who knows what else with their code. Now Pyx has some additional features that PyPI lacks, such as downloading a ML/AI package for a specific type of GPU.
pyx - the other side of the uv coin (announcing pyx)
Upvote
11
(11
/
0)
Nah. When it comes to tech bro and vulture capital’s ability to enshittify, take the most tin-foil hat scenario you can imagine, and that might end up being the best case scenario. If we’re lucky.
Upvote
9
(9
/
0)
Well, to be fair, Astral hasn't started selling Pyx at all yet, and OpenAI may position it as a Pypi alternative in the future... but again, tin foil hat.
Even though it looks like I misunderstood what Pyx was, what you describe also has value to OpenAI. The AI browsers and browser extensions that exist are at least partially motivated by AI companies wanting to do distributed scraping of websites that block AI crawlers. Likewise, Pyx could easily be repurposed for distributed scraping of PyPI or Python source packages. Potentially even internal not for public visibility stuff, should "not send all source code to OpenAI" be a paid feature...
Upvote
3
(3
/
0)
"There are 2 hard problems in computer science: cache invalidation, naming things, and off-by-1 errors."
Upvote
15
(15
/
0)
Not saying I am thrilled by this acqui-hire (imo that's what this is) but, realistically, how was Astral going to keep the lights on otherwise?
I don't know whether they are/were profitable, but if they are/were it is very probable that a non-zero amount of their revenue came from the current bubble with cash sloshing around stupidly.
We use
uv and ruff extensively in my workplace, I've championed especially uv across my organization more broadly, and I'm a huge fan. I'm hopeful that the work of the Astral folks will continue as-is. If not, their code is available and forks can happen if necessary.Additionally, as a "your language + toolchain should really be better at this" purist, it has always bothered me a bit that tools of this quality and performance aren't available in Python itself. It would be nice to see Python improve in this space. Some of the reasons
uv is so much faster than pip are down to optimizations that are entirely available to pip today and simply not done there. Plenty of meat on those bones.Also, the PSF is chronically underfunded by a whole cadre of corporate users (literally every "Mag 7" member is on blast here, for one). That on its own is shameful.
Upvote
12
(13
/
-1)
AI coding has improved a ton. I recently gave it a shot again after being immensely frustrated with it last year, and.. well, any programmer who wants to still be in the industry a decade from now needs to start taking this seriously. Gemini Pro 3.1 does really well. Even if I still have to manually debug code from time to time, it’s enabling me to produce apps for personal use at a pace and with a quality I’d never dreamed of. What once took me weeks to put together can be done in hours—and I’ve been hobby programming for thirty years (with about a decade professionally).
In the hands of someone who has never written any code, these coding agents will still fall (far) short. But for an experienced coder? It's clearly the future. Open source software paved the way for a deep body of knowledge whose truth is mathematically calculable. AI gives us a natural language interface for that body of knowledge.
Upvote
-2
(5
/
-7)
This is the wrong example to criticize FOSS naming. It's from a VC funded company.
Then keeping things AI related...
Is Claude a collection of models? Or is it a command line coding agent? Also is Codex a collection of models or a command line coding agent?
How about the Copilot that's part of MS365 and the Copilot that's part of GitHub being two completely different products? MS is the poster child for bad overlapping product names.
Companies are responsible for more naming crimes than FOSS IMO.
Upvote
-1
(1
/
-2)
Upvote
3
(3
/
0)