Google AI Classroom Part 1: Solving the $25-Per-Exam Problem — Architecting a Sub-$0.02/Student Serverless AI Classroom with Gemini 4 & Firebase
Author: Cyrus Wong (黃俊彥) — Google Developer Expert - Cloud (AI), Senior Lecturer at HKIIT ...
Tutorials
Does the Red Banner Flicker or Change Colour in My Bama Bound AI Clip? Tracking Banner Hue Frame by Frame in Python
Checked November 9, 2026 (15:00 CST, UTC+8) against the live pages linked below. I build...
Your Agent Is Only as Good as Your Codebase
Coding agents copy what they find. In a tangled repo they copy the tangle. Here's why structure changes what an agent writes, and what RayStack's new dev-refactor skill does about ...
LLM-as-a-judge: building an AI grader you can actually trust
Short answer: An LLM judge works when you've calibrated it against human labels. Without calibration...
Why Your AI Agent Is Brilliant for 20 Minutes and Then Starts Lying
The first twenty minutes, my agent was perfect. It read the repo, found the failing service, drafted...
We Moved to a Bigger Model to Debug the Test. Then We Went Back to the Smaller One.
From the agentpolicy v0.1.0 release: a local-first, Cedar-based policy decision point for AI agents...
AI Can Reverse Engineer Your Code. CORE License Helps Keep Its Future Open.
AI Is Changing the Rules of Software Development For decades, software developers have...
Breaking My Own Agents #1: I Fed 5 n8n Workflows 30 Malformed Inputs. 102 of 150 Runs Failed Silently
My n8n lead qualifier sent a "🔥 Hot lead (85)" alert to Slack for a form whose message field was four...
Serving a stateful video world model on one RTX 4090: memory, flags and seeds in Hy-Nomas
Hy-Nomas keeps a video world between calls. On one RTX 4090 a turn took about 90 s, two decode flags ran out of memory, and Ada was required.
Claude Code Leaked 2,539 Secrets Onto GitHub. I Checked My Own Repos the Same Morning
I found out about it the way you find out about most things that ruin your morning: a link in a group...
What a passing test score didn't tell me about my AI agent
The eval suite said 17 of 18. Reading the transcripts showed a made-up email address and a model that never cached its prompt.
78 Claude Code runs on three tickets: the spec mattered more than the model
Current coding agents are good enough that "just describe the ticket" usually works. So is writing a...
Vibe engineering won't be enough
Vibe coding works. I built a photo collage tool that way myself. Vibe engineering goes a step further: the agent writes and the engineer steers. But I think even that won't be enou...
Why does a small local LLM repeat itself when you ask it for 100 rows of JSON?
Asking a language model for synthetic training data usually means one prompt: "Generate N rows as a...
Voice Robot for Outbound Calls: A Latency Budget from ASR to TTS
A voice robot that confirms appointments can be perfectly accurate and still feel broken. The...
Build Buddy, Day 2: two AI fixes compiled, only one was correct
Day 2 of MLH Global Hack Week: Builders Week. Yesterday my tool Build Buddy fixed 6 of 6 simple C#...
How we let an AI read production databases without trusting it
Every engineering team knows this ticket: "I was charged twice." Someone stops what they're doing,...
I built a bridge so AI agents can order groceries in Hinglish, and can't order without your yes
Haat translates between the Personal Agent Protocol (Poppy) and ONDC/Beckn. Notes on Indian-language parsing, an approval gate, and why 'unknown' is not 'failed'.
How to make a training dataset with AI when you have no data
Training a small classifier for your own inbox (refund requests, shipping questions, spam) starts...
I checked every quote Claude's deep research cited. One wasn't what the source said.
AI research tools are good at sounding sourced. Every paragraph ends with a link, every key point...
26 AI Breakdowns of the Same 10-Second Reel: One Hook, 25 Different Topics
A scheduled check has been sending one fixed Instagram Reel to the same vision model every day for sixteen days. hook.source and the on-screen lines came back identical 26 times ou...
The Compiler Trusts the Graph: An Adversarial Evaluation of a Toy ML Compiler
The Compiler Trusts the Graph: An Adversarial Evaluation of a Toy ML Compiler Abstract....
I replayed two weeks of Claude Code transcripts to see what my parallel sessions did to each other
For two weeks I ran Claude Code the way a lot of people now do: several sessions at once, usually one...
OpenSpec Tutorial (2026): Spec-Driven Development in Claude Code, Step by Step
Key Takeaways OpenSpec is a free command-line tool that makes your coding agent write a short plan...