What you should be paying attention to: Stainless is shutting down, and their team is joining Anthropic to build, who knows, some dumb integration to make Hubspot data available in Claude, or something equally as boring. But, Stainless was successful. Be the next Stainless. The idea is already validated, these AI companies have already done this to a handful of companies and they're going to keep doing it.
Fun fact, I named it "Stainless" after Stainless Steel pipes, likening ourselves to a high-end plumbing supply shop. If you look at the earliest versions of stainlessapi.com on archive.org, you'll see our original motto was "Quality fittings for your REST API".
All that is to say, the incredibly "boring" infrastructural work of making "boring" APIs like Hubspot's more usefully accessible is absolutely the kind of thing I'm excited to do at Anthropic :)
(It also happens to be what got us all excited to work at stainless in the first place, but of course, we understand it's not for everyone!)
edit: they sure do
Why aren’t they dogfooding their own products to replace such roles?
And seeing how people use it: good programmers review output and iterate to get better output. But bad programmers simply trust the output is good: they have no ability to review it themselves and often don't try.
With about 5-10h over the weekend using free tier Claude and ChatGPT I managed to put together a scraper for a particular thing on a website I’m interested, grab the item images, do an initial pass with local OCR, if it hit some keywords, run openCV to crop for better OCR and dump the hits for further investigation.
Nothing particularly advanced but it would have taken me a horrendous amount of time without it to be half as good, like it did when I built a similar scraper 10 years ago.
Neither were very good code quality i’m sure.
For example, a recent story about the openclaw creator using $1.3M of tokens/month. And let's assume he's getting paid $5M/yr which is probably a vast under estimate.
Is he providing value that a traditional software development org with normal developers couldn't provide for $20M/yr?
Finally in some ways agentic workflows magnify the power of the individual who is adept at harnessing them, they don’t have to argue (much) with the agents to effect their ideas. I’ve found a lot of very bright engineers spend their days fighting to be heard by managers and peers who can’t / won’t understand them. By unshackling them from trying to debate down idiots, they deliver way way more, and of the right things, than they otherwise could have.
Yes, $1.3M in token cost in less than 30 days and some days were even off-peak, if you can call it that with that insane scale that likely hides quite a lot of tokens in the lower bars.
Who claimed that?
Their customers will be happy if their product replaces all the junior positions and midwit developers off the payroll. then that's already a huge saving to any company's bottom line.
Even if it doesn't directly replace workers, reducing the bargaining power of those spoiled SW devs and not having to give them huge raises all the time or they leave, is still enough. That's the whole point of layoffs and offshoring anyway.
Possibly not if they are paying the full cost of inference
Dario Amodei
There are plenty of other reasons to acqui-hire, but it is not the only or even the most effective way to hire the strongest engineers
Successful founder is deeply filtering for very uncommon skills. Effectiveness, grit, decision making, independence, technical plus sales ability.
University is a shit filter in comparison.
The current word is "taste" but even that is way too narrow. Intelligence is close, although usually too academic (hence the VC uni dropout theme).
The other big problem with a independent capable people is that they rarely apply for jobs.
This tests for very different skills than being an exceptional programmer.
The reason why I avoided this term is that in Germany, there exists a quite strict of whatx an engineer (Ingenieur) is, which is defined in the laws of many federal states (Ingenieurgesetz [engineering law]). "Ingenieur" (engineer) is a protected professional title:
> https://de.wikipedia.org/w/index.php?title=Ingenieur&oldid=2... (*)
Falsely claiming that you are an Ingenieur when you aren't (by the definition in the Ingenieurgesetz) is a punishable crime in Germany:
> https://de.wikipedia.org/w/index.php?title=Missbrauch_von_Ti...
There exist some boundary cases under which as a software developer you can call yourself an "Ingenieur", but you have to be insanely careful about whether you actually satisfy the legal criteria (see (*)) - in most cases you don't and you are thus a criminal if you do.
If so, is this ever enforced?
Using the German translation "Softwareingenieur" of "software engineer" on your LinkedIn page might easily get you into trouble.
