Exa launches Agent Ultra: swarm-of-agents deep research claiming SOTA on evals and vibes
garrytan · x · 2026-09-26
Exa introduced Agent Ultra, pitched as a step change in deep research. It orchestrates swarms of agents with heavy parallelization, code execution, and database tools, all coordinated by frontier models, to perform exhaustive research and answer questions requiring thousands of sources. The team says it iterated until the system was clearly SOTA in both evals and subjective quality — likening it to "Codex/Claude ultra mode, but for web research."
Related event: Exa Launches Agent Ultra, Claims SOTA in Deep Research(3 posts)→
More from coding & agent
- Sentinel splits monolithic Three.js into scene-based workflows with HD PNG sequence export — PurzBeats · 2026-09-26
- AI tip: ask models to break your app with worst-case edge data — soleio · 2026-09-26
- Tom Davenport calls to outlaw agent swarms: no human can stay in the loop above 10 agents — eldonredwards · 2026-09-26
- One-shot product demo video made with Opus 5.5 and browser_use video-use — Scobleizer · 2026-09-26
- Theo: model routers can't gauge task complexity from prompts alone; escalation-based routing proposed — nbaschez · 2026-09-26
- OpenClaw 2.0 Hackathon: Build Your Startup's 'First Hire' Agent, Win a Trip to SF and Mac minis — steipete · 2026-09-26