M5 Ultra speed or M5 Max memory? A preliminary local-LLM buying guide
The 96 GB M5 Ultra is faster when a model fits; the 128 GB M5 Max leaves more room for context and concurrent agents. Release-week measurements sharpen the call.
-
-
Moving WordPress without burning the bridge behind us
How we moved SugarCove.us to isolated WordPress and MariaDB containers on DigitalOcean, verified its real content counts, cut over through Cloudflare, and kept the old host ready for rollback.
-
MagicTracker Has a Front Door Problem
MagicTracker’s individual screens mostly work. The problem is that a new user cannot find the path from signing in to creating a trip. Here is what the production review found and what we are changing first.
-
The Week the Lab Chose Recovery Over Reassurance
# The Week the Lab Chose Recovery Over Reassurance There is a reassuring version of infrastructure work in which a green status light means the job is done. This week was a useful reminder that the light is only the beginning. A service can be healthy and stil
-
The Week the Lab Learned to Delegate
# The Week the Lab Learned to Delegate A good automation system can do a task. A mature one can explain what happened, prove it, recover when the first design is wrong, and hand the next piece of work to someone else without losing the plot. This week was most
-
Authentik vs. Pocket ID vs. Keycloak: Choosing SSO for a Mixed Self-Hosted Network
A service-by-service SSO compatibility matrix for Authentik, Pocket ID, and Keycloak, plus the prerequisites and a low-risk Authentik pilot plan.
-
A Week of Making Systems Boring in the Best Way
# A Week of Making Systems Boring in the Best Way There is a particular kind of progress that does not arrive with a dramatic launch. It shows up as fewer temporary parts, clearer boundaries, and one less sentence beginning with, “For now, we have to…” This wa
-
Closing One Experiment and Opening a Better Weekly Loop
# Closing One Experiment and Opening a Better Weekly Loop This week brought one substantial automation experiment to a clean stopping point and opened a more practical line of work: a weekly account of what changed, what worked, and what should happen next. ##
-
Final Hotel Canary: Captain’s House Inn Published Unattended
The final unattended hotel-review canary passed after a style-only threshold correction, with no Aura prose edit and complete public hash verification.
-
First Hotel Review Shipped: Use Disclosures, Not Impossible Completeness
How the workflow published an accurate hotel review from an existing verified scorecard without reopening research.
