Companion files for the blog post "Keeping interactive tutorials up to date with an orchestrator and AI subagents".
| File | What it is |
|---|---|
readme-tester.md |
Small version: OpenCode subagent that follows a README literally, cannot edit |
readme-fixer.md |
Small version: OpenCode subagent that fixes the failing lines, cannot run commands |
tutorial-executor.md |
OpenCode subagent that walks the tutorial literally in the live console |
| #!/usr/bin/env bash | |
| # agent-state: set the hterm terminal background based on an AI agent's status, | |
| # so ChromeOS Overview shows at a glance which terminals are working vs done. | |
| # | |
| # Usage: | |
| # agent-state <working|done|waiting|error|reset> [agent-name] | |
| # | |
| # States: | |
| # working -> yellow background (traffic light: agent is working) | |
| # done -> green background (agent finished) |
A simple benchmark comparing numpy (vectorized) and numba (JIT-compiled parallel loop) for a sum-of-squares computation over a large array.
On a representative machine the JIT-compiled numba version is roughly 12–15x faster than the vectorized numpy version:
A small benchmark that compares the wall-clock performance of equivalent numeric
operations written with plain numpy and with numba @njit-compiled code.
On a large 1-D float array (10M elements) it compares five operations:
| Operation | Description |