changelog
the curriculum is alive: every lesson carries a version, every content change is a dated entry. this is the full log, newest first.
76 entries in the curriculum log
2026-07-15
- what is a coding agent?takeaway block added; read time shown separatelyv1.1
- system requirementstakeaway block added; read time shown separatelyv1.1
- accounts and coststakeaway block added; read time shown separatelyv1.1
- installing claude codetakeaway block added; read time shown separatelyv1.1
- installing codextakeaway block added; read time shown separatelyv1.1
- the first mini-buildtakeaway block added; read time shown separatelyv1.1
- the good PRD: what should the agent build?takeaway block added; read time shown separatelyv1.2
- the good PRD: what should the agent build?claims validated against external sources and linked (Anthropic memory docs, agents.md standard, Osmani, O'Reilly)v1.2
- the good PRD: what should the agent build?free preview lesson of module 1 (per the module-preview rule)v1.2
- CLAUDE.md: the project's constitutiontakeaway block added; read time shown separatelyv1.2
- CLAUDE.md: the project's constitutionclaims validated against external sources and linked (Anthropic memory docs, agents.md standard, Osmani, O'Reilly)v1.2
- context: what the agent actually seestakeaway block added; read time shown separatelyv1.2
- context: what the agent actually seesclaims validated against external sources and linked (Anthropic session-management post, usage docs, best practices, Osmani)v1.2
- context: what the agent actually seesfirst publishedv1.2
- the anatomy of a good requesttakeaway block added; read time shown separatelyv1.2
- the anatomy of a good requestclaims validated against external sources and linked (Anthropic session-management post, usage docs, best practices, Osmani)v1.2
- the anatomy of a good requestfirst publishedv1.2
- the first production sessiontakeaway block added; read time shown separatelyv1.1
- the first production sessionmodule gained two new lessons (context, good request) — renumberedv1.1
- the first production sessionfirst published — sources: Anthropic best-practices guide + internal-teams case studyv1.1
- verification: what "done" meanstakeaway block added; read time shown separatelyv1.1
- verification: what "done" meansmodule gained two new lessons (context, good request) — renumberedv1.1
- verification: what "done" meansfirst published — sources: Anthropic verification guide, Addy Osmani spec framework, O'Reilly SDD analysisv1.1
- the fix loop and a clean headtakeaway block added; read time shown separatelyv1.1
- the fix loop and a clean headmodule gained two new lessons (context, good request) — renumberedv1.1
- the fix loop and a clean headfirst published — sources: Anthropic session-management post, Claude Code usage docs, best-practices guidev1.1
- skills: packaged expertisetakeaway block added; read time shown separatelyv1.1
- skills: packaged expertisefirst published, claims validated against the official skills docs and the Agent Skills open standardv1.1
- hooks: rules that always runtakeaway block added; read time shown separatelyv1.1
- hooks: rules that always runfirst published, claims validated against the official hooks guide and referencev1.1
- subagents: parallel handstakeaway block added; read time shown separatelyv1.1
- subagents: parallel handsfirst published, claims validated against the official subagents and parallel-agents docsv1.1
- MCP: the agent's door to the worldtakeaway block added; read time shown separatelyv1.1
- MCP: the agent's door to the worldfirst published, claims validated against modelcontextprotocol.io and the official MCP quickstartv1.1
- codex next to claude codetakeaway block added; read time shown separatelyv1.1
- codex next to claude codefirst published, claims validated against the official Codex docs (AGENTS.md discovery, customization layers)v1.1
- picking the model: intelligence for the pricetakeaway block added; read time shown separatelyv1.1
- picking the model: intelligence for the pricefirst published, prices and aliases validated against the official Anthropic model and pricing docsv1.1
- deploy: from repo to live URLtooling note added: connect the Vercel MCP so the agent drives deploys from live docs and real deployment statev1.1
- deploy: from repo to live URLfirst published, claims validated against the official Vercel git and environments docsv1.1
- env vars and secretstooling note added: the agent manages variable names via CLI and MCP; values stay in the platformv1.1
- env vars and secretsfirst published, claims validated against the official Vercel environment-variables docsv1.1
- auth: sign-in without passwordstooling note added: a docs MCP keeps the agent on current Better Auth documentationv1.1
- auth: sign-in without passwordsfirst published, claims validated against the Better Auth magic-link, security, and options docsv1.1
- payments: checkout and the webhooktooling note added: Stripe MCP server and agent skills for prompt-driven payment setupv1.1
- payments: checkout and the webhookfirst published, claims validated against the Stripe webhooks, quickstart, and subscription-webhooks docsv1.1
- the database in productiontooling note added: the Neon MCP server covers the branch-migrate-verify loop as agent toolsv1.1
- the database in productionfirst published, claims validated against the Neon branching and branch-management docsv1.1
- monitoring and the fix loop in productiontooling note added: the Vercel MCP turns log reading and deployment state into agent toolsv1.1
- monitoring and the fix loop in productionfirst published, claims validated against the Vercel runtime-logs, observability, and rollback docsv