Signal SentrySearchUpdated Nov 2, 17:30 UTCPM Drop
Agents · industry

Agents in General

Agent deployments and launches in General, from the Drops.

From the Drops

AgentsGeneralTestingHuman in loop: unclear

Preprint: goal-stack recovery lifts AppWorld success up to 22.07 points over ReAct

Researchers report the Verify-Backtrack Loop, a training-free ReAct extension, keeps subgoals in a bounded stack and checks feasibility on failure. It beat ReAct and reflection on every model and benchmark pair they tested.

Why it matters: Long agent runs fail at recovery, not length, the authors argue. If your agent retries blindly after errors, goal tracking in the harness is a cheap, training-free change to test.

On the Frontier watch →
arXiv · Nov 2
From Drop #062 · Mon, Nov 2, 2026 · Evening
AgentsGeneralAnnouncedHuman in loop: yes

OpenAI to merge ChatGPT's Chat and Work tabs; Codex will share the token allowance

TechCrunch reports an OpenAI email says Chat and Work will merge, first for select users. ChatGPT and Codex will draw from one token pool, and OpenAI says some users may hit resets more often.

Why it matters: Teams that lean on Codex may find chat use counts against the same limit. Check how heavy agent users on your plan will be affected before the change reaches you.

TechCrunch · Nov 2
From Drop #062 · Mon, Nov 2, 2026 · Evening