All 20 dispatches from the lab.
Investigations, essays, and release notes. Sorted newest first. Every number on this page is real, every claim is footnoted in the piece it belongs to.
-
№ 020
AI & Development
17 min read →
The 5-Minute Baseline: What We Found in Claude Code’s Tools Array
Update (2026-04-13): Our original recommendation to pin Claude Code to v2.1.81 is now outdated. Anthropic has partially fixed the resume-scatter bug in v2.1.90 and further improved cache stability in v2.1.104.
-
№ 019
AI & Development
8 min read →
Building a Fetch Interceptor to Fix Claude Code’s Cache — Without Touching Claude Code
In Part 2, we mapped out three bugs that break Claude Code's prompt caching: resume block scatter, fingerprint instability, and non-deterministic tool ordering. Each one independently busts the cache. Together,
-
№ 018
AI & Development
7 min read →
How Claude Code’s Prompt Cache Actually Works — And Three Ways It Breaks
In Part 1, we showed that Claude Code sessions were burning through quota at 26-28% per turn — because the prompt cache was being rebuilt from scratch on every API
-
№ 017
AI & Development
8 min read →
We Taught Our AI Agents to Take Coffee Breaks (And It Saved Us Real Money)
We Taught Our AI Agents to Take Coffee Breaks (And It Saved Us Real Money) Here's something nobody tells you about Claude Code: when you step away from your keyboard,
-
№ 016
AI & Development
8 min read →
We Burned Through 100% of Our Claude Code Quota in Two Hours. Here’s What We Found.
Claude Code is, in our experience, the best AI coding tool available today. Anthropic's models are genuinely excellent — the reasoning is sharp, the context window is enormous, and when
-
№ 015
AI & Development
10 min read →
When AI Agents Argue: What One Adversarial Reviewer Caught That Four Collaborative AIs Missed
The Premise What happens when you deploy seven specialized AI agents on a long-term research project — and one of them is specifically designed to tear the others’ work apart?
-
№ 014
General
6 min read →
Adding Test Coverage to 20-Year-Old Koha Code — With a Little Help from AI
If you’ve worked on a large open-source project, you’ve seen the pattern: critical modules that have run in production for years with minimal test coverage. Not because nobody cares, but
-
№ 013
General
5 min read →
When a 20-Year-Old Perl Bug Comes Knocking: How We Used AI to Track It Down
At Veritas Supera, we maintain PDF::Reuse — a Perl module on CPAN that lets developers efficiently create PDF documents by reusing existing components. It’s roughly 7,000 lines of code spanning