Note that this is fundamentally different from the Astral acquisition. At the end of their announcement, they stated:
> Cirrus CI will shut down effective Monday, June 1, 2026.
And earlier in the article:
> Joining OpenAI allows us to extend the mission we started with Cirrus Labs: building new kinds of tooling and environments that make engineers more effective, for both human engineers and agentic engineers.
It isn't a product-led acquisition, but more a talent one.
Yeah. Much like Astral - acquiring both the product (because they need to use it internally, but don't care about trying to resell / market), and they also want the talent to keep maintaining it / add features they want.
Just want to note that we will continue maintaining and improving our virtualization solutions actually with even greater attention. SaaS options like Cirrus CI and Cirrus Runners will eventually wind down so we can focus on incorporating pieces internally.
If your scope includes making the Codex web app environments have additional functionality I look forward to it. More enterprise features and yaml backed pipelines.
You’ll be pleasantly surprised. Updates in the coming weeks.
> In the coming weeks, we will relicense all of our source-available tools, including Tart, Vetu and Orchard under a more permissive license. We have also stopped charging licensing fees for them.
This is kind-of neat too, at least in the near term:
> In the coming weeks, we will relicense all of our source-available tools, including Tart, Vetu and Orchard under a more permissive license. We have also stopped charging licensing fees for them.
>It isn't a product-led acquisition, but more a talent one.
I am pretty sure OAI mostly cares about their virtualization IP for MacOS. They already extensively use WSL2 for sandboxing Codex on Windows, and I imagine they want something similar for Codex on Mac.
I’m happy for the founders, they’re great folks. I contributed to CirrusCI a bit in the past and it was a great experience. I even advocated for Cirrus in a couple of my last $DAYJOBs (with varied success). Congrats Fedor!
I’m very sad they’re shutting down, though. IMHO CirrusCI was very close to a perfect CI system (I wrote a blogpost about it [0]). I’ll now have to find something to replace it with in my personal projects. I guess I’ll run their cirrus-cli in GitHub Actions for a while. But GitHub Actions is really poor. I heard some good things about Buildkite.
Having tried many other CI systems, all of which ultimately turned out to be subpar, it makes me incredibly sad to discover only now that Cirrus CI is (was?) quite a bit better than them. :( Thanks for the blog post, though!
Well, check out RWX! We don't do exactly everything like CirrusCI did -- most notably, we aren't open source and we are still managing all of our own infrastructure for running CI right now -- but we have a great "local" story and a validation CLI (`rwx lint`), and a lot of fundamental changes that make us better than other CI systems too :)
Same; the reason everyone ran out to buy Mac Minis last month is it gave their Claw access to iMessage, their browser cookies, and a residential IP. Cirrus provides a way to provision and orchestrate MacOS VMs, which is exactly what I did for running Openclaw (for a minute …).
Not to sell Tart short (it is quite good), but it's "just" a wrapper around Virtualization.framework with a few extra pieces. This is the kind of thing that Codex driven by experts _should_ be able to build very easily.
interesting that was what i thought this was, it keeps boggling my mind the sums being paid for what really could be built by experienced devs on their own teams
I agree that is the reality of why this has happened, but it is completely at odds with the story that that AI maximalists are telling the world, which is that software development is over because LLMs can do anything with a couple of specs and a Ralph Wiggum loop.
Eh, we’re approaching that world (I am "an AI maximalist", I guess)
But someone with the knowledge to guide an AI will have more success than someone without, at least today. I don’t think that will necessarily be true in a year or two.
You don't want a 40-men strong team that needs to be managed, you want 2 guys that already did it and are hungry for the next 10 problems all on their own.
"Hey guys, make our agents verify tool use before responding to the user. See you in 2 months. Here's 2$b"
Agreed. The benefit from not having anyone else or any partial (container) solutions in the computing chain is huge for secure isolation. Getting rid of the intermediary solves a universe of possible problems.
That said, I've been free-riding on tart because they've often surfaced issues I needed to address. Free riders like me are possibly the reason these companies can't make their own way.
Note that apples terms do not allow someone to sell something like an agent running on macOS. They have explicit cut outs for 24-hour minimum leases of full hardware, but they prohibit this with vms
> Then, unprompted, Altman offers up a kind of shocking timeline for the groundbreaking feature of counting: “Maybe another year before something like that works well.” Per Altman, ChatGPT’s voice model doesn’t have the capability of starting a timer or keeping track of time. “But we will add the intelligence into the voice models,” he said.
Every great journey must have an end. As far as I understand Cirrus CI was struggling due to Github Actions eating their market. Cirrus CI was in my opinion much better than Github Actions but it is hard to compete with a bundled solution.
Ability to trivially use custom VM images was quite nice. The amount of CI time spent installing dependencies or copying a cache of installed stuff is nontrivial. Particularly for Windows the time difference is often very substantial. But even for plain Linux, there's no point in apt-get update && apt-get install the same set of things in every run (when using containers, cirrus could build them in-demand too, with little notational overhead).
Defaulting to throw-away-VMs for everything is also the right choice for something where the threat model includes attackers submitting patches/PRs. I'll never understand why folks were ok with just container separation for that (and often have no separation in runners).
