meta-research

All models are wrong, but some are useful.

Meta Engineering / Meta Research #

It’s a compelling perspective to see everything as a system. Within this view, systems that generate new systems—what we might call ‘meta-systems’—hold particular importance. While these foundational meta-systems are deeply personal and unique to their creators, a comparative look often reveals common underlying components and patterns that humans frequently employ in their systemic thinking and building.

Following researchers and engineers write about there systems in a lot of detail. I hate the fact that this is highly concentrated in the domain of CS and Machine Learning. But a large majority of the subsystems apply across domains. Check out their writings here :

• You and Your Research (Richard Hamming) — Classic essay on doing significant research.
• An Opinionated Guide to ML Research (John Schulman) — Practical advice for new researchers.
• Principles of Effective Research (Michael Nielsen) — Timeless principles for research productivity.
• Becoming an Independent Researcher (Andreas Madsen) — The struggles of independent researchers.
• Eugene Yan’s Blog — A treasure trove for meta-research and meta-engineering.
  • Design Docs for ML — How to structure and write effective ML design documents.
  • Design Patterns — Patterns for effective research and engineering.
• Applying ML — Curated resources for production ML and research workflows.
• You Don’t Need Another MOOC (Eugene Yan) — Practical advice on learning and building expertise.
• Kilian Weinberger (YouTube) — How to do Research — Insightful talks on the research process.
• Reproducing Deep RL (Amid Fish) — Lessons from reproducing deep reinforcement learning results.
• ML Productivity (Amid Fish) — Tips and frameworks for productive ML research.
• Harvard CS197: Communicating Computer Science Research — A comprehensive resource on research communication.

Note : This is an incomplete list. I wish to keep it updated. Keep checking this page for further edits.