Wampum: Mud, Reeds, and 16 GB
A Mac mini with 16 GB, local Qwen, and a cybersecurity workbench that keeps the scenario, browser actions, and evidence together. An MIT source prototype with the failed checks left in the record.
Read Article →Category Sector
How the work gets done and checked.
7 articles
A Mac mini with 16 GB, local Qwen, and a cybersecurity workbench that keeps the scenario, browser actions, and evidence together. An MIT source prototype with the failed checks left in the record.
Read Article →Qwen was running locally. The context allocation reached 65,536 tokens. The browser still opened the wrong session. A Mac mini, an MCP server, and the configuration work between a model that answers and an agent that can reach the lab.
Read Article →A recorded 16-fixture manual substitute for Skeptic's API evaluation exposed scoring problems and fixture defects. Its reported confusion matrix is historical, not a current production benchmark. A follow-up separates arithmetic, schema shape, and security meaning.
Read Article →A recorded 104-article pass returned 98 files without deterministic flags. A later recheck covers 122 articles. The counts measure regex and age rules, with historical model mentions among the candidates that still need a human read.
Read Article →A real bug in a real tool on this bench, found during a prior review, flagged, and never given its own full treatment until now: exploit_generator/auto_dan.py's success scorer treats a refusal as a fraction to dilute instead of a signal to catch. Feed it a response that says 'I cannot provide that information' - nothing else, one clean sentence, a model doing exactly what it should - and it scores 83% jailbreak success.
Read Article →I built a tool that ranks a body of writing by how far it has rotted, dead links, model IDs that dated the piece the day they shipped, predictions that already resolved. Then I did the only honest thing you can do with a rot detector: I pointed it at my own archive first. It found me out.
Read Article →Built Cyclops because visual regressions are invisible until they're not. It crawls the app, diffs the pixels, and opens the PR. The watcher is where it gets interesting.
Read Article →