Posts warn of data‑center backlash, model fatigue and youth job losses; our tracker shows steady demand in engineering and infrastructure, with OpenAI and Anthropic adding roles.
Several influential voices in the last two days painted a picture of public resistance, model fatigue, and even bad news for the biggest labs. Zvi Mowshowitz declares that “Support for data centers keeps cratering.” Simon Willison relays that “Anthropic’s best AI model struggles to attract users as cheaper tools thrive.” Gary Marcus writes “BREAKING: More bad news for the frontier AI companies.” Azeem Azhar warns, “The canary keeps coughing.”
Our tracker shows something simpler in the labor market. There are 4,234 open AI roles across 174 employers. In the last 30 days, employers opened 412 modeling and engineering roles, 204 data roles, 67 infrastructure roles, 35 research roles, 26 product and design roles, and 9 evaluation and safety roles. Top current openings include Accenture (614 open, 216 opened in 30 days), Capital One (184; 35), Amgen (179; 18), OpenAI (153; 67), Anthropic (116; 35), PwC (115; 43), Waymo (101; 15), and Databricks (89; 19).
Data centers: politics versus postings
Mowshowitz argues that data centers are a proxy target for wider distrust of tech: “Thesis: People Mostly Dislike Data Centers Because They Dislike and Distrust AI, Tech Companies, Big Money And Building Things.” Ethan Mollick, citing @randomwalker, adds a policy angle: “state-level data center bans have very little impact overall on AI progress due to rising efficiency. If your concern is AI development, concentration of power, future use, or anything else other than "is a datacenter near me?" this is not a substitute for policy.”
Our numbers do not register any hiring collapse in roles most closely tied to standing up and running systems. Infrastructure roles are 393 open across 90 employers, with 67 opened and 18 closed in the last 30 days. That is a net expansion. If opposition were already choking investment, we would expect to see a sharper slowdown in new infrastructure postings. We do not see it.
Frontier labs: pressure headlines, expansion footprints
Willison summarizes an FT report with striking claims and a market narrative: “A few interesting numbers in this FT story gathered from "people with knowledge of the matter": Anthropic's "annualized revenue" for July is up to $65bn…” He also quotes that “It also told investors that it had 6,000 customers that spend $100,000 annually or more.” And on OpenAI: “annualised revenue has jumped 35 per cent in the quarter to date and is now over $40bn, with the launch of GPT 5.6 in July jolting the company’s performance after a sluggish start to the year”.
Set those revenue assertions next to our hiring data. OpenAI has 153 open roles and opened 67 in the last 30 days. Anthropic has 116 open roles and opened 35 in the last 30 days. That is consistent with growth, not retrenchment. Marcus’s “BREAKING: More bad news for the frontier AI companies” frames a bearish turn, but our listings do not show a freeze at the labs named in these posts. On hiring, the footprint is expanding.
Pricing and model mix are changing the work, not the demand
Willison also quoted Drew Breunig’s account of how model pricing changes behavior: “Prior to Fable, it felt silly to waste too much time improving your coding harness or context strategies. A new model would arrive at the same price (or cheaper!) and paper over most of your problems. But then Fable landed. It was (and still is!) incredible. But the cost was so high and Opus was good enough (as was 5.6, K3, and even GLM) for most of the code we needed. So we started to think about what work went where.”
If teams are splitting workloads by price and capability, you would expect demand to skew to the people who implement, integrate and optimize. That is what we see: modeling and engineering roles are the largest single family at 1,648 open, with 412 opened and 106 closed in 30 days. Data roles are 819 open, with 204 opened and 22 closed. Product and design sits at 104 open, with 26 opened and 2 closed. These mixes suggest companies are still architecting and shipping, even as they rebalance which models they use where.
A related signal comes from Willison’s own tooling update: “This release of the Anthropic plugin for LLM mainly provides compatibility with the recently released anthropic v1.0.0 Python library, which switches from httpx to httpx2.” Churn in SDKs and runtime stacks tends to show up as hiring for engineers who can keep pace. Our modeling and engineering openings support that reading.
