Company Agent Bench: agents know ownership (73–91%) but not people (42–54%)
philhchen · x · 2026-11-13
In this thread, philhchen shares results from Company Agent Bench: on authorship questions about ownership and expertise, company assistant products scored 73–91%, but on questions about a person's interests and tendencies they only reached 42–54%. Agents understand who owns what far better than what people actually care about.
Related event: Company Agent Bench: Understanding People Is Enterprise Agents' Biggest Gap(4 posts)→
More from coding & agent
- Agent commerce startup AgentCard raises $8M from Peak XV and Robinhood to let agents buy from 5,026 stores — yenkel · 2026-11-13
- Keenable 1.0 tops AI web search benchmarks, 330ms queries, 50% fewer tokens — Scobleizer · 2026-11-13
- Designing IL-36 miniprotein binders in a day with Codex and a few dollars of GPU — CatAstro_Piyush · 2026-11-13
- Dual-agent AI harness optimizes chemistry codes, cutting TDDFT cost by up to 94% — CatAstro_Piyush · 2026-11-13
- Claude builds 3D sets via MCP, Nano Banana and Seedance render an action teaser — azed_ai · 2026-11-13
- User moves long Codex sessions to T3, calls it a big upgrade over stock GUI — natesiggard · 2026-11-13