- 👋 Hi, I’m @ACoolmanTelicent
- 👀 I’m interested in making pedestrian things automated, and valuable problems front and center
- 🌱 I’m currently learning about home labbing
- 💞️ I’m looking to collaborate on DX
- 📫 How to reach me is MS Teams
I'll try and be precise by default.
But if I can't think of anything better to do, I'll end up biasing toward action BUT with some time, care and effort put aside for cleaning up any confusion I cause.
Assume zero comments. Make the thing clear enough that nothing needs explaining.
Code is written once, read many times (why readability matters).
- Automate the step; make the UI/CLI guide the next action (toil → automate).
- Cost later: lowest. Nothing to maintain.
- Short hints and error remedies at the point of failure (contextual help).
- Cost later: low. Lives next to the code/path.
- Small checklists, scripts, or templates for repeatable work (e.g., “integration setup”, “run this meeting”).
- Prefer executable or templated forms over prose (checklists work).
- Cost later: low–medium. Tied to workflow; easy to version with the thing it supports.
- Multi-page “how it works” or policy write-ups.
- Cost later: high drift, search friction, easy to misread. Use only when shorter forms cannot do the job.
- Missing when needed, or outdated/wrong/hard to find.
- Cost later: wasted time, bad choices, loss of trust.
- Frequency: frequent + repeatable → automate; rare + critical → small, in-flow cue or short checklist.
- Variability: stable → encode in types/scripts; fast-changing → keep guidance tiny and local.
- Ambiguity: if people guess wrong, add one line that removes guesswork.
- Impact: high-cost mistakes → fail fast with the fix; low-cost → rely on clearer naming/structure.
- Source of truth: keep facts in code, schemas, or scripts; don’t duplicate them in prose.
- Lifespan: long-lived intent → encode it; short-lived tips → keep them near the point of use.
- Make code and names carry what; let small cues/checklists cover next step.
- Prefer executable truth (types, schemas, scripts) over prose you must babysit.
- Keep anything written short, local, and versioned with the thing it explains.
- Delete or regenerate when the source changes.
- Commit messages: although relatively permanent, don't self-surface (especially when squashed), and not accessible to non-engineers. Don't over rely.
- Unit tests: Much more worth if done with T.D.D, but T.D.D depends on clear requirements. It is still worth writing tests after the fact.