Research Library
The homepage should decide the next learning move for you: continue a module or read one strong article now.
From Engineering Fundamentals
20 min • 0% through this track
Use expensive frontier models to map bugs, plans, architecture, and verification contracts; let cheaper agents execute bounded work with an inspectable trail.
Recommended as a strong standalone read that can move your model forward quickly.
Use expensive frontier models to map bugs, plans, architecture, and verification contracts; let cheaper agents execute bounded work with an inspectable trail.
Clark and Chalmers' 1998 paper: Otto's notebook is part of his mind, not just a tool. If cognition extends into environment, tools, and other people — then workspace design is cognitive intervention, externalising thought is extending the system, and AI assistance is restructuring how you think. The most radical and practically useful idea in philosophy of mind.
Reliable systems do not assume actions happen exactly once. They are built so repeated actions are safe. That design property is idempotency, and it sits underneath payments, retries, job systems, and failure recovery.
Once systems live on more than one machine, timestamps stop being enough. Vector clocks help distributed systems track causality: not just when something happened, but what could have influenced what.