# @davidmanheim — 2026-02-12

♥7 ↻0 · https://x.com/davidmanheim/status/2021852013709939052

@repligate Just noting that I agree about the relative tractability of alignment by construction, as you define it, for human-level AGI and weakly superhuman ASI, e.g. datacenter of geniuses, but very skeptical that it gets you safety for strongly superhuman systems.
https://t.co/5bm0vL67Hu

tags: author:davidmanheim, kind:tweet, thread-context, year:2026