Karpathy Skills vs MemPalace

Karpathy Skills is a CLAUDE.md plugin distilled from Andrej Karpathy's public observations on LLM coding pitfalls. Drop the single file into your project root (or install it as a Claude Code skill) and every Claude Code session starts pre-loaded with the four principles Karpathy identified as most commonly violated: think before writing, prefer simplicity, make only targeted changes, and close loops with explicit verification. The project has accumulated 1,450+ GitHub stars in under two weeks. The implementation is intentionally minimal — it's a structured system prompt, not a framework. Each principle is spelled out with concrete anti-patterns to avoid: no premature generation, no over-engineering simple tasks, no cascading refactors when a surgical fix suffices, no ending a session without verifying the goal was actually met. It's Karpathy's "Software 2.0" thinking applied to the agent workflow meta-layer. What makes this compelling isn't the technology — it's the curation. Karpathy has spent more time thinking about LLM behavior patterns than almost anyone outside the major labs. Packaging that into something installable in 30 seconds lowers the floor for teams who want more reliable agent outputs without extensive prompt engineering work.

MemPalace is a free, MIT-licensed AI memory framework that stores LLM conversation data verbatim locally — no AI summarization step, no per-query API costs. It integrates with Claude Code, ChatGPT, and Cursor via MCP, and claims the highest LongMemEval benchmark score among free memory frameworks at 96.6% (initially claimed 100% before community pressure forced a correction after GitHub issue #29 exposed test-set tuning). The project went viral on GitHub with 23,000+ stars in under 48 hours, partly because it was built by actress Milla Jovovich and developer Ben Sigman — an unusual origin story that dominated early coverage. But the technical pitch is real: competing paid solutions (Mem0 at $19–249/month, Zep at $25+/month) do similar things and charge for the privilege. MemPalace runs fully local, connects to any POSIX filesystem, and the verbatim storage approach avoids hallucination artifacts introduced by AI-summarized memory. The catch: verbatim storage means much higher storage overhead than summarization-based approaches, retrieval latency grows with context size, and the benchmark controversy raised questions about the team's methodology. For personal projects and small teams, the zero-cost angle is hard to argue with. For production systems where memory quality is critical, wait for independent benchmarking.