log

Week of 2026-10-12

What I did

• Set up this repo as a real learn-in-public system: Astro site on GitHub Pages, four content types (blog, notes, log, projects), templates, and a weekly cadence — 1 published post per week, logs are extra.
• Completed the foundation task (Task 0 / Score 5): took bin/miren from “works on one laptop” to a proper systemd service behind HTTPS on EC2. Wrote up learnings in a blog post (SSM tunneling, systemd with Restart=always + enabled at boot, stale binary debugging, containerd root requirement and tradeoffs, TLS via Let’s Encrypt + Cloudflare DNS, kill-9/reboot resilience proof).

What broke

• Service kept going inactive (dead) after start due to an outdated/stale bin/miren binary.
• Tried to run as a strict non-root service user, but containerd requires privileged/root access for managing containers — not possible for this setup. Had to document the constraint and compensating controls instead.

What I learned

• The proof is what survives: kill -9, then a reboot, then the laptop dying. Anything less is a demo, not a deployment.
• Verify the binary (hash/timestamp) before blaming systemd. SSM port forwarding is handy when direct SSH isn’t exposed. Least privilege is the goal, but real runtime constraints matter.

Next week

• Finish the foundation task write-up (first blog post).
• Get the site deployed and the first post published.