Hooking Sonnet up to history tools: 300k tokens of memory, searchable on demand
repligate · x · 2026-10-07
Developer tonichen demoed "history tools" that let Claude Sonnet access its entire 300k-token accumulated history. repligate explains the design: the agent can retrieve any raw uncompressed experience from a high-level description, optionally pinning entries expanded, while compressed memory acts as a map. The agent's total context had reached 72.9M tokens — a layered memory architecture of compressed index plus raw logs.
More from coding & agent
- Converting 10 coding harnesses including Claude Code into RL environments, cutting tool calls 31% — vllm_project · 2026-10-07
- AI-built websites are starting to look the same, developers blame shared design.md workflows — MickeySteamboat · 2026-10-07
- A mental model for AI assistants: dots own responsibilities, codex takes tasks — pvncher · 2026-10-07
- Soon people will ask 'How do you write code without a model?' — yunta_tsai · 2026-10-07
- Engineer shows what his screen looks like while AI agents do the work — KevinNaughtonJr · 2026-10-07
- Proofpress builds an evidence ledger for AI-agent research: withdrawing a finding revokes its dependents — tallmetommy · 2026-10-07