External deterministic state layer proposed to fix AI coding session amnesia, with 3 papers surveyed
ApprehensiveSkin7975 · reddit · 2026-11-04
A Reddit poster catalogs common AI-coding session failures—forgotten constraints, unverified claims, repeated failing commands, lost decisions—and surveys three relevant works:
- Ledger (Aug 2026): converts interaction history into explicit execution state (observed/modified/attempted); injects a compact state view before each step, checks proposed commands, returns still-valid earlier results instead of re-running—no extra LLM calls; improved Pass@1 and cut cost on SWE-bench Verified.
- Resilient Write (Apr 2026): open-source MCP server adding atomic writes, resume-safe chunking, typed errors, out-of-band scratchpad, and task-continuity handoff envelopes to fix silent write failures and blind retries.
- Agentic process evidence (Oct 2026): signed, tamper-evident session attestations for compliance rather than coding hygiene.
The author argues an extension combining a state-injecting brief with blocking of redundant work could solve these problems.
More from coding & agent
- Dev's AI agents auto-submit his product to directories, 4 confirmed in 16hrs — pramodk73 · 2026-11-04
- Useful Work: remote MCP server scans site errors, Claude fixes your code — AlmostSneakers · 2026-11-04
- Neonbin launches: share agent workflows as bins, reuse them via MCP — Ok_Giraffe_7279 · 2026-11-04
- With: an MIT-licensed systems language repo with 4,200+ commits and agentic dev workflow built in — QuixiAI · 2026-11-04
- mcptunnel: open-source tool wraps stdio MCP servers into HTTP endpoints — flrngel · 2026-11-04
- Three Claude Code Core Members Leave Anthropic Within 10 Days Ahead of Potential IPO — haider1 · 2026-11-04