The level of aqui-hires is getting interesting - at this point, it appears that if one wants one's career to progress, you need to start some kind of tiny startup like Astral or Bun and hope to be notable enough you can get acquired by someone like OpenAI or Anthropic.
It certainly makes the idea of a career progression / promotion more challenging than it used to be, but perhaps it also opens up some new opportunities. It becomes far more "high stakes" since you have to take the risk of starting and running a startup that ultimately fails if it does not get acqui-hired.
It’s been true for a while, but AI seems to have exaggerated it. To me it reinforces the idea that LLMs created a “K” shaped talent market. Those whose are good become even more valuable.
It also kills competition and disincentivizes providing long term value. I’m not sure how making the job market more like gambling does anything positive for the majority of people
> at this point, it appears that if one wants one's career to progress, you need to start some kind of tiny startup like Astral or Bun and hope to be notable enough you can get acquired by someone like OpenAI or Anthropic
This has been popular for 25+ years. Likely before, but that's when I first started noticing a significant number of companies that were clearly in business solely TO BE BOUGHT.
- gitea
- woodpecker CI
- my own docker registry
- portainer running on my docker swarm
I then define the docker stack in the git repository, and CI builds the images and pushes to build the new image to the docker repository. The portainer API allows to deploy a stack, with the image tag as a parameter.
That is something of a broken promise in the SaaS world; it was supposed to be convenience but instead it became a hundred broken integration points as you struggle to keep up with deprecated APIs and acqui-hires killing off shims you wouldn't have needed if everything had been on-prem to begin with.
Been there, done that, and I still find value in 3rd party services despite the occasional need to migrate.
Self hosting is the way to go if you need to keep monthly services spend as low as possible but you have extra time to spend, such as with a hobby project.
Whenever I’ve worked on real startup projects, self-hosting became a constant source of little tasks for the engineering team to mix into our weekly workload. There were always little tasks to upgrade this service, investigate why that one server was slow, or to migrate something to a bigger server because we were bottlenecked on some resource. Then we had to manage backups and do our recovery drills, along with changing the backup strategy every 6 months because someone had a better idea.
When we started to add up all of the time spent managing everything it starts to look like spending dollars (of engineer time) to save pennies on SaaS bills.
Probably not a popular thing to say on HN, but I now try to stay away from teams that go to extremes to self-host everything because I just want to get my work done, not also be constantly involved in running the underlying services. I do it for my own hobby projects at home but I don’t want to be doing it at work where we have money to spend to lighten the load. If the cost is the occasional migration to a different 3rd party service that’s not a big workload relative to everything involved in self-hosting.
>if you need to keep monthly services spend as low as possible but you have extra time to spend, such as with a hobby project.
We did not have any budget constraints at all.
Depending on 3rd party meant:
1. Begging for issues where project owner has marked them "won't fix"
2. Navigating hundreds of features we don't use, don't need and not having features we absolutely need but they don't have.
3. Weird gotchas and cost saving implemented by them, where we do not want to save any money.
>When we started to add up all of the time spent managing everything it starts to look like spending dollars (of engineer time) to save pennies on SaaS bills.
We actually factored all costs, our self hosted solution are "very lean" and still come out ahead when you factor in time/cost.
>spending dollars (of engineer time) to save pennies on SaaS bills.
I ran by this for long time fasely believing "engineers are way more expensive than cloud bills, so these thousands of dollars we are spending on these services don't matter"
but when we actually got down to fixing all , we realised now it costs us 1/10th of what it used to with more flexibility and reliability.
>Probably not a popular thing to say on HN, but I now try to stay away from teams that go to extremes to self-host everything because I just want to get my work done, not also be constantly involved in running the underlying services
we found exact opposite tbh, integrating into others hurt our performance, wasted our time, made us frustrated and desperate.
Only external service we use now are Anycast for serving our sub millisecond api and "billing", yes billing because it has lots of edge cases, taxes and all wicked things we can't manage on our own.
Still looking to rollout inhouse biller soon though.
Cirrus gave a ton of support for years to open source projects. I congratulate them on cashing out. Running a business like Cirrus did is always a hard road and I will never fault folks who gave time and resources on their platform away for taking the money.
I wish Fedor and everyone at Cirrus the best of luck and OpenAI and thank them immensely for the years of free CI they gave to us in the Pony programming language despite it not having any marketing value to them.
Wow Cirrus was like the one cool CI thing with first-class Podman support. RIP. Guess I'm looking elsewhere (and not at Dagger which refuses to support rootless Podman).
A pity. Cirrus has been providing quite decent CI facilities, for free. One of the advantages (among many) compared to GitHub Actions is the large variety of runner images, e.g., Debian, Fedora, Alpine, FreeBSD, ...
This is a VC site. Morality or changing the world for the better is window dressing on earning as much money as possible and damn anyone who gets in the way. Morality is only good in as much as it’s good for business to be seen as moral. Money people have been majority running things for over 2 decades at least but the mythos lives on.
I noticed on Altman’s recent announcement about someone being mean to him, he said that OpenAI had “changed the world” - conspicuously lacking any mention of “better”.
Interesting move. Cirrus Labs has been doing solid work in mobile CI/CD infrastructure — curious to see how their expertise gets integrated into OpenAI’s tooling and developer ecosystem.
It's hard not to imaging that OpenAI is attempting to build an developer tools eco-system. It makes sense as it's one of the few fields in AI that are currently able to generate sales.