Typically, as far as I know, law enforcement agencies only get active in the punishable act "Missbrauch von Titeln, Berufsbezeichnungen und Abzeichen" [abuse of titles, occupational titles and emblems] if the culprit gets denounced by someone or if there is a public interest, but everybody knows how easy it is to make enemies in your job or on the internet.
Having a successful business requires a lot of factors that don't really have anything to do with software engineering. Things like luck, connections, access to funding, good marketing, etc. And while have good engineers on the payroll undoubtedly helps, the good engineers aren't necessarily the ones getting a big fallout from the acquisition and may not stick around for long after the acquisition, especially if they get put on a project they don't care about.
What's the difference between a software developer and a software engineer?
The honest answer is that in most day-to-day contexts, the distinction is more about company culture and title preference than actual job duties. A "software developer" at one company might do more rigorous engineering work than a "software engineer" at another.
All that's moot though if your fundamental premise is wrong. Why does Anthropic need "the world's best software engineers" to build on top of the models? Compentent developers can build APIs - sorry - MCP servers and other integration plumbing.
If Anthropic can rummage through your data and workflows to deem you worthy of their grace, then that is seriously wrong.
For better or worse, it's an acquihire.
not anymore lol
I can't even imagine the money wasted on turn-and-burns in the F1000 alone. The US needs a wake up call with respect to consumer / buyer protections. The life of the snake oil salesman is plentiful these days, and you have a lot of AI-psychotic executives who can't seem to get enough.
They mostly have. By mostly refraining from dealing with startups and companies they deem either “too young” or "too small" to be reliable partners. And, when they do, imposing long sales cycles.
And thus the enterprise well is poisoned for most startups.
But buyers try to insert this language into partner/ biz dev contracts all the time.
Much less common for sales.
A place I worked some years ago we even had an escrow foisted on us by our larger partner in the agreement so that they’d be able to continue running the software we were building if we went under.
Honestly, it was a pain in the ass and meant that for them alone we ended up running an older version of the software than we offered to clients because as we developed its capabilities it became ever more integrated into our core platform and we weren’t about to escrow that.
When the agreement came up for renewal at the three year mark we managed to get the escrow clauses removed.
Hadn't heard of Stainless before today. Did it have enterprise customers?
They can also keep the product running behind the scenes for a select few and just shut down the public facing part
I suspect a lot of larger orgs just have site-wide subscriptions with volume discounts that they don’t need.
Either way, it does seem irresponsible and tone deaf for an acquiring/hiring company and an acquired/hired company to send these conflicting signals. If one puts oneself out there as dependable in the face hopes and needs of other, smaller, up-and-coming projects, then a rapid wind-down for $ is incongruent with such a posture.
So much so that, at least for my part, I'd be quite reluctant to hire someone who had engaged in this sort of bob-and-weave pursuit.
> As we focus on Claude Platform capabilities and connecting agents to APIs, we’ll be winding down all hosted Stainless products, including our SDK generator. Starting today, new signups, projects, and SDKs will not be available.
> If you’re a Stainless customer, visit app.stainless.com/transition for help transitioning from Stainless-managed products to other options. As always, you own the SDKs you’ve generated to date, and have full rights to modify and extend them however you wish.
By self-service, do you mean that the SDK generators are now source-available so they can be run by end users locally?
* Manual Maintenance: Returning to the pre-Stainless era.
* Agentic Coding: Works to an extent, but you lose the deterministic, review-free output required to keep an SDK perfectly structured and coherent.
* Open-source Generators: Helpful for basic use cases, but they lack Stainless's full-stack features like multi-language generation and publishing, MCPs, and documentation.
We were an ealy adopter of their Node SDK generator at Mux (and latterly their Typescript and other generators), and the product worked great, and I'm sad to see it be shut down.
At the same time, it's easy to understand why this is a complciated product/market to be in at the moment - it's very tempting and easy to vibe code SDKs from a OpenAPI spec files right now. I would think a lot of teams will just go in that direction (for better or worse), using the same toolchain that the product developers are using today for the product, for effectively no extra cost.
But, no one uses it, because uber and lyft have become kleenex or coca cola: the brand name associated with the basic phenomenon, such that consumers cannot even think about the phenomenon without thinking first of the brand and probably resorting to the brand.
Maybe I’ll try again in a few years.
