Alignment to Evil
Published on February 21, 2026 3:29 AM GMTOne seemingly-necessary condition for a research organization that…
Recent Articles
View AllReporting Tasks as Reward-Hackable: Better Than Inoculation Prompting?
Published on February 21, 2026 1:59 AM GMTEpistemic status: untested but seems plausibleTL;DR: making honesty…
Robert Sapolsky Is Simply Not Talking About Compatibilism
Published on February 21, 2026 1:27 AM GMTImagine someone wrote a 500-page book called Taking…
TT Self Study Journal # 7
Published on February 21, 2026 1:22 AM GMT[Epistemic Status: This is an artifact of my…
India’s Sarvam launches Indus AI chat app as competition heats up
Sarvam's Indus chat app is currently available in beta.
How will we do SFT on models with opaque reasoning?
Published on February 21, 2026 12:00 AM GMTCurrent LLMs externalize lots of their reasoning in…
Hodoscope: Visualization for Efficient Human Supervision
Published on February 20, 2026 11:41 PM GMTThis is a link post for our recent…
Carrot-Parsnip: A Social Deduction Game for LLM Evals
Published on February 20, 2026 11:06 PM GMTSocial Deduction games (SD games) are a class…
The creator economy’s ad revenue problem and India’s AI ambitions
The creator economy is evolving fast, and ad revenue alone isn’t cutting it anymore. YouTubers are launching product lines, acquiring startups, and…
Can Current AI Match (or Outmatch) Professionals in Economically Valuable Tasks?
Published on February 20, 2026 9:38 PM GMTA Demonstration Utilizing OpenAI’s GDPval BenchmarkSaahir Vaziranisaahir.vazirani@gmail.comAbstractThis project…