Skip to the document
Madhuopen lab

§5 · Writing

Series, not a feed.

Long-form, from the ground up: analogies first, numbers second, and the experiment that produced every number. Each series is written to be read in order — start at the first post, and each one assumes only the ones before it.

18essays
3series
38kwords
3 hof reading
What is Context Recall?latest · 27 June 2026

A series

RAG

Retrieval-Augmented Generation, built naive and improved one measured step at a time — Novice to Ninja.

6 parts · 1.2 h

Start here

How RAG actually works, start to finish

The map for the whole series: a naive RAG system improved one idea at a time, every step measured on the same hard test. Start here.

Read · 15 min →
  1. 02How do you actually measure a RAG system?12 min
  2. 03How does chunking change RAG?10 min
  3. 04How query rewriting helps (and hurts)11 min
  4. 05How does contextual retrieval work?12 min
  5. 06How does a semantic cache work?11 min

A series

Measurements

The four metrics that tell you whether a RAG or LLM system actually works — each one explained on its own.

5 parts · 0.6 h

Start here

What is Context Recall?

Of all the facts the right answer needs, how many did retrieval fetch? Recall is the ceiling everything else sits under.

Read · 7 min →
  1. 02What is Context Precision?Don't bring the junk.7 min
  2. 03What is Faithfulness?Don't make it up.7 min
  3. 04What is Answer Relevancy?Answer the actual question.7 min
  4. 05What is Retrieval Hit-Rate?Did you fetch any right doc at all?6 min

A series

Models in the browser

Gemma 4 running inside a browser tab on WebGPU: no server, no API key, no credit wall — and what that does to the frontend job.

2 parts · 0.7 h

Start here

There's an LLM Running in Your Browser Right Now — No Internet, No API, No Credits

Every app shows you that 0/3 credits screen. Gemma 4's TurboQuant + PLE make it possible to run an LLM entirely in the browser — no internet after first load, no token counter, no credit wall. Here's the full build.

Read · 23 min →
  1. 02UI Out of Thin Air — What Frontend Engineers Will Actually Do After Gemma 416 min

Build notes

Build notes

Short write-ups of things I built, most of them in a weekend.

5 notes

OptionMantra: VIX Data Pipeline - A Practical, Everyday Build

Building a cost-effective VIX data pipeline with Python, SQLite, and Google Cloud Storage. From $500/month to $2/month.

Thinking in Agents: The New Mental Model Every Developer Needs

We used to write code that told computers what to do. Now we're writing code that tells agents what to think about.

Introducing WordStash: Your Personal Vocabulary Companion

Building a dead-simple iOS app for capturing and learning new words. From copy-paste to AI-powered definitions in seconds.

Spec-Driven Development: Building ContentFlow CMS in 2 Days with AI

How I built an entire CMS—SDK, portal, server, deployment—in 32 hours by treating specifications as the primary artifact and letting AI handle 90% of the implementation.

Building ContentFlow CMS: A Journey into AI-Powered Development

How I built an entire content management system using Level 4 AI development—SDK, CMS, server, and deployment—in just 32 hours.