Mnemos MCP author preps V3 rework: no setup, no multi-call turns, full engine complexity
RileyRalmuto · x · 2026-09-27
The developer behind Mnemos MCP says several updates shipped this week and a fundamentally reworked V3 is underway, aiming to restore much of the original Cognitive Engine's complexity without requiring user/agent setup or multiple model calls per turn. They claim three "very significant breakthroughs" with input from Superhuman Intelligence, and a team running recursive testing to tune combinatorial mechanics. The author also argues models can learn and evolve over time without weight access. The post is largely a teaser with no concrete benchmarks or technical details, so claims remain unverified.
Related event: Mnemos MCP Announces Fully Rebuilt V3 With Zero-Config Setup(2 posts)→
More from coding & agent
- Viral demo: Claude Opus 5.5 nails a color-fill doodle animation in pure JavaScript, first try — sven_ai · 2026-09-27
- Ops optimization for an AI staff team: human bandwidth is the bottleneck — RachelVT42 · 2026-09-27
- MongoDB launches Agent Skills to help AI coding assistants write better database code — TheTuringPost · 2026-09-27
- GitHub's Ash: agents are running a DoS attack on code reviews — jonmarkgo · 2026-09-27
- Using Codex Computer Use to monitor Claude Code's progress end-to-end — vista8 · 2026-09-27
- Local agent eval: challenger model wrote tool calls as prose 5/80 times — a silent failure class — Grimmoner · 2026-09-27