Category: Development
-
How to Learn Mechanistic Interpretability Without a PhD: A Self-Taught Developer’s Path
A five-stage path from zero to contributing in mechanistic interpretability, for working developers with no research background: Neuronpedia in the browser, a transformer from scratch, the ARENA curriculum, reproducing one real result, then SPAR or MATS. The tools, the habits, and how much maths you actually need.
-
Circuit Tracing: How Researchers Follow an AI’s Thoughts Step by Step (The Marble Run Explanation)
Circuits and attribution graphs let researchers watch a model plan a rhyme, add 36 + 59 in an unexpected way, and sometimes invent a story about its own reasoning. How circuit tracing works, the three most quoted findings, what it cannot do yet, and how to try it at home, explained with a marble run.
-
Neurons, Features and Superposition: Why an AI’s Neurons Don’t Mean One Thing Each (The Toy Box Explanation)
Researchers expected each neuron in an AI model to mean one thing. Instead, most mean many things at once. Neurons, features, superposition and the sparse autoencoder, the tool that untangles them, explained with a jumbled toy box and 500 cups. Plus what you can see for yourself on Neuronpedia.
-
Why Should You Care About Mechanistic Interpretability? 7 Reasons for Founders, Developers and Curious Humans
You are already trusting a machine nobody can inspect. Seven plain-English reasons to learn the basics of mechanistic interpretability: debugging AI features, catching a bluffing model, enterprise and regulatory questions, steering, reading AI news, a field you can start this weekend, and the question of minds.
-
What Is Mechanistic Interpretability? Looking Inside an AI’s Head, Explained Like You’re Six
Nobody fully understands how the AI models we use every day work, not even the people who built them. Mechanistic interpretability is the science of opening them up and finding out. What it is, the Golden Gate Bridge experiment, what researchers have found so far, and why it matters, explained with examples a six-year-old could…
-
How Much Does It Cost to Fix a Vibe-Coded App? Diagnostic, Hardening and Rebuild Explained
The cost of fixing an AI-built app depends less on how broken it is than on whether the data model can carry the roadmap. The three tiers — diagnostic, hardening, core rebuild — what moves the number inside each, and how to get a quote you can trust.
-
Supabase RLS for Vibe-Coded Apps: The Policies Your AI Forgot to Write
If your AI-built app uses Supabase, every user can probably read every other user’s data — because Row Level Security is off or the policy says true. The 60-second check, the four policies every table needs, and the prompt that gets the AI to write them properly.
-
Adding AI Features to a Vibe-Coded App: Cost Control, Prompt Injection and the Three Ways It Goes Wrong
“Summarise this” is the magic feature — and the one that can empty your account overnight, leak another user’s data through a chat box, or get your provider account suspended. Twelve controls for the three failure modes AI features have that ordinary features don’t, plus the audit prompt to run on your own code.
-
Vibe-Coded Mobile App Rejected by the App Store or Google Play? The 10 Usual Reasons and How to Clear Them
AI-built iOS and Android apps hit store review with a predictable set of problems: crashes on the reviewer’s device, no demo login, missing Sign in with Apple, Stripe inside the app, junk permissions, placeholder screens, no account deletion. The ten reasons, how to clear each, and the one-hour pass before you submit once.
-
Is Vibe Coding Safe for a Startup? The 12 Questions Investors and Enterprise Customers Will Ask
Nobody serious penalises you for building with AI. They penalise not knowing what you have. The twelve technical-diligence and security-questionnaire questions a vibe-coded startup will face — code ownership, tests, tenant isolation, secrets, backups, monitoring, data model, AI review — with what a good answer sounds like and how to prepare in a week.