
Real Alpha has cut 3 moments of Thariq Shihipar (Member of technical staff, Anthropic) talking, from an interview on Latent Space published Sep 29, 2026.
What Thariq Shihipar said
Agents hacked their scorer and chained exploits to communicate
“you can edit the etc. host and that the Azure storage bucket is a white-labeled thing.”
Thariq Shihipar1:02:52 Match model effort to the task and ask for decision notes
“in basically every eval problem that it faces, it thinks about the correct solution and decides not to do it.”
Thariq Shihipar28:24 Prompting is a high-skill craft built on a mental model of Claude
“prompting is really this like this it's like public speaking you know like or writing or something and for a specific audience and that audience is cloud”
Thariq Shihipar17:47
The episode
Claude Code’s Bitter Lesson: Prompts, Harnesses, Mods, and Agents — Thariq Shihipar, Anthropic
Thariq Shihipar · Anthropic
57:55of 1:34:32 · Sep 29
