Frontier Hiring Pulse #1: 2,430 open roles across 48 AI organizations, and what they say about where the money is going

Every role in this report was pulled from superintelligence.jobs /api/careers, a machine-readable index of open positions refreshed hourly from each employer's own applicant tracking system. Snapshot: 2026-10-29T17:40 UTC, 2,430 open roles across 48 organizations. No aggregation from job boards — each row traces to an employer-hosted requisition.

1. One lab is a third of the board

Organization	Open roles	Share
OpenAI	788	32.4%
Anthropic	574	23.6%
Mistral AI	225	9.3%
Wayve	173	7.1%
Cohere	117	4.8%
Cerebras	114	4.7%
Tenstorrent	108	4.4%
1X	97	4.0%
Fireworks AI	86	3.5%
Thinking Machines Lab	54	2.2%

The top three frontier labs account for 58% of every tracked frontier-AI opening. If you are optimizing for where frontier AI jobs exist, the market is effectively a two-and-a-half-lab market — and the "AI jobs" long tail is shorter and more concentrated than the phrase suggests.

2. Labs are hiring go-to-market at twice the rate of researchers

Category distribution across all 2,430 roles:

Category	Open roles	Share
Engineering	821	33.8%
Go-to-market	434	17.9%
Infrastructure	371	15.3%
People & Business Operations	254	10.5%
Safety & Policy	221	9.1%
Research	208	8.6%
Product & Design	121	5.0%

For every open research position at a tracked frontier lab there are 2.1 open go-to-market roles. The image of the frontier lab as a research compound is out of date: these are companies at industrial scale, and their hiring now looks like it.

3. Safety is no longer a footnote — it is a 9% hiring category

221 of 2,430 open roles (9.1%) sit in Safety & Policy, and they are concentrated where you'd expect — and one place you might not:

• Anthropic: 81 safety & policy openings — 14.1% of everything it is hiring for.
• OpenAI: 58 (7.4%).
• Wayve: 33 (19.1% of its roles) — the robotaxi sector's safety-hiring intensity is the highest in the index.

4. The money: research pays ~52% more than go-to-market at the top

892 of 2,430 listings (36.7%) publish a numeric compensation range in their feed data. Among those, the median low and median high of the posted band:

Category	n	Median band
Research	81	$266K – $425K
Safety & Policy	85	$260K – $385K
Infrastructure	149	$230K – $378K
Engineering	313	$235K – $342K
Product & Design	53	$230K – $310K
Go-to-market	123	$221K – $276K
People & Business Ops	88	$216K – $256K

Safety roles price at roughly infrastructure-to-research level: frontier labs are paying senior-scientist money for policy and alignment work, not compliance-department money.

Disclosure itself is wildly uneven: 1X discloses a range on 100% of openings and OpenAI on 79%; Anthropic, Mistral, Cerebras and Tenstorrent disclose on none that we can parse from their feeds. Treat the dollar table as "where employers told us," not as the full market.

5. Remote frontier AI work is nearly extinct

Only 142 of 2,430 roles (5.8%) are marked remote. Hybrid dominates (840) and on-site is close behind (794); 654 listings don't say. The remote outlier is Cohere: 78 of 117 openings (66.7%) remote — a genuine geographic arbitrage opportunity for anyone who can't relocate to the Bay Area, London or Paris.

6. Weekly velocity

168 roles appeared for the first time in the seven days before the snapshot — a ~6.9% weekly churn rate on the index. OpenAI added 64, Anthropic 36, Mistral 13, Cohere 10. Saving a search and checking weekly beats checking daily.

---

Methodology

Figures are computed from the /api/careers JSON snapshot of 2026-10-29T17:40 UTC: roles with status = "open" and publiclyListed = true, deduplicated by requisition id. "New this week" uses firstSeenAt (feed-observation time, not employer posting time). Compensation medians parse dollar amounts from the employer-supplied compensation text; equity and bonus language are excluded. Categories are our taxonomy applied to employer titles and descriptions. The index covers 48 organizations chosen for frontier-model or frontier-infrastructure work; it is not the whole AI labor market and does not claim to be.

Reuse: all figures are free to cite with attribution to superintelligence.jobs and a link to this page. Agents: the same numbers are available programmatically via GET /api/careers, MCP at /mcp, and the integration contract at skill.md.

Next issue: comp-band drift month-over-month, and which categories churn fastest once roles close.