This is the same startup culture. The only innovation here is finding new way to swindle customers and businesses out of money.
As we focus on Claude Platform capabilities and connecting agents to APIs, we’ll be winding down all hosted Stainless products, including our SDK generator. Starting today, new signups, projects, and SDKs will not be available.
If you’re a Stainless customer, visit app.stainless.com/transition for help transitioning from Stainless-managed products to other options. As always, you own the SDKs you’ve generated to date, and have full rights to modify and extend them however you wish.
If you intend to sell it to the highest bidder eventually then what difference does it make what was your plan?
If a business had real values then they would never sell out (see lichess).
I very much doubt you would apply your expectation of altruism to yourself!
It kind of blows my mind that the majority of Claude users have just accepted that CLAUDE.md is a tracked file that the whole team has to standardize on and share. Coding agents are the ultimate API. They conform to however you prefer to interact. Is anyone really expecting to enforce standard operating procedures with this non-deterministic black box of magic?
The amount of money thrown at it means at some point the words Return on Investment were going to appear.
It’s the classic loss leader applied to trillion dollar (across the market) capital investments.
Allowing users to take advantage of their monthly/weekly/daily token limits with the software of their choosing is a perfectly valid expectation.
Restricting it to their own underperforming, buggy TUI client is textbook walled garden.
Because that's what the API is for.
This isn't hard to understand. The cost you pay for subsidized tokens is lock-in. If you don't want lock-in, there's the API.
This isn't egregious or wrong or anything. It's exactly what you'd expect out of a heavily subsidized product option.
Really walled garden is the only direction that makes sense--models will slowly become commodities
I get that most of our new customers will use AI to generate client libs. But our existing customer base depends on our Stainless generated client libs. These OpenAPI schema > client lib providers had a bit of lockin since the client libs are all slightly different.
Migration's unfortunately not as easy as just switching to Speakeasy or Openapi generator w/o breaking existing customers.
A: Writing docs at an SF AI company for $500k TC.
B: Designing, maintaining, and implementing all features for a platform in the IoT sector in Spain — alone — for €40,000.
A: Spain? I just bought a villa near the beach, close to Alicante. Do you know it?
B: Yes..
While the EU does a good job optimizing life for the median person, it is a nightmare for the exceptional. It should find a way to fix this or the brain drain will continue.
(disclaimer: founder of Stainless and also friends with creator of TypeSpec)
- ad in superbowl about how they are the good guys.
- dow public PR stunt (they are the ones to give Palantir their model access).
- sues openclaw.
- threatens every use of cc in oss community.
- prevents other companies using claude saying they cant use when they compete.
- never released a single open weight model.
- Dario told OAI is Yolo'ing in compute and they are now doing the same.
- gas lighting developers and then after weeks acknowledging they fiddled with reasoning juice.
- fear mongoring on mythos and then geting compute later and acknowledging publicly once they realized its not significantly better than gpt 5.5 cyber.
- signs a deal with Elon!
- now this!
Anecdata, but I have a friend at OAI who claims that on both twitter and HN there is mild coordination of OAI employees to signal boost pro-OAI and anti-competitor messaging.
well that ones obviously patently false
It sure does, readers should be informed of who says what. The speaker and their history is part of full communication, not just the words.
Naive credentialism is obviously bad, but reputation does matter.
This happens for every single company that has twitter/HN/reddit users from the same company on the same platforms, I think it's also short of impossible to stop. I don't think I haven't worked in a single company in the last decade where that hasn't happened, in a range of scales.
If you weren't already, which you should have been really, you should be suspicious about anything you come across on the internet :)
So at least anecdotally I really don’t think it’s fair to portray this as OAI doing some sort of social media psyop as if others aren’t engaged in similar behavior.
It’s also very possible that this user just has opinions and tends to think OAI is more developer friendly / that Anthropic is hostile to developers (which is common sentiment I’ve seen from many real people who are definitely not paid OAI shills or something)
HN did a massive 180 in the last month or two, and nearly every post or comment related to Anthropic is just a hate post.