Tart is an amazing tool, and I have been very grateful for it. There's almost no other way I've found to stand up ephemeral CI/CD macOS VMs for self-hosted Git forge solutions. I really hope this doesn't mean that Tart will eventually die the death of unmaintained projects (e.g. Realm post-Mongo-acquisition.)
> We are no longer accepting new customers for Cirrus Runners but will continue supporting the service for existing customers through their existing contract periods.
I mean, it’s no different than any other company who goes out of business. Yeah, their star rating on Yelp is gonna take a major hit, but on the other hand, they don’t exist anymore so shrug feel free to hold a grudge against them, I guess.
This is the risk we run when we get services from companies that aren’t Microsoft, IBM, etc. (I would say Google but LOL, they obviously kill products even faster than the startup death cycle).
Thank you! Full journey is too long for a comment here.
It was hard and I was lucky with my previous pre-IPO gigs at Airbnb and Twitter so I had some bootstrap fund. In retrospective a dev tools startup in 2017 with no network and no VC support was a crazy idea but I was young and didn’t think thought too much.
Then it was long 8 years of raw work and constant questioning this choice. Then finally a third component: luck. In 2024-2025 it kind of grew organically due to market changes and back in October 2025 I finally stopped questioning the future of Cirrus Labs.
My only advice if I may, try to get your first dollar from your startup while you are employed.
Didn’t downvote, but I’m assuming dogging the team’s transition while waiting for their freebies is where people are getting caught up with your comment.
To create an ephemeral (docker-like) MacOS VM on a Mac with full performance and access (e.g. GPU) you have to use a virtualization API provided by Apple.
For most CI use, you can choose between:
Anka, a "contact-us for pricing" closed-source projet, where you have to pay expensive license (easy 3000 USD/yr per machine)
or
tart, which is a lightweight wrapper around the official Apple API.
But you have to know that on MacOS, there is an artificial limit of 2 VMs per Mac... but well:
I just love how companies like this gaslight the whole world with announcements like this.
We started a company to make a big difference in the world and build an engineer’s dream company, and that’s why we have now decided to do the exact opposite and become employee numbers 32,463 through 32,510 at one of the largest tech companies in the world because money is nice.
Look, I’d have done the same thing, I’m not criticizing the choice. I just think we don’t need this kind of weird unnatural rhetoric.
Please just stop with the tech industry puffery. You’re not Steve Jobs, you’re just the DevOps team at OpenAI now. You’re dumping your worthless code on GitHub, and you’re kicking your customers to the curb.
There’s no PR spin left to do anymore. You’re not a company anymore and you’re not a founder anymore.
I just don't think there's much upside to telling it how it is in the press release that gets buried after a week and everyone moves on. For better or worse.
Maybe it’s just my age or how my brain works, but positivity with a touch of BS is just business-speak. My company sold itself to a crappy company who fired most of us, but my departure message for my colleagues was full of “I’ve enjoyed this journey, it’s been a privilege, blah blah” and not “I’ve actually been counting down the days to be done with this nonsense since even before the acquisition.”
You say this stuff because we all agree to sugar coat this type of thing. We all love money and most of us would sell a clearly-never-going-to-be-unicorn company we founded to just about any company if they offered a price that would make you or your whole team set for life. Like, I’d sell to Salesforce, or IBM, or Oracle, or Microsoft or Apple, or Monster Cable, or Ticketmaster. At least half of those, most of us would say are not innovative, and would ruin the product. But no one wants to work with or hire someone who comes right out and says “I hope we sell the company to some rich jerks (ANY rich jerks, really!) so I can retire to the Bahamas.”
I would actually love to work for someone with that much honesty.
I will just go ahead and rewrite the post because it seems like my message isn’t landing. Maybe seeing what it could have looked like would be helpful:
Cirrus has been acquired by OpenAI with the goal of integrating our technology into their products. We will continue to serve existing customers until the expiration of their contracts, at which point Cirrus will cease operations as a separate entity. We are open sourcing our current products and hope they prove to be useful. Thank you to our customers and employees.
> But no one wants to work with or hire someone who comes right out and says “I hope we sell the company to some rich jerks (ANY rich jerks, really!) so I can retire to the Bahamas.”
So just lie to them long enough to extract the value you need to sell out, right?
Thank you for the unabashed insight for the reader, but, as a founder: shame on that.
> we all agree to sugar coat this type of thing. We all love money
We don’t all agree. We don’t all “love money”.
We all don’t step on each other. That’s incorrect.
You’re justifying your choices,
which were made in the pursuit of presenting you in a particular way to others,
There is a theme in a lot of products (and open source projects) of plastering AI all over the description, even if it isn't really related to AI. The first thing I see on the CirrusLabs website is "Accelerate AI Adoption Without Compromising Safety". How am I supposed to know this is a CI product, and even if I figure that out, how am I supposed to know it is a genral CI tool and not something AI specific?
I don't understand why these sites just put a bunch of buzzwords instead of telling you what it actually is.
We have AI companies constantly fear-mongering that their next model is somehow too dangerous to release. But they just continue to go on an acquisition spree.
