AI Safety Strategy·Exploring·Last reviewed May 1, 2026
This page is a stub. I’ve marked the territory but haven’t written my views here yet. The headings below are placeholders — the actual beliefs, uncertainties, and evidence are still in my notes. If you want my current take on this topic before it lands here, get in touch.
Where I currently stand
<Headline view: how I think the talent pipeline actually functions today, where the obvious failure modes are, and what I'd tell someone trying to break into the field. 3–4 sentences. Maps onto your own ERA experience and the advice you give people.>
Current beliefs
- <e.g. Structured fellowships (MATS, ERA, ASB) are doing the load-bearing work of converting strong generalists into safety researchers; nothing else has matched their throughput.> ~XX% — <one-line why>.
- <Claim about whether mid-career engineers transitioning in are systematically underused versus undergrads with safety-from-day-one trajectories.> ~XX% — <why>.
- <Claim about what "research-taste" actually is and whether it's transferable in <2 years of focused work.> ~XX% — <why>.
Uncertainties
- What's the actual conversion rate from fellowship to durable researcher, and how does it compare to PhD pipelines? Why it matters: changes the cost-effectiveness story for fellowship funding.
- Are governance / strategy / policy careers facing the same bottlenecks as technical careers, or fundamentally different ones? Why it matters: a generic pipeline fix may not work for both.
What would update me
- Long-run tracking of fellowship alumni — five years out, are they still in the field and doing original work? — would push my views significantly in either direction.
- Clear evidence that one specific intervention (e.g. compute access, mentorship hours, paper-writing support) was the binding constraint would let pipeline funders concentrate.
Recent reading
- <date> — <title> — <takeaway>.
Related writing
No essays tagged with this topic yet.