How to make Claude Code remember its own mistakes
Claude Code forgets everything when the session ends. The setup I use: a post-mortem command that writes to CLAUDE.md, skill files for the quirks of the stack, and a daily pass over the logs.
Tag: Backend
Vibe coding a podcast app in a weekend, and what it costs
PodClip was built over a weekend with Replit: Spotify OAuth, Whisper transcription, Postgres. The scaffolding held up. The Spotify playback limits and the transcription costs are where a human had to step in.
Agentic RAG failure is a loop bug with a bigger bill
Agentic loops fail quietly: the same query retried with new wording, retries with no circuit breaker, a context window full of raw JSON. Three patterns, how to measure them, and a PHP budget gatekeeper.
How WordPress Playground MCP connects AI agents to your site
WordPress Playground MCP connects Claude Code or Gemini CLI to a Playground instance running in your browser. The agent reads files, runs PHP and queries the database itself, which cuts out most of the copy-paste during plugin testing and debugging.
How OpenAI prompt caching actually cuts your token bill
OpenAI prompt caching takes 90% off repeated tokens, but only for prefixes over 1,024 tokens and only when the static part of your prompt comes first. Here is a Python example plus the ordering mistake that kills cache hits.
Nonlinear constrained optimization with PWL approximations
How to linearize a nonlinear objective with piecewise linear approximations, keep the segments adjacent using SOS type 2, and hand the result to an ordinary LP/MIP solver like Gurobi.
What managed WordPress hosting fixes in an agency stack
Most agencies build their own stack and end up debugging Nginx instead of shipping. What plugin bloat really costs, and what Encircle Technologies changed when they moved to managed hosting.
Causal inference and the accuracy trap in ML
A readmission model with 94% accuracy made things worse because it found a correlation, not a cause. Where Pearl’s ladder of causation comes in, and how DoWhy’s four steps keep you honest.
Why NumPy and Pandas give different variance results
Same ten numbers, two different variances. NumPy defaults to the population formula and Pandas to the sample one. Here is what ddof does and how to set it so your pipeline and your prototype finally agree.
Handling outliers and missing values in credit scoring
Mean imputation and unclipped outliers are what break credit risk models. This covers splitting first to avoid leakage, IQR clipping with training-set bounds, and what to do when the missingness itself is informative.