Scientific acceleration, cyber risk, and who is hiring for safety
Jack Clark highlights areas where AI seems to be speeding things up, especially in security. He quotes METR’s finding that “Cyber vulnerabilities: Major acceleration.” and further that “The rate of vulnerabilities reported across many projects has dramatically accelerated in 2026 compared with 2025, both for specific projects (cURL, OpenSSL, Firefox, and Microsoft) and for aggregate vulnerability databases (the US NVD, and OSV)”.
If vulnerability discovery and exploitation are accelerating, we would expect at least some movement in hiring for evaluation and safety or for research linked to defense. Our tracker shows 41 open roles in evaluation and safety across 17 employers, with 9 opened and 2 closed in 30 days. Research roles total 223 open, with 35 opened and 4 closed in 30 days. The numbers are not zero, but safety hiring remains a small fraction compared to engineering and data. On job postings alone, we do not see a surge in safety demand to match the “Major acceleration” Clark cites.
Are young workers losing ground?
Azhar flags a worrying trend: “Employment of young workers (ages 22–25) in AI‑exposed occupations is now 19% below trend. Up from 15% last year.1” He frames it starkly: “The canary keeps coughing.” Noah Smith similarly observes a broader malaise: “It’s hardly news that Americans are not happy with the state of their nation.”
Our tracker cannot validate or refute age-specific employment outcomes. We count current openings, not who fills them. What we can say is that employers are posting a lot of roles that could be early-career entry points in principle, especially in data (819 open) and modeling and engineering (1,648 open). Whether these are accessible to 22–25‑year‑olds is a separate question our data cannot answer.
Hype wars and the developer grind
Some posts this week push back on breathless claims. Emily M. Bender writes, “Everything you talk about tech as "superhuman" at anything you sound as ridiculous as NBC here. But the robots slamming into the padding at the end and falling down is a nice metaphor for something.” Alex Hanna adds, “Well now they’re just saying anything. We dissect Altman’s doublespeak on Mystery AI Hype Theater 3000’s latest episode “Rumors of a Singularity Have Been Grossly Exaggerated”. (with @emilymbender.bsky.social)” Mollick cautions methodologically: “It is not inherently bad to publish research on the impact of AI that only refers to older models, but it requires a very careful discussion and has to be very clear to non-technical readers. 1/” He also notes a more mundane side effect: “Annoying side effect of all the AI bots on the site is that they are all “well-read”…”
The hiring picture is more prosaic than the hype battles. Employers are staffing teams to ship systems and keep them running. Among the top posters, OpenAI and Anthropic both have dozens of newly opened roles in the last month. Professional services and adopters like Accenture, PwC, Capital One, Amgen, Waymo and Databricks are also hiring at scale. That is what a diffusion phase looks like in job postings, even if the discourse oscillates between doom and miracle.
What to watch next
Two tensions stand out. First, politics versus buildout: despite claims that “Support for data centers keeps cratering,” new infrastructure postings outnumber closures. Second, narrative pressure versus operating reality: claims of model adoption challenges and “bad news for the frontier AI companies” coexist with continued expansion in headcount footprints at those same labs.
We will keep testing these claims against the numbers. Our headline jobs-created indicator is calibrated to the World Economic Forum’s Future of Jobs Report 2025 (11M AI roles created, 9M displaced, by 2030). For now, the hiring data we track is clear: 4,234 open roles across 174 employers, with the bulk of new demand in modeling and engineering, significant demand in data and infrastructure, and modest but steady hiring in research, product, and safety.
What we read
Every quote above is taken verbatim from one of these posts.
- Jack Clark: Import AI 470: No rights for machines; automating environment generati
- Zvi Mowshowitz: The American People Really Hate Data Centers
- Noah Smith: What if our biggest problems aren’t economic?
- Simon Willison: Anthropic’s best AI model struggles to attract users as cheaper tools , llm-anthropic 0.27, Quoting Drew Breunig
- Azeem Azhar: 📈 Data to start your week
- Ethan Mollick: Annoying side effect of all the AI bots on the site is that they are a, It is not inherently bad to publish research on the impact of AI that , From @randomwalker.bsky.social: state-level data center bans have very
- Gary Marcus: BREAKING: More bad news for the frontier AI companies
- Alex Hanna: Well now they’re just saying anything. We dissect Altman’s doublespeak, 'I just care about academic freedom and plagiarism' is the new 'it's a
- Emily M. Bender: Everything you talk about tech as 'superhuman' at anything you sound a