This just confirms to me that we are no where near AI being able to write any complicated software. I mean, if it could woudln't OpenAI just prompt it into existence? ;)
> We have AI companies constantly fear-mongering that their next model is somehow too dangerous to release
I'm guessing you're referring to this recent report of the security vulnerabilities Mythos found and submitted patches for? That just seems like they don't want the negative press and/or liability if their new model ends up being used to create 0-days that cause widespread damage.
The projects that tend to get acqui-hired — Astral, Bun, Cirrus, etc. — share something specific: they solved a real problem that the acquirer depends on internally, and nobody else had solved it well enough. That's a different dynamic than "be notable and hope someone bids."
What's actually happening is that AI labs have infrastructure needs that don't map cleanly onto existing commercial products, so the fastest path to having the tool they want is to bring the team building it in-house. That's closer to procurement than a traditional acqui-hire.
Whether that's net positive for the ecosystem is genuinely unclear. You get a better-resourced tool in the near term. But you also get organizational risk: if the acquirer pivots or the team gets reassigned, the institutional knowledge goes with it. Tart being relicensed more permissively is the hedge against that scenario, and it's a smart move.
Note that this is fundamentally different from the Astral acquisition. At the end of their announcement, they stated:
> Cirrus CI will shut down effective Monday, June 1, 2026.
And earlier in the article:
> Joining OpenAI allows us to extend the mission we started with Cirrus Labs: building new kinds of tooling and environments that make engineers more effective, for both human engineers and agentic engineers.
It isn't a product-led acquisition, but more a talent one.
It could also be a suite of product acquisition, the CI could be a product OpenAI is interested in having, but not sell.
Yeah. Much like Astral - acquiring both the product (because they need to use it internally, but don't care about trying to resell / market), and they also want the talent to keep maintaining it / add features they want.
Just want to note that we will continue maintaining and improving our virtualization solutions actually with even greater attention. SaaS options like Cirrus CI and Cirrus Runners will eventually wind down so we can focus on incorporating pieces internally.
If your scope includes making the Codex web app environments have additional functionality I look forward to it. More enterprise features and yaml backed pipelines.
If you are interested in yaml backed pipelines check out this open source tool I built for exactly this purpose:
https://github.com/smogili1/circuit
What are your plans for tart licensing going forward?
https://github.com/cirruslabs/tart/blob/main/LICENSE
https://tart.run/licensing/
That sounds like a British phrase for pimping.
You’ll be pleasantly surprised. Updates in the coming weeks.
> In the coming weeks, we will relicense all of our source-available tools, including Tart, Vetu and Orchard under a more permissive license. We have also stopped charging licensing fees for them.
MIT or Apache2 or FreeBSD licenses would be preferable in my case, but GPLv2 or even AGPLv3 (if you have to) would work.
If you’re taking requests…
For now.
Is Sam or family an investor in them anyways?
We were 100% bootstrapped with no outside capital or support/advisory.
Congrats, I am sure you and team made some $$$.
This is kind-of neat too, at least in the near term:
> In the coming weeks, we will relicense all of our source-available tools, including Tart, Vetu and Orchard under a more permissive license. We have also stopped charging licensing fees for them.
This is making my current work 100 times easier. Very welcome timing.
This is awesome because I flippin love tart.
That part is amazing. Back when I first heard of tart I thought it was amazing, with the one downside being the license.
Hopefully development on it continues, or a community maintained version keeps it going.
This is a huge deal! I secretly hoped for this. :)
I mostly clicked the link because I was curious if Cirrus Labs operates Cirrus CI and if so how that would be impacted.
Looks like I’ll need to move the FreeBSD CI jobs for open source projects I maintain to another solution. Anyone have suggestions for alternatives?
I guess qemu over Ubuntu-latest from GitHub Actions running freebsd, but it will be a bit flaky
Same for me. I liked the native BSD runners, qemu on gh is much slower.
>It isn't a product-led acquisition, but more a talent one.
I am pretty sure OAI mostly cares about their virtualization IP for MacOS. They already extensively use WSL2 for sandboxing Codex on Windows, and I imagine they want something similar for Codex on Mac.
This does raise some concerns in major open-source projects:
https://github.com/scipy/scipy/issues/24990
https://www.postgresql.org/message-id/3ydjipcr7kbss57nvi67no...
https://github.com/libprima/prima/issues/291
Did anyone else think this was Cirrus Logic?
https://en.wikipedia.org/wiki/Cirrus_Logic
Yes, it brought back some video card memories
For a hot second yes, and I even use the modern one's tart cli
I definitely had to look and see if this was the same company, yes.
Edit: TIL “Apple makes up 89% of the company's revenue in 2025”
Wow, this is surprising.
I’m happy for the founders, they’re great folks. I contributed to CirrusCI a bit in the past and it was a great experience. I even advocated for Cirrus in a couple of my last $DAYJOBs (with varied success). Congrats Fedor!
I’m very sad they’re shutting down, though. IMHO CirrusCI was very close to a perfect CI system (I wrote a blogpost about it [0]). I’ll now have to find something to replace it with in my personal projects. I guess I’ll run their cirrus-cli in GitHub Actions for a while. But GitHub Actions is really poor. I heard some good things about Buildkite.
[0]: https://garden.pacia.tech/cirrus_ci_is_the_best.html
> I wrote a blogpost about it [0]
Having tried many other CI systems, all of which ultimately turned out to be subpar, it makes me incredibly sad to discover only now that Cirrus CI is (was?) quite a bit better than them. :( Thanks for the blog post, though!
