Was anyone else affected by the Claude outage last evening April 29? When it came back I lost access to my account. This evening I spent two hours convincing Fin (the support AI) that it was an Anthropic issue, not my computer or browser. Now Fin says they will investigate but I don't get a support ticket number and Fin won't confirm a time frame or even promise that I will get a response. I realize my $20 does not buy concierge service but losing access indefinitely is very discouraging. If I want to keep going I have to start again in a new account. I have Github repos for the project and the project context but I only committed half of yesterday's work, and I no longer have access to the session I was working in or all the prior ones. I won't just sit here, but what should I do? Should I show my repos to some other tool? Bite the bullet and drop a new $20? Those sessions and my past ones will never meet.

You learned the framework. Now try it on something you're actually working on. Vote below, then drop your prompt structure in the comments. Not the output. The structure. Show us how you set it up.

You built your first folder. Vote below, then drop a screenshot in the comments so we can see what you came up with.

https://x.com/felixrieseberg/status/2044128194647994585?s=46

I start a new role next week and need to absorb 5 technical books on agent-building — 50 chapters total. Reading them cover to cover would take weeks I don't have. Skimming them would give me summaries I can't use. So I built two things. 1. A research pipeline that reads the books for me. Shell script + JSON state files. Each chapter gets its own autonomous Claude session that reads the PDF pages, examines the companion repo code, extracts patterns, and maps them against the leaked source from Claude Code. One config file per book, one state file tracking progress. If a session fails, it moves on and I can re-run later. The whole orchestrator is ~100 lines of bash. 2. A Socratic teaching system that makes sure I actually learn what the pipeline extracts. Three markdown files — a teaching engine that asks me questions instead of explaining things, a setup tool that builds a learning plan from source material, and an adversarial interviewer that pressure-tests what I think I know. The pipeline gives me structured data from 50 chapters. The teaching system makes me work through it. Not read it — work through it. There's a difference. The goal isn't "summarize 5 books." The goal is to understand how these systems work well enough to build them, debug them, and explain the tradeoffs. The pipeline is just intake. The teacher is what closes the gap between having information and actually understanding it.