Turning a Long-Term Learning Roadmap into an Evidence-Driven Feedback Loop
How idempotent daily plans, evidence-gated completion, spaced review, and bounded load changes turn a long roadmap into an executable learning system.
LOGBlog
Engineering Notes
Database-backed investigations, migration lessons, production recovery, and practical AI-assisted engineering.
Page 6 / 9
Making AI Actors More Believable with Enforceable Production Checks
How reviewed fictional actors, actions driven by visible causes, pre-generation frame review, model routing by shot importance, and file hashes make realism checks repeatable and version-specific.
Assigning Incrementing Job IDs Without Duplicating Paid Work on Retry
How a stable allocation key and a process lock let a recurring paid pipeline create several jobs per period without duplicating work when a scheduled run retries.
Filter Before You Rank and Aggregate: Taking a 27-Second Query Down to About One Second
How moving selective business filters ahead of window functions and dependent aggregates cut a paginated Oracle query from 27.49 seconds to about 1.40 seconds.
Hibernating a Cloud Deployment to Zero Paid Resources Without Losing Recovery Proof
A recovery-first procedure for stopping a cloud workload, releasing every paid resource, and preserving evidence that the system can be restored.
A Business Sentinel Is Not Empty Data
A valid sentinel was converted to an empty string, breaking display, required-field validation, and request propagation until normalization was narrowed to true empty values.