OpenAI and Anthropic jobs: 1,472 openings, 702 for engineers

Short answer: On October 26 2026, OpenAI listed 790 open jobs and Anthropic listed 682, a combined 1,472. Our title count puts 702 of them in engineering and 305 in sales and go-to-market. That answers the Hacker News question of the week ("if the models are so good, why doesn't OpenAI just start 100 companies?"): the labs are not trying to run 100 companies. They are hiring hundreds of humans to sell the tool to the people who will.

TL;DR: We pulled every live posting from both public job boards this morning: 1,472 openings, with posted pay on 1,149 of them. The median posted pay midpoint is $307,500 at OpenAI and $347,500 at Anthropic, and the top bands reach $850,000. About two thirds of the open jobs (978) were first posted after July 28, so the hiring is speeding up, not slowing down, while the same companies sell "AI that writes the code."

Key numbers

What	Number	Source, checked Oct 26 2026
OpenAI open jobs	790	OpenAI on Ashby
Anthropic open jobs	682	Anthropic on Greenhouse
Titles with "engineer", "developer" or "architect"	702 (OpenAI 421, Anthropic 281)	Our count of both boards
Sales, partner, customer, solutions, marketing titles	305 (OpenAI 146, Anthropic 159)	Our count of both boards
Research and scientist titles	142 (OpenAI 67, Anthropic 75)	Our count of both boards
Median posted pay midpoint	$307,500 OpenAI, $347,500 Anthropic	Pay ranges on 618 + 531 postings
Highest posted pay band top	$850,000 (both)	Both boards
Open jobs first posted since Jul 28	978 of 1,472 (66%)	Posting dates on both boards
Base pay if every job filled at the median-ish midpoint	about $492M a year (our estimate)	Our math: average midpoint x open jobs

The question that broke Hacker News has a boring answer

The Ask HN thread "If the models are so good, why doesn't OpenAI just start 100 companies?" sat on the front page all weekend with more than 670 points and 740 comments. Most replies argued about whether the models are actually good.

We skipped the argument and read the job boards. Both labs publish every opening through public APIs (Ashby for OpenAI, Greenhouse for Anthropic), with team, location, posting date and, mostly, pay. You can check every row in the data download below.

What the boards say: these are not companies trying to replace their own staff with their own product. They are classic hypergrowth software vendors, hiring engineers to build, researchers to train, and a sales army to close.

Sales is the biggest team hiring at both labs

At OpenAI the single biggest department on the board is Go To Market, with 149 openings, ahead of Applied AI (120) and Scaling (100). Research proper has 53. User Operations, which is customer support at scale, has 48.

At Anthropic it is the same shape: Sales leads with 121 openings, then AI Research and Engineering (67), Applied AI (63) and Finance (54). Anthropic is also hiring 44 people for Safeguards, its trust and safety team, and 15 for Public Policy.

Follow the money and the answer to "why not 100 companies" writes itself. A company that starts 100 companies has to carry 100 sets of customers, lawsuits and payrolls. A company that sells tokens to a million founders gets paid by all of them and carries none of the risk. Of course they hire salespeople. Selling shovels beats digging.

"AI writes the code" and they still want 702 engineers

By title, 421 OpenAI openings and 281 Anthropic openings are engineering roles. That is 702 engineers at the two companies most loudly telling the world that coding is being automated.

The panic crowd will call that hypocrisy. We read it the other way. The best-equipped engineers on Earth, sitting on unlimited free access to frontier models, still turn into more output per head, so the labs want more heads, not fewer. Leverage does not delete the worker. It raises what each one is worth.

The pay says the same thing. The median posted midpoint is $307,500 at OpenAI and $347,500 at Anthropic. The top band at both companies is $850,000, on research and GPU performance roles. Nobody pays $850,000 a year for a job they expect a model to do next quarter.

The hiring is speeding up, not winding down

Two thirds of today's open jobs were first posted in the last three months. At OpenAI, 127 still-open postings date from August, 209 from September and 183 from the first 26 days of October. Anthropic shows 144, 146 and 158 for the same months.

One honest caveat: older postings drop off once they are filled, so a snapshot always skews young. Even so, a still-open count that triples from July to September at both labs at once is not a hiring freeze. It is a land grab.

Who wins and who loses

• Wins: anyone who can sell to enterprises. 305 sales, partner and solutions jobs at two labs is a hiring market of its own, and the labs are funding it with customer money.
• Wins: the founder with a laptop. The labs chose not to build the 100 companies. That leaves the 100 companies to you, built on the same models they sell for cents per million tokens (see our LLM price war breakdown).
• Loses: the doom narrative. If the people closest to the models are on a hiring spree, "AI will take every job by Christmas" is a story someone is selling, not a forecast.
• Loses: the incumbents outside tech. As we showed in Mag 7 vs the S&P 493, the payoff so far lands with the people who ship software, not the ones who buy a seat and wait.

Why the labs will never start your company

Our bet, and nobody has proven it yet: the labs stay vendors on purpose. Every company they launch would compete with a paying customer, and the 978 new postings since July are built around serving customers, not competing with them. The OpenAI for Gov team (15 openings) and Forward Deployed Engineering (30) are the tell: engineers who sit inside other organizations to make the product stick.

That is the opening. The 100 companies the Hacker News thread asked about will be started by individuals with leverage, and the labs are staffing up to sell to them.

More on the business of AI on our news desk, the AI industry hub and our LLM guides.

What this means for you

Stop waiting for the labs to build the product you have in mind. They just told you, in 1,472 job posts, that they would rather sell you the model. Pick one of the 100 companies, ship it this month, and use the same tools their $307,500 engineers use.

If something in your AI-built app is broken and holding up the launch, our paid fix gets it working: fixed within 24 hours or a full refund.

Not financial or career advice; check the live boards before you apply anywhere.

Sources

• OpenAI job board, Ashby public API, pulled Oct 26 2026
• Anthropic job board, Greenhouse public API, pulled Oct 26 2026
• Ask HN: If the models are so good, why doesn't OpenAI just start 100 companies?, Hacker News, October 2026
• Tensorust data download: every posting with title, team, location, first-posted date and pay band, Oct 26 2026