Yadnesh Salvi

Story · @omarsar0

Meta research: agents learn harness policies instead of hand-authored harnesses (@omarsar0)

X post · Story page

New Meta research has agents learn harness policies offline, then deploy them to construct and update external harness state online, instead of relying on hand-authored harnesses that are hard to tune for long-horizon tasks.

In plain words

  • Meta researchers taught agents, artificial intelligence systems that act independently, to manage the supporting setup they need for lengthy tasks.
  • During preparation, each system learns rules for building and changing information kept outside itself.
  • When working later, it applies those rules as the task changes, replacing a fixed setup written by people.
  • This could help system builders tune lengthy tasks because the supporting setup can adapt instead of remaining hand-built.

Appeared in

Subscribe

Get the brief in your inbox

Pick daily, weekly, or both. Nothing is gated either way: every issue is on the site and in the feeds.

  • Weekdays at 8:45am IST, one lead story and 6 to 9 items.
  • Sundays, an argued synthesis rather than a recap.
  • One click to leave, and quiet days say so in the subject line.
How often

Weekdays 8:45am IST + Sundays. Unsubscribe in one click.

You're asking for The Agentic Brief by email at the cadence you picked. You can unsubscribe in one click from any issue, and your address is never sold or shared.

Subscribe by RSS →