Well, check out RWX! We don't do exactly everything like CirrusCI did -- most notably, we aren't open source and we are still managing all of our own infrastructure for running CI right now -- but we have a great "local" story and a validation CLI (`rwx lint`), and a lot of fundamental changes that make us better than other CI systems too :)
Why would anyone move to a closed-source tool like yours? It's 2026.
FTA:
> In 2022, we built Tart, which became the most popular virtualization solution for Apple Silicon, along with several other tools along the way.
from Tart's github:
> [Tart is for] macOS and Linux VMs on Apple Silicon to use in CI and other automations
My (naive?) hypothesis is this kind of expertise is why OpenAI chose to acquihire.
Same; the reason everyone ran out to buy Mac Minis last month is it gave their Claw access to iMessage, their browser cookies, and a residential IP. Cirrus provides a way to provision and orchestrate MacOS VMs, which is exactly what I did for running Openclaw (for a minute …).
Not to sell Tart short (it is quite good), but it's "just" a wrapper around Virtualization.framework with a few extra pieces. This is the kind of thing that Codex driven by experts _should_ be able to build very easily.
Yes, but it’s also currently the best one. They have OCI compatible Mac VM images that are prebuilt. It’s quite good.
interesting that was what i thought this was, it keeps boggling my mind the sums being paid for what really could be built by experienced devs on their own teams
Short term: they don’t need devs to build it, it’s already built
Long term: they now have experienced dev(s?) to build their next products and features
I agree that is the reality of why this has happened, but it is completely at odds with the story that that AI maximalists are telling the world, which is that software development is over because LLMs can do anything with a couple of specs and a Ralph Wiggum loop.
Eh, we’re approaching that world (I am "an AI maximalist", I guess)
But someone with the knowledge to guide an AI will have more success than someone without, at least today. I don’t think that will necessarily be true in a year or two.
You don't want a 40-men strong team that needs to be managed, you want 2 guys that already did it and are hungry for the next 10 problems all on their own.
"Hey guys, make our agents verify tool use before responding to the user. See you in 2 months. Here's 2$b"
Agreed. The benefit from not having anyone else or any partial (container) solutions in the computing chain is huge for secure isolation. Getting rid of the intermediary solves a universe of possible problems.
That said, I've been free-riding on tart because they've often surfaced issues I needed to address. Free riders like me are possibly the reason these companies can't make their own way.
Note that apples terms do not allow someone to sell something like an agent running on macOS. They have explicit cut outs for 24-hour minimum leases of full hardware, but they prohibit this with vms
So AI company buys devs again, but devs are dead
They want to kill the last good ones
Part of my $DAYJOB is working on timers.
@sama if you need someone to buy to implement timers for ChatGPT I’m your guy - my price is 2 billion dollars.
A timer?
> Then, unprompted, Altman offers up a kind of shocking timeline for the groundbreaking feature of counting: “Maybe another year before something like that works well.” Per Altman, ChatGPT’s voice model doesn’t have the capability of starting a timer or keeping track of time. “But we will add the intelligence into the voice models,” he said.
--
https://gizmodo.com/sam-altman-says-itll-take-another-year-b...
I started writing “why would you want that?”, thinking about Alexa users…
Then I remembered scripting existed ^_^
I cant wait till chatgpt sets me 14 minute timers when I request a 40 minute timer, just another year!
I liked “our incredible journey” more when it wasn’t rushing headlong into OpenMawAI
Every great journey must have an end. As far as I understand Cirrus CI was struggling due to Github Actions eating their market. Cirrus CI was in my opinion much better than Github Actions but it is hard to compete with a bundled solution.
What was good about it? It looks pretty ordinary to me: https://cirrus-ci.org/features/
Much better UX for viewing logs, more supported platforms. Github Actions in particular is also very unstable.
It gives a better docker based experience and it also has measurements of memory and cpu usage to help you dimension things quickly.
Ability to trivially use custom VM images was quite nice. The amount of CI time spent installing dependencies or copying a cache of installed stuff is nontrivial. Particularly for Windows the time difference is often very substantial. But even for plain Linux, there's no point in apt-get update && apt-get install the same set of things in every run (when using containers, cirrus could build them in-demand too, with little notational overhead).
Defaulting to throw-away-VMs for everything is also the right choice for something where the threat model includes attackers submitting patches/PRs. I'll never understand why folks were ok with just container separation for that (and often have no separation in runners).
The level of aqui-hires is getting interesting - at this point, it appears that if one wants one's career to progress, you need to start some kind of tiny startup like Astral or Bun and hope to be notable enough you can get acquired by someone like OpenAI or Anthropic.
It certainly makes the idea of a career progression / promotion more challenging than it used to be, but perhaps it also opens up some new opportunities. It becomes far more "high stakes" since you have to take the risk of starting and running a startup that ultimately fails if it does not get acqui-hired.
It’s been true for a while, but AI seems to have exaggerated it. To me it reinforces the idea that LLMs created a “K” shaped talent market. Those whose are good become even more valuable.
