Seven sites · one network
Learn to code when the code writes itself.
The syntax tutorial is a solved problem — your assistant already knows it. What nobody teaches is the new job: configuring an agent for a language, giving it a feedback loop it cannot fake, and reviewing what it hands back. That is what these sites are for now.
The network
Every site is free, ad-light, and has three tracks: foundations, agent workflow, and review.
learn-python.com
Learn Python
Python for people whose code is half-written by an agent.
learn-javascript.org
Learn JavaScript
JavaScript fundamentals, and the agent workflows built on top of them.
learn-typescript.org
Learn TypeScript
The type system is the best prompt you will ever write.
learn-go.org
Learn Go
A small language an agent can hold entirely in its head.
learn-c.net
Learn C
The language where you cannot afford to skip the review.
learn-bash.net
Learn Bash
The language your agent runs the most and understands the least.
learn-nestjs.com
Learn NestJS
A framework with enough structure that an agent can't get lost.
Across the network
The questions that are not about one language.
Cross-network · comparison
Which language should you hand to an agent?
Seven languages, ranked by how much of a generated mistake survives to production. The answer is mostly about what the toolchain refuses to accept.
Cross-network · configuration
Customizing your coding agent
Instructions, slash commands, subagents, skills, output styles, plugins. Six ways to shape a harness — and a rule for which to reach for, so you build capability instead of clutter.
Cross-network · deep dive
Agent hooks: preprocess prompts, allow and deny, audit
Instructions are advice. Hooks are enforcement. A working reference for the extension points in Claude Code and Codex, and what to build with them.
Cross-network · guardrails
Sandboxing a coding agent
Hooks are policy: they stop what you anticipated. A sandbox is containment: it bounds what you did not. You want both, and most people have neither.
Cross-network · guardrails
Prompt injection is the guardrail problem
Every other guardrail assumes the threat is the model doing something odd. This one is different: the instructions come from the data, and the model is doing exactly what it was told.
Cross-network · operations
Reading the record of what your agent did
Every session you run is already recorded in full. Almost nobody reads it — and the record is the fastest way to find out why a run went sideways.
Cross-network · economics
What tokens actually cost, and where the money goes
The pricing model has five moving parts, and most bills are dominated by the two nobody thinks about. Here is the mental model, before the language-specific implementation.
Why we rebuilt all of this
These sites used to be what every tutorial site was: a list of syntax pages. In 2026 that page loses to a two-second question in an editor, and it should.
So the foundations stayed — they are short, runnable and honest about what you can skim — and we built two new tracks on top of them. One covers the tooling: AGENTS.md files, agent permissions, MCP servers, the test and type loops that actually constrain a model. The other covers review: the specific, repeatable ways generated code fails in each language, and the checks that catch it.
Every page is also published as plain Markdown at <url>index.md, and each site ships an llms.txt. If an assistant is reading this on your behalf, it can read the source directly.