Governance·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: compute is the most legible and most enforceable handle on frontier AI development; FLOP thresholds buy you a real regulatory tool but only if the threshold tracks the underlying risk, which it does poorly today; the most interesting near-term work is hardware-enabled mechanisms because they let you decouple regulation from training-stack details.>
Current beliefs
- FLOP-based thresholds will become rapidly less useful as algorithmic efficiency improves. ~XX% — <why>.
- Compute governance is currently the strongest available regulatory tool in expectation, even with that flaw. ~XX% — <why>.
- <Claim about export controls / chip controls / KYC.> ~XX% — <why>.
Uncertainties
- Does compute governance survive a decentralised training paradigm? Why it matters: pivotal for the long-term viability of the whole approach.
- Can credible cross-border enforcement actually be built, or does this only work as a single-jurisdiction tool? Why it matters: determines whether compute governance is national or international.
What would update me
- A successful demonstration of cross-jurisdiction compute attestation would shift my view on international tractability.
- A mainstream training paradigm that materially weakens the FLOP-capability link would force a reframing.
Recent reading
- <date> — <title> — <takeaway>.
Related writing
No essays tagged with this topic yet.