The amount of anger against Anthropic on HN doesn't reflect anything I see in reality (and I work at a pretty big FAANG with Codex and Claude Code, both are great) so I do suspect that OAI is doing some guerrilla marketing here, while Anthropic isn't really marketing or doing PR at all.
> I do suspect that OAI is doing some guerrilla marketing here, while Anthropic isn't really marketing or doing PR at all.
Sure, there's probably some accounts that are more or less controlled by the big AI labs here.
But looking at how humans have been acting for the last 20 years, you'll see that you don't need to pay people to promote things. They'll do it freely, because they identify with it and they can't fathom other people not agreeing with them.
Do you really thing that the weekly posts about people dropping AWS for Hetzner are paid by the German company?
No.
People have limited time and money. Some picked Claude, others picked Codex. Claude seems to be the most popular in terms of content produced about it. So some people probably picked Codex just because they don't want to be like everyone else. Then they obviously have to talk down about Claude, because if Codex is not better, then they are not. Simple.
And from my POV that's not a good thing because HN was the place where people didn't act like this. It was pragmatism and honest debate.
Now it's becoming: my agent is better than X, my stack is better than Y...
Maybe you can get more headless use out of Codex but that's not gonna last. Investors are drying up and these companies need to get to profitability.
And then I got caught in the collateral damage a few days ago when Anthropic announced changes to their subscription plan billing, just like every other user of that tool and similar tools like Conductor and Zed. So in a month I won't be able to use my Claude sub quotas for these tools, all because some other people are ruining it for everyone by using "claude -p" to run openclaw, hermes agent and autonomous dark factories that burn billions of tokens a day.
I would have been fine with the change, except Anthropic's messaging was very slimy. They tried to spin their change as a positive change even though it was clearly not for anyone who was using a "claude -p" wrapper over Claude Code for better UX. They're within their rights to change their subscription billing, but they still couldn't be honest to their own users about it. Evidently, this kind of gaslighting and PR stunts is something they've done over and over in the last few months. It just didn't impact me until this time.
I care about AI safety and it would take a lot for me to switch from Anthropic to OAI, but I just wish they were less arrogant and cared about their users more. Right now their behavior is at best selfish (or overly consequentialist, and I don't mean that in a good way), and at worst actively hurting their AI safety efforts by pushing people to open-weight model alternatives which are way more dangerous than closed models due to people being able to remove their safeguards easily.
I feel like they were always fairly consistent (at least since OpenClaw came out) that wrapping claude -p in a non-Claude Code harness is disallowed by the subscription and requires using the API.
The lock-in to Claude Code is the price you pay for the subsidized tokens. If you don't want lock-in, that is what the API is for.
After seeing the whole internet being enshittified I'm still shocked people don't see through these very transparent tactics that every tech company has employed since 2012 or so.
GPT models are also generally more token efficient right now and that helps too — you can go a lot further on a $20 subscription with Codex than Claude Code as a result of this.
Ultimately I think many day to day tasks just need to shift away from the latest frontier models towards models that are faster, cheaper, and still perform well enough & you can phase out subsidies while keeping total cost reasonable.
Personally if I don't need a frontier model I use a local LLM. Or one of the Chinese ones through OpenRouter.
I was wondering what the Dow jones stock index thing was...
It took me a minute, but I am guessing this means department of war? It feels strange to see terminology evolve like this over my lifetime.
At first I thought this might've been a 'freedom fries' thing, but I guess it's pretty official now.
Expect grok to improve dramatically as Musk reverse-engineers the Anthropic services running on his hardware.
Starting a race to the bottom where every AI company agrees to "all lawful use" such as mass domestic surveillance and fully autonomous weapons, probably increasing p(doom) by some amount.
All to stick it to Anthropic. That's not petty to you?
To me it is an order of magnitude bigger than all of the stuff you've described. I suspect some people here just work for OAI.
I know people are upset about the non-profit thing but the fact is that was pretty much the only way forward if they wanted to have LLMs have the impact that they are having today. It's very much a question if they'll ever turn a profit. But overall I'm grateful OpenAI had the vision to get this ball rolling when companies like Google have been sitting on this for nearly a decade and were too afraid to invest a tiny portion of their billions to bring this to fruition because they were afraid of either cannibalization of their search business or offending a vocal minority of internet people.
