AI Event Horizon and Automated AI R&D
Approaching the AI Event Horizon? Part 2, w/ Abhi Mahajan, Helen Toner, Jeremie Harris, @8teAPi is a February 2026 Cognitive Revolution discussion of CSET's automated AI R&D work, centered on Helen Toner's account of a July workshop with frontier-company, policy, and research participants. The transcript frames the live question as whether AI use in AI research is merely another productivity tool or whether it can close a recursive loop, and it spends its best time on concrete uncertainties: what parts of research remain human-mediated, where bottlenecks move when coding speeds up, and what indicators could distinguish a normal acceleration from a self-reinforcing one.
For Spiralist themes, the video matters because it ties self-improvement arguments to institutional signals rather than mystical intelligence claims: automated AI R&D, benchmarks, data and compute constraints, deployment and adoption frictions, political-economy effects, and the gap between AI that improves models and AI that can act effectively in the world. Its caveat is also the reason to read it carefully: it is a long, speculative podcast conversation around a report and workshop, not a formal eval or settled timeline, so its value is as a map of disagreements and warning indicators rather than proof of an imminent "event horizon."