It also kills competition and disincentivizes providing long term value. I’m not sure how making the job market more like gambling does anything positive for the majority of people
> at this point, it appears that if one wants one's career to progress, you need to start some kind of tiny startup like Astral or Bun and hope to be notable enough you can get acquired by someone like OpenAI or Anthropic
This has been popular for 25+ years. Likely before, but that's when I first started noticing a significant number of companies that were clearly in business solely TO BE BOUGHT.
yes selling things unheard of in business. quite unusual to participate in economy
I've moved most companies away from using others stuff
Today we use Hertzner and OVH and roll out our own solution whenever possible.
Running lean and mean.
Depending on such third party services is a trap.
What software do you use to run your CI?
Not OP, but to me the answer is:
I then define the docker stack in the git repository, and CI builds the images and pushes to build the new image to the docker repository. The portainer API allows to deploy a stack, with the image tag as a parameter.
not op either, github actions, self hosted runner on bare metal with lxc containers for each runner.
Cost savings are insane and the speed of latest amd epycs are miles ahead of the default ci instances on github and other places.
Are your self-hosted runner on premise or are you using a cloud service for it?
This is the way.
That is something of a broken promise in the SaaS world; it was supposed to be convenience but instead it became a hundred broken integration points as you struggle to keep up with deprecated APIs and acqui-hires killing off shims you wouldn't have needed if everything had been on-prem to begin with.
Been there, done that, and I still find value in 3rd party services despite the occasional need to migrate.
Self hosting is the way to go if you need to keep monthly services spend as low as possible but you have extra time to spend, such as with a hobby project.
Whenever I’ve worked on real startup projects, self-hosting became a constant source of little tasks for the engineering team to mix into our weekly workload. There were always little tasks to upgrade this service, investigate why that one server was slow, or to migrate something to a bigger server because we were bottlenecked on some resource. Then we had to manage backups and do our recovery drills, along with changing the backup strategy every 6 months because someone had a better idea.
When we started to add up all of the time spent managing everything it starts to look like spending dollars (of engineer time) to save pennies on SaaS bills.
Probably not a popular thing to say on HN, but I now try to stay away from teams that go to extremes to self-host everything because I just want to get my work done, not also be constantly involved in running the underlying services. I do it for my own hobby projects at home but I don’t want to be doing it at work where we have money to spend to lighten the load. If the cost is the occasional migration to a different 3rd party service that’s not a big workload relative to everything involved in self-hosting.
... please keep doing backups and DR tests.
>if you need to keep monthly services spend as low as possible but you have extra time to spend, such as with a hobby project.
We did not have any budget constraints at all.
Depending on 3rd party meant:
1. Begging for issues where project owner has marked them "won't fix"
2. Navigating hundreds of features we don't use, don't need and not having features we absolutely need but they don't have.
3. Weird gotchas and cost saving implemented by them, where we do not want to save any money.
>When we started to add up all of the time spent managing everything it starts to look like spending dollars (of engineer time) to save pennies on SaaS bills.
We actually factored all costs, our self hosted solution are "very lean" and still come out ahead when you factor in time/cost.
>spending dollars (of engineer time) to save pennies on SaaS bills.
I ran by this for long time fasely believing "engineers are way more expensive than cloud bills, so these thousands of dollars we are spending on these services don't matter"
but when we actually got down to fixing all , we realised now it costs us 1/10th of what it used to with more flexibility and reliability.
>Probably not a popular thing to say on HN, but I now try to stay away from teams that go to extremes to self-host everything because I just want to get my work done, not also be constantly involved in running the underlying services
we found exact opposite tbh, integrating into others hurt our performance, wasted our time, made us frustrated and desperate.
Only external service we use now are Anycast for serving our sub millisecond api and "billing", yes billing because it has lots of edge cases, taxes and all wicked things we can't manage on our own.
Still looking to rollout inhouse biller soon though.
It looks like OpenAI has no clue what to do and does what every software company without a plan does: create new dev tools and a new dev stack.
So they want an integrated solution with CI, Python packaging and vibe coding.
That is a $100 million valuation at best, not a $1 trillion one.
yeah super confused, it looks like some tools to manage vms in ci? what is unique about that vs lxc or docker or apple's native container cli?
“what’s the advantage of kubernetes over docker?”
orchestration
docker / virtualization framework is for one machine
this stuff is for 100s - 1000s machines
“We will replace white collar workers”
Proceeds to buy white collar workers.
Cirrus gave a ton of support for years to open source projects. I congratulate them on cashing out. Running a business like Cirrus did is always a hard road and I will never fault folks who gave time and resources on their platform away for taking the money.
I wish Fedor and everyone at Cirrus the best of luck and OpenAI and thank them immensely for the years of free CI they gave to us in the Pony programming language despite it not having any marketing value to them.
Wow Cirrus was like the one cool CI thing with first-class Podman support. RIP. Guess I'm looking elsewhere (and not at Dagger which refuses to support rootless Podman).
Thank you! Cirrus CLI is still around and can run your tasks locally in either Podman or Docker. Can also be used in any other CI.
I mean, yes it is, but we all know what happens to these projects when their primary devs move on or get acquihired. Keybase, anyone?
A pity. Cirrus has been providing quite decent CI facilities, for free. One of the advantages (among many) compared to GitHub Actions is the large variety of runner images, e.g., Debian, Fedora, Alpine, FreeBSD, ...
FreeBSD CI testing is the part I’ll miss the most… time to find an alternative for open source projects.
