Chat_175 - Building in the AI Era with Luís Novo & Matt Ahlborg

— Open Notebook is an open-source research and learning environment that collects sources (PDFs, videos, links) and uses LLMs to extract insights, build AI-assisted notes, and create organized "thinking layers" for knowledge management.

— The Coldcard hardware wallet security incident highlighted why open-source auditing and verification tools are now critical; the speaker rethinks sponsorships and security dependencies after realizing he had not re-verified hardware wallet code despite changing threat landscape.

— AI models (Claude, GPT-4, others) now write and review code autonomously; developers are deferring 90%+ of code review to AI while maintaining test coverage and validation layers, fundamentally changing software engineering workflow.

— Local versus cloud tradeoffs are shifting as Apple silicon, unified memory, and falling compute costs make local inference viable; cloud scaling economics are breaking for providers, creating space for private, on-device AI tools.

— Building defensive and offensive teams (validation vs. shipping speed) is essential for scaling engineering; aggressive development paired with strict testing prevents catastrophic failure while maintaining velocity.

— PairCore/PairDrive represent a peer-to-peer framework for organizing personal data across multiple devices without relying on centralized cloud providers.

— Audit critical infrastructure (hardware wallets, software) using current AI tools; ask LLMs to review code you previously trusted, especially if threat model or tooling has changed since your last verification.

— Set up local, open-source alternatives to cloud AI services (local Ollama instances, Claude via MCP) to avoid censorship, maintain privacy of context, and reduce cognitive load from outsourcing judgment to closed platforms.

— Start small with OpenNotebook or similar tools by picking one extraction type (dense summary, Q&A, contradiction detection) and applying it to existing content collections; automate ingestion via API if you have multiple sources.