Weekly Review: A Belief Dressed as a Gate

Two goals this week: apply every day, and close the cloud gap — Kubernetes, Terraform, CI/CD. Both scored partial. Three applications went out, and my RAG project went from “runs on a laptop” to a containerized app on GKE, provisioned by Terraform, live on a domain I now own. Steps five through seven — CI/CD, evaluation, the write-up — carry to next week.

Last week’s review found a real gate: applications deliberately held until a project shipped, because the resume wasn’t credible without it. This week I caught the counterfeit version. “The cloud deploy has to finish first” held my applications back for roughly four days — and when the review forced the question, I had to admit nothing in those applications actually required the deploy. A real gate blocks the work; a belief dressed as a gate just reorders it to feel better. They look identical from inside the week. The only test I’ve found is to ask, per blocked task: what concretely fails if I do this today? This week’s honest answer was “nothing.”

The week’s loudest fight was a screening video in spoken English for a freelance platform. The deadline re-sequenced everything, the language gate stretched two days of work into three, and the recording shipped late — but it shipped, and the practice rig it forced me to build (paid shadowing tools, custom-text TTS) stays.

The recall blind spot repeated, on schedule: I could name every visible artifact, but the commit log had to remind me that my job-hunting harness got a full audit this week — which found that 23 of 32 “live” postings in my tracker weren’t real anymore. The system that counts my funnel can now be falsified, not just incremented. That’s the biggest single improvement of the week, and I retrieved none of it from memory.

The keeper frame, pulled from using my own decision tooling: a plan is a prediction, not a promise. A broken promise only produces guilt; a wrong prediction produces data — record the miss, adjust the model, forecast better. Next week runs on that: one fixed morning spine — orient, apply (two-hour cap), twenty minutes of spoken practice, then build — with a no-exceptions rule and a mid-week slot where practice becomes a live drill. The plan is filed as a prediction. Saturday checks it.