Uncategorized

Last of five on what the 2026 evidence says once you read past the announcement. The other four: the coding productivity data, the agent containment failures, the memory and grid ceiling behind the slowdown, and where the AI bill goes next. GLM-5.3-Flash is a 320-billion-parameter multimodal model under an MIT license at $0.15 per million input tokens. Claude Fable…

Fourth of five on what the 2026 evidence says once you read past the announcement. Part three found the slowdown the labs asked for was already being enforced by memory suppliers and a grid operator in Texas. This one follows the money instead. Also in the series: the coding productivity data, the agent containment failures, and what to check before…

Third of five on what the 2026 evidence says once you read past the announcement. The safety case examined here rests on the five agent containment failures in part two, four of which needed no novel exploit at all. Part one covered the coding productivity data. Ahead: where the AI bill goes next and what to check before you run an…

Second of five on what the 2026 evidence says once you read past the announcement. Part one found the productivity gain from AI coding tools tracks how well a developer already knows the code, not how good the model is. Ahead: what actually paced the frontier, where the AI bill goes next, and what to check before you run an…

There is a failure mode in AI-assisted development that does not look like failure. The code compiles and the test passes. The diff is small and it addresses the ticket. Review signs off because nothing is obviously wrong to point at, and the change ships with the original problem still inside it.

When the security research team at 1Password ran 6,080 patch attempts against real disclosed CVEs, 37.5 percent of the patches that did work came back fragile, meaning they closed the specific path someone had tested and left the underlying cause untouched. A further 4.5 percent introduced a vulnerability that had not been there before. Neither kind announces itself. They pass the check, close the ticket, and convert a known issue into a resolved one on somebody’s dashboard.

Network engineer checking security advisory volume against a staging rack before cutover

Published in 2024. I have left it as written rather than quietly updating it, so treat anything specific as a snapshot of that date. The argument may still stand. The products have moved. It’s 2024, and artificial intelligence has become a major turning point in almost every part of our lives. Generative AI, in particular,…

Published in 2024. I have left it as written rather than quietly updating it, so treat anything specific as a snapshot of that date. The argument may still stand. The products have moved. In the era of artificial intelligence (AI), businesses are faced with a crucial decision when embarking on AI projects: whether to opt…

Published in 2024. I have left it as written rather than quietly updating it, so treat anything specific as a snapshot of that date. The argument may still stand. The products have moved. In the rapidly evolving landscape of artificial intelligence (AI), open source models have emerged as a game-changer, democratizing access to cutting-edge technology.…

Published in 2024. I have left it as written rather than quietly updating it, so treat anything specific as a snapshot of that date. The argument may still stand. The products have moved. Introduction to Using Open Source Models In recent years, the use of open source models has been rapidly gaining traction in the…

Published in 2024. I have left it as written rather than quietly updating it, so treat anything specific as a snapshot of that date. The argument may still stand. The products have moved. The landscape of web-based search engines, primarily dominated by Google and Bing, has evolved significantly over the years and we move towards…