I don't think people have been keeping track but OpenAI has been hiring a murderer's row of developers for their Codex team.
Meh, Gehrig and Ruth are long since dead.
Incredible how many people are perfectly fine working for a company making AI powered murder bots.
This is a VC site. Morality or changing the world for the better is window dressing on earning as much money as possible and damn anyone who gets in the way. Morality is only good in as much as it’s good for business to be seen as moral. Money people have been majority running things for over 2 decades at least but the mythos lives on.
Killing bad people is changing the world for the better
Yeah, the ‘bad’ people.
Like journalists, or civilians in Gaza and Iran?
> Morality or changing the world for the better
I noticed on Altman’s recent announcement about someone being mean to him, he said that OpenAI had “changed the world” - conspicuously lacking any mention of “better”.
Interesting move. Cirrus Labs has been doing solid work in mobile CI/CD infrastructure — curious to see how their expertise gets integrated into OpenAI’s tooling and developer ecosystem.
It's hard not to imaging that OpenAI is attempting to build an developer tools eco-system. It makes sense as it's one of the few fields in AI that are currently able to generate sales.
Tart is an amazing tool, and I have been very grateful for it. There's almost no other way I've found to stand up ephemeral CI/CD macOS VMs for self-hosted Git forge solutions. I really hope this doesn't mean that Tart will eventually die the death of unmaintained projects (e.g. Realm post-Mongo-acquisition.)
It won’t and I think it will thrive even more. ;)
That is a bit sudden, it would be great if it was possible to get an extra month for migration.
Earlier post:
https://news.ycombinator.com/item?id=47704497
Why are they acquiring seemingly random things, they acquired Astral and now this
bloating up for the IPO
Tart is really cool, impressive, and useful. Best of luck to the team!
> I wanted to work on fun and challenging engineering problems, in the hope of bootstrapping a business as a byproduct.
> We never raised outside capital
I guess it worked out though
Am I reading this right?: a CI company that shuts down CI services with such short notice?
Do service providers not think customers have other things to do than simply maintain their existing infrastructure?
Same thoughts. Guess the migration team responsible will have to kick up the gear.
I don’t think they care about their customers at all, from the statement they consider their business a “byproduct”.
Realistically this is what you agree to when you want to use someone else's computer. They can just as easily ran out of money.
but they're not
> We are no longer accepting new customers for Cirrus Runners but will continue supporting the service for existing customers through their existing contract periods.
Cirrus CI was on a downhill in terms of users and revenue for years now. Most of the customers moved to GHA already.
Plus migration is super easy with Cirrus CLI -- tool to run our CI task definitions locally or in any CI. See https://github.com/cirruslabs/cirrus-cli
I mean, it’s no different than any other company who goes out of business. Yeah, their star rating on Yelp is gonna take a major hit, but on the other hand, they don’t exist anymore so shrug feel free to hold a grudge against them, I guess.
This is the risk we run when we get services from companies that aren’t Microsoft, IBM, etc. (I would say Google but LOL, they obviously kill products even faster than the startup death cycle).
Congratulations!
Can you talk a bit more about your journey without raising funds?
Also what does HN think of that path today when trying to launch a new AI startup?
Thank you! Full journey is too long for a comment here.
It was hard and I was lucky with my previous pre-IPO gigs at Airbnb and Twitter so I had some bootstrap fund. In retrospective a dev tools startup in 2017 with no network and no VC support was a crazy idea but I was young and didn’t think thought too much.
Then it was long 8 years of raw work and constant questioning this choice. Then finally a third component: luck. In 2024-2025 it kind of grew organically due to market changes and back in October 2025 I finally stopped questioning the future of Cirrus Labs.
My only advice if I may, try to get your first dollar from your startup while you are employed.
That is good advice!
I am too late for that as I am full time but also lucky to have had a previous exit.
I am planning to go the VC route this time, because the problem I am going after feels VC size.
However I feel bootstrap or small F&F round gives you more flexibility when it comes to exits.
Congrats again!
As a tart cli user, I'd love to know more about their "more open" licensing of that particular project that they call out that will follow the merger.
That said, I find their aqui-hire by OpenAI disappointing for a number of (mostly personal) reasons. However, I wish them the best regardless.
Genuinely curious on why I'm getting down voted for the above comment
Didn’t downvote, but I’m assuming dogging the team’s transition while waiting for their freebies is where people are getting caught up with your comment.
What was the USP of their CI service?
Ability to virtualize on Apple devices and linux with GPUs https://github.com/scipy/scipy/issues/24990
To create an ephemeral (docker-like) MacOS VM on a Mac with full performance and access (e.g. GPU) you have to use a virtualization API provided by Apple.
For most CI use, you can choose between:
or
But you have to know that on MacOS, there is an artificial limit of 2 VMs per Mac... but well:
https://github.com/cirruslabs/orchard/commit/3cfa2445500f45f...
With https://khronokernel.com/macos/2023/08/08/AS-VM.html
Some people might find it very attractive:
Even without bypassing the limit it is great actually
Better UX than their competitors and support for many different images.
Looks like tech and talent grab for "Computer use" project
https://developers.openai.com/api/docs/guides/tools-computer...
https://github.com/cirruslabs/mtell
yea yea yea, purchase every last company you find, no one wants OpenAI
Great
AI companies need a surprising amount of people.
