Still using Cursor's `thermo-nuclear-code-quality-review` skill regularly:
github.com/cursor/plugins/bl…
Tests are still worth it, maybe more than ever. I use this rule to keep my agents from death spiralling:
At work we enabled GitHub status notifications in Slack.. then disabled it hours later due to spam 🙃
Flirted with tmux, but back with super.engineering. Its solid
Enjoying Pi. Took 10 mins to set up the stuff I liked from CC and now I can run both Claude and Codex in the same TUI/session
Planning an MMORPG with Godot and Elixir. Sol went architecture astronaut over it. 13 grilling sessions, just insane stuff.
Luna on max seems to be doing a good job, staying pragmatic, good questions and more focused.
Eval'd GPT models building a React UI tonight. All test passed, looked similar etc.
Sol/med, no subagents ~$2.50
Sol/med orchestrator, Luna/max implementors: ~$1.85
Luna/max orchestrator and implementors: ~$0.25
Gonna use Luna a lot more
This reduced my initial token use from 33k to 19k.
Cool cool cool
Here's a step-by-step process to kill all the bloat from your Claude Code system prompt:
1. Run a proxy so you can see exactly what gets sent to Claude Code (included in the article)
2. "Fuck, there is so much cruft in there"
3. Use my settings.json to kill all the bloat
Down to a clean 13K tokens to start each session with. Nice.
Full process here:
aihero.dev/how-to-kill-the-b…
Just realised - instead of bloating context with rules that get ignored, Claude can write custom linter rules that enforce my preferences.
e.g. a Rubocop cop requiring resourceful controllers (index, new, create, etc only)