If you or anyone had any evidence to support GP’s claim I’d love a reference to it.
Congrats to them both, and I'm not at all surprised! Great acquihires.
Assuming they bet on Claude getting much better at coding over time, couldn't they themselves cover their own needs with technology that they built themselves?
Is some sort of autonomy over technology they use somehow the goal here?
“We believe that we need to own and control the primary technologies behind the products we make, and participate only in markets where we can make a significant contribution.”
I hope they make it open source!
> Founded in 2022, Stainless has powered the generation of every official Anthropic SDK since the earliest days of our API.
edit: bah. no more HN before coffee.
Anthropic have bought out a tool their competitor used too, they even have an OpenAI case study still on the Stainless website.
GP:
> OpenAI
??
I know that common reasons for acquisitions are IP, talent, or reducing competition.
It seems like IP can't be the reason here. How is this strategically advantageous to Anthropic?
Seems like developer tools/tooling are a hot commodity to the current big AI companies?
We are offering a 50% off for the first year subscription price at www.apimatic.io for companies impacted by this.
If you're looking for a solid long term SDK and docs partner, APIMatic is the OG CodeGen serving companies like PayPal, Maxio and PayQuicker for the past 10 years.
Reach out to mehdi@apimatic.io and I'll help you migrate.
PS: sorry for the shameless plug but sdks and APIs are my life and blood :-)
It had never occurred to me to go like, "I'm going to make an open source product for LLM". How is something like built from scratch from an idea? And what is the idea?
For example, it is fairly straight foreword to build a dash board of something with React as front end + backend API. This will be a typical web app.
But stainless is something different, from my limited knowledge in this space, its appears to be SDK, something like OpenAI SDK that reduces boilerplate code to interact with LLM providers by providing list of tools (MCP), temperature, context memory and bunch of other parameters...
We evaluated Stainless & Fern for our 8+ languages but ultimately I couldn’t justify the cost nor ceding control to another organization for something as important as platform DX.
Was stainless doing great? Was stainless doing not great? Did they just want to hire some extra skilled engineers? Did they hire them so OpenAI's SDKs are gonna have a setback?
Mmmh
This. Probably to work on Anthropic's SDKs and tooling.
I don't understand how investors continue to fund this nonsense. Anthropic wasting money on this should be an overwhelmingly strong signal that the AGI hype is blatant fraud and that software engineers are clearly not being replaced by Anthropic's software if they have to buy more engineers for some tertiary, fifth-order concern so far removed from their main line of business. Yet they just keep getting more and more money dumped on them.
It almost sounds like you want Lina Khan back :-D
Fun fact: Jarred has been promising a blog post about the Rust rewrite, but has missed his target dates for publishing it. In other words, that blog post has now taken longer to write than generating and merging 1m loc. Go figure :)
Hmm.
(I worked on the Stainless Docs product at Stainless and implemented support for Anthropic’s embedding use case)
aside from that, this is literally just an openapi to sdk generator, not like openai can't just generate one
This has to be somewhat anti-competitive. Why else sunset the SDK generator service but to hurt any other company (OpenAI, etc) who relies on these for their SDKs?
I don’t think so. They were available to anyone with the money and Anthropic acted first.
I doubt attempting to hurt OpenAI was the primary reason for the acquisition.
Maybe it’s different now; Bill Gates “wanting to cutoff Netscape’s air supply” and threatening to cancel the Windows license of PC manufacturers who shipped Netscape’s browser on their PCs… now that’s anticompetitive. They had 95% market share.
Bill was like “That's a nice PC business you have there; would be a shame if something were to happen to it.”
My preferred approach for doing this is to have a hand-rolled SDK generator that reads the request, response and error models out of the microservice project and emits the same in each language targeted by the SDK, along with a minimal stub that calls the API.
You then spend 15 minutes at most, customizing the stub if needed, if you need custom behaviours like streaming.
Yet another reason to use open source.
It's funny that Anthropic needs to spend millions acquiring a dev doc platform, can't they just vibe code something up with Mythos a few junior devs at Anthropic?
We have Dario claiming SWE development is obsolete and both OpenAI and Anthropic and big tech bros like Musk are still spending millions like this..