It's kind of like electric cars charged with electricity from coal power plants.
I just love how companies like this gaslight the whole world with announcements like this.
We started a company to make a big difference in the world and build an engineer’s dream company, and that’s why we have now decided to do the exact opposite and become employee numbers 32,463 through 32,510 at one of the largest tech companies in the world because money is nice.
Look, I’d have done the same thing, I’m not criticizing the choice. I just think we don’t need this kind of weird unnatural rhetoric.
Please just stop with the tech industry puffery. You’re not Steve Jobs, you’re just the DevOps team at OpenAI now. You’re dumping your worthless code on GitHub, and you’re kicking your customers to the curb.
There’s no PR spin left to do anymore. You’re not a company anymore and you’re not a founder anymore.
Making a statement like this is generally part of the terms of the acquisition.
Sure, but I imagine the terms of the acquisition doesn’t say you have to write it in this specific style.
I’m sure there’s a way to say the same thing without coming across as a bullshitter.
I just don't think there's much upside to telling it how it is in the press release that gets buried after a week and everyone moves on. For better or worse.
Why is telling it like it is not the rational default?
If this is a forgettable press release isn't it lower effort than coming up with this type of nonsense?
Maybe it’s just my age or how my brain works, but positivity with a touch of BS is just business-speak. My company sold itself to a crappy company who fired most of us, but my departure message for my colleagues was full of “I’ve enjoyed this journey, it’s been a privilege, blah blah” and not “I’ve actually been counting down the days to be done with this nonsense since even before the acquisition.”
You say this stuff because we all agree to sugar coat this type of thing. We all love money and most of us would sell a clearly-never-going-to-be-unicorn company we founded to just about any company if they offered a price that would make you or your whole team set for life. Like, I’d sell to Salesforce, or IBM, or Oracle, or Microsoft or Apple, or Monster Cable, or Ticketmaster. At least half of those, most of us would say are not innovative, and would ruin the product. But no one wants to work with or hire someone who comes right out and says “I hope we sell the company to some rich jerks (ANY rich jerks, really!) so I can retire to the Bahamas.”
I would actually love to work for someone with that much honesty.
I will just go ahead and rewrite the post because it seems like my message isn’t landing. Maybe seeing what it could have looked like would be helpful:
Cirrus has been acquired by OpenAI with the goal of integrating our technology into their products. We will continue to serve existing customers until the expiration of their contracts, at which point Cirrus will cease operations as a separate entity. We are open sourcing our current products and hope they prove to be useful. Thank you to our customers and employees.
> But no one wants to work with or hire someone who comes right out and says “I hope we sell the company to some rich jerks (ANY rich jerks, really!) so I can retire to the Bahamas.”
So just lie to them long enough to extract the value you need to sell out, right?
Thank you for the unabashed insight for the reader, but, as a founder: shame on that.
> we all agree to sugar coat this type of thing. We all love money
We don’t all agree. We don’t all “love money”.
We all don’t step on each other. That’s incorrect.
You’re justifying your choices,
which were made in the pursuit of presenting you in a particular way to others,
for your own psyche.
More like employees number 32,463 and 32,464 - they’re two people from what I seen on GitHub over the years. (Incredibly strong two people)
So they have even less of a reason to put out BS like this. They have no employees they need to talk to.
Could have been a private email to customers.
Another one bites the dust.
Wow. I have rarely seen a company website with so many buzzwords. Still not sure what they do, except "AI". Good riddance
Cuirrus does not do AI, they do CI.
There is a theme in a lot of products (and open source projects) of plastering AI all over the description, even if it isn't really related to AI. The first thing I see on the CirrusLabs website is "Accelerate AI Adoption Without Compromising Safety". How am I supposed to know this is a CI product, and even if I figure that out, how am I supposed to know it is a genral CI tool and not something AI specific?
I don't understand why these sites just put a bunch of buzzwords instead of telling you what it actually is.
You are checking the wrong Cirrus Labs.
We have AI companies constantly fear-mongering that their next model is somehow too dangerous to release. But they just continue to go on an acquisition spree.
This just confirms to me that we are no where near AI being able to write any complicated software. I mean, if it could woudln't OpenAI just prompt it into existence? ;)
> We have AI companies constantly fear-mongering that their next model is somehow too dangerous to release
I'm guessing you're referring to this recent report of the security vulnerabilities Mythos found and submitted patches for? That just seems like they don't want the negative press and/or liability if their new model ends up being used to create 0-days that cause widespread damage.
https://red.anthropic.com/2026/mythos-preview/
The projects that tend to get acqui-hired — Astral, Bun, Cirrus, etc. — share something specific: they solved a real problem that the acquirer depends on internally, and nobody else had solved it well enough. That's a different dynamic than "be notable and hope someone bids."
What's actually happening is that AI labs have infrastructure needs that don't map cleanly onto existing commercial products, so the fastest path to having the tool they want is to bring the team building it in-house. That's closer to procurement than a traditional acqui-hire.
Whether that's net positive for the ecosystem is genuinely unclear. You get a better-resourced tool in the near term. But you also get organizational risk: if the acquirer pivots or the team gets reassigned, the institutional knowledge goes with it. Tart being relicensed more permissively is the hedge against that scenario, and it's a smart move.