In the news
- I've been quiet for a while because I've been building some new things while heavily optimizing my workflow. What came out of it: the Eigenwise Toolshed, a free marketplace of Claude Code plugins, and Sidequest, the one at the center of it. Sidequest is a Kanban board where you never drag a card, because the cards get up and do the work themselves. The best part? It has model routing out of the box, so when you're mid-task and you notice something else is broken, Claude immediately saves it to the board and, depending on the
- The first thing I install in every Claude Code project isn't a workflow. It's a map. I've tried the Ralph loops, the spec-driven setups, all of it. And I keep genuinely trying the new stuff. But the thing I now install before writing a single line is a small plugin I built called codebase-mapper. Free, MIT-licensed. The problem it fixes is boring and constant: every new session, Claude starts blind. Same project, new chat, and it goes off to re-figure-out how Docker runs here, or re-reads half the codebase to answer a question it
- It's funny how everybody just blindly repeats marketing material all the time. After having been convinced to give GLM 5.2 a shot because everybody was saying it reaches opus level intelligence at a much cheaper price, I have to say, I am disappointed at the usage limits. Everyone is always complaining about Claude's usage limits, but I have to try very hard to hit the limit on my 20x plan. On GLM I seem to hit it twice as fast. Perhaps one of the reasons is that it actually requires more tokens to think to get to the same level of
- The US government just forced Anthropic to switch off its two best AI models for everyone who isn't American. Including Anthropic's own non-American employees. The stated reason: a "jailbreak." But every AI model ever built can be jailbroken. It's a mathematical property of how they work. You cannot patch it out, and a jailbreak-proof model does not exist anywhere on earth. So the precedent the Trump administration just set is this: it can pull any AI model, at any time, over a flaw that is ALWAYS present in every single one of
This page shows public professional information only, each fact cited. Is this you? send a correction, or ask for removal within 24 hours, no questions asked.