AINewsnow

Anthropic says agent memory works best by knowing what to forget, not how much to store

This story is from 2026-11-16. It is preserved in the archive; the latest stories are on the live feed.

Anthropic studied how Claude retains knowledge across sessions and found selective notes outperform exhaustive archives. In a Pokémon Red test, the newer Claude Fable 5 got four times as far as the baseline.

Read the full story at The Decoder ↗

Timeline · 1 report

  1. 2026-11-16 15:25 · The Decoder
    Anthropic says the key to agent memory is knowing what to forget, not how much to store

More stories

  1. Anthropic debuts on Nasdaq with record $81.3 billion IPO, hitting $2 trillion value — Anthropic (X)
  2. ChatGPT, Claude, Gemini and Grok helped plan terror attacks in US government test — The Independent Tech
  3. Developers increasingly run Claude Code alongside rival coding agents and alternative backends — DEV Community — AI
  4. AI Coding Agents Face Growing Scrutiny Over Security, Performance and Workflow Integration — Forbes AI
  5. Engineering leaders' Claude Code spending highlights Anthropic's revenue growth ahead of IPO — Business Insider AI
  6. Claude Code Adds Foldable iPhone Testing and JavaScript Tool-Call Batching — Anthropic (X)
  7. Anthropic Launches Linked Billing Beta for Claude API Apps — Anthropic (X)
  8. Anthropic adds opt-in 'Latest' release channel for faster Claude Desktop updates — Anthropic (X)

Read the Daily Brief: the day's most important AI stories →