Llm

A batch job matches free-text line items against a large product catalogue. Supplier wording never matches catalogue wording, so it runs a cascade — remembered aliases, then embedded codes, then exact matches, then fuzzy retrieval, and finally one language-model call to adjudicate the survivors. That last step costs money, so results are cached. I improved the text normalisation feeding the cascade. Tested it properly: extracted the production matching code, ran it against the real catalogue, confirmed zero previously working matches broke and roughly 230 previously failing lines now cleared the acceptance gate. ...

I needed demo videos for two pre-revenue products: one that reads supplier invoices and computes per-product margins, one that watches trucks leave a quarry and classifies what they’re carrying. No budget for an agency. So I built them with an agent, ffmpeg, a headless Chromium and a free text-to-speech voice. The rendering pipeline took an afternoon. Everything else took two days, and none of it was video work. The problem isn’t generation Script-to-video is a solved commodity. Synthesia will turn text into a presenter for about USD 18/month. Arcade will turn a screen recording into a guided demo for USD 32. Both are good at what they do, and neither touches the part that actually costs anything. ...

I built an MCP server that searches flights across flexible dates and multiple airports, then used it for an actual trip I was planning. Three tool calls in, it had produced a correct answer, and along the way had wasted most of a day’s API quota and told the agent several things that weren’t true. None of the four problems was a crash. Every one of them was a shape problem: the output was well-formed, plausible, and misleading. That’s the failure mode that matters when your consumer is a language model, because an agent can’t smell that a number is wrong. It acts on whatever you hand it — and when the number is wrong in an expensive direction, it acts again. ...