Vibecoding 023 — AFL Masterclass: Coverage-Guided Fuzzing From First Principles
A self-paced course, with points and progress, on AFL: the tool that finds bugs by throwing millions of slightly-wrong inputs at a program and watching what breaks.
100 days. 100 projects. One LLM prompt at a time.
Each post here is a project built with AI assistance — browser extensions, bots, tools, dashboards, analyzers. Some took an afternoon. Some fought back. All of them taught me something about what these tools are actually good at, and where they fall short.
The rule: build something real, ship it, write about it honestly.
A self-paced course, with points and progress, on AFL: the tool that finds bugs by throwing millions of slightly-wrong inputs at a program and watching what breaks.
A private reader that strips out trackers and hands you just the article text, before any of the tracking junk reaches your phone.
One search box, two social networks, and a live stream of what the security community is talking about right now.
A hands-on lab where you try ten of the most common website attacks yourself, with an AI tutor that nudges you along without giving away the answer.
Blackout poetry from the Hacker News front page: an AI picks the words that survive, and the rest get inked out.
Notepad++ has a hidden easter egg: type the word 'random', select it, press F1, and a random quote appears. I pulled out all 247 of them.
A browser tool that looks up a piece of malware in several public databases, then has AI write a rule to detect it and explain that rule line by line.
A small web app that stress-tests any claim or article: the strongest case for it, the weakest version, and a balanced verdict.
A timed speaking drill for security leaders: 26 tricky situations across three modes, with AI feedback on jargon, clarity and talking like a business person.
A bot that posts P.G. Wodehouse quotes to Bluesky twice a day, with a different kind of quote each day of the week, and costs nothing to run.