Arav Srinivas: building a multi-model, multi-agent harness toward an automated researcher
AravSrinivas · x · 2026-11-18
Arav Srinivas says he intends to keep building on a multi-model, multi-agent harness for harder tasks: answering open-ended questions, designing experiments, building hypotheses, testing and ablating them in parallel, and incrementally refining predictions — all aimed at an automated researcher.
He adds that multi-agent setups aren't just about cost and performance but raw wall-clock speed: parallelism and higher credit spend translate into real, measurable time saved.
Related event: Perplexity Finds Mixed-Model Agent Teams Beat Identical Model Copies(4 posts)→
More from AGI Musings
- Chamath Palihapitiya: 'All existing software needs to be rewritten' for the AI age — Polymarket · 2026-11-18
- FAI Publishes 'Futurecraft': 15 Policy Essays on Governing the Day After AGI — jachiam0 · 2026-11-18
- Researcher proposes 'Qualiaphobia': giving AIs justified belief in their non-consciousness — erikphoel · 2026-11-18
- Twitter users surprised by how many people welcome being 'succeeded by the machine god' — erikphoel · 2026-11-18
- AI infrastructure spending now dwarfs railroads and the Internet combined, but civic investment lags — blaiseaguera · 2026-11-18
- Reverse Alignment coalition proposes identity verification, Polis-style consensus and data attribution upgrades — blaiseaguera · 2026-11-18