Skip to content
Muhammad Musa

Notes from the work

A question. Then a closer look.

Things I’ve built, ideas I’ve tested, and the decisions I’d make differently. The technical details, with the reasons behind them.

Simulation / Optimization / Computer vision / Game systems / Simulated annealing · 7 min read

Three cats. Fifteen turns. One useful model.

Building a screenshot-to-plan tool taught me that a search algorithm is only as useful as the world it is searching. Try the planner, then follow the decisions behind it.

Reading notes / Anthropic / Dynamic workflows / Agents / Evaluation / Orchestration · 5 min read

Dynamic workflows still need a stopping rule

Reading Anthropic’s dynamic-workflows article alongside the decisions I made in Resume Gauntlet: separate the work, verify the artifact, and make failure count.

Open source / MCP / TypeScript / Performance / Testing · 4 min read

Only send the schema a tool needs

A Notion MCP pull request reduced the tool-list response by pruning unreachable definitions. The interesting work was proving what remained.

AI safety / Evaluation / RAG / False positives / Research · 5 min read

When a safety score hides the cost

A defense reached zero attack success in one test. The more useful question was what happened to the clean answers.

Knowledge systems / Search / Provenance / Product engineering / Publishing · 5 min read

A bookmark is not a research system

What I changed in Kiln to keep ideas useful without losing the evidence, the uncertainty or the original source.

Graphics / Three.js / Shaders / Interaction design / Debugging · 4 min read

A spotlight without the rings

A small visual defect in this site turned into a useful distinction: lighting the background and revealing its geometry are different operations.