Skip to content
fellowcoder

New writing every week

Code, AI, and the craft of building software.

Long-form articles and hands-on tutorials on AI engineering, systems design, and the unglamorous decisions that keep codebases alive. No hype, no listicles — just the working notes of people shipping real things.

Start readingBrowse tutorials
Articles
06
Tutorials
05
Topics
19
~/fellowcoder — zsh

A publication for engineers who build with AI, and want to understand what they are building. Topics: AI engineering, systems, and craft. Principles: ship small, measure honestly, read the source.

ai-engineeringpostgressystemspythonragtypescriptagentscraftevalsllmstreamingcachingcontextdatabasesperformancesearchsqltestingtypes

Latest

Fresh off the branch

All articles
Article6 min

Context windows are not memory

A million tokens of context does not give a model memory. It gives it a very large short-term working set — and confusing the two is why your agent forgets what it decided twenty minutes ago.

Read the piece
ai-engineeringllmcontextagents
Tutorial9 minadvanced

Measuring retrieval: recall@k, MRR, and a golden set you'll maintain

Build the harness that tells you whether a change helped. Bootstrap a golden set from your own corpus, separate retrieval failures from generation failures, and stop shipping on vibes.

Start tutorial
Tutorial8 minintermediate

Hybrid retrieval: fuse keyword and vector search

Vector search alone misses exact identifiers and rare terms. Add Postgres full-text search, fuse the rankings with RRF, and rerank the top slice.

Start tutorial
Tutorial9 minintermediate

Chunking that survives contact with real documents

Build the ingestion half of a retrieval pipeline: structure-aware splitting, contextual headers, and a schema that lets you re-chunk without re-embedding the world.

Start tutorial

Tracks

Tutorials that build on each other

All tutorials

Retrieval that survives production

3 parts
  1. 01Chunking that survives contact with real documents9 min · intermediate
  2. 02Hybrid retrieval: fuse keyword and vector search8 min · intermediate
  3. 03Measuring retrieval: recall@k, MRR, and a golden set you'll maintain9 min · advanced

Archive

More from the blog

Browse everything

Write the eval first

Prompt engineering without an eval is just vibes with extra steps. Here's how to build a scoring harness in an afternoon, and why twenty examples beat two thousand.

Article6 min

The cache you think you have

Your hit rate is 4% and you can't see why. A field guide to the invalidations nobody writes down — in HTTP caches, in Postgres, and in LLM prompt prefixes.

Article6 min

Types that earn their keep

Most TypeScript adds ceremony without removing bugs. Four patterns that make illegal states unrepresentable instead of merely annotated.

Article6 min

Streaming is a state machine, not a string

The token loop is the easy part. Reconnects, partial JSON, backpressure, and the abandoned-request problem are where streaming implementations actually break.

Article7 min

Postgres is probably enough

Before you add Redis, Elasticsearch, a vector database, and a queue: Postgres does all four adequately, and 'adequately' is the correct target until it isn't.

Article6 min

Free, forever, no signup

Every article ships with runnable code and the reasoning behind it.

Subscribe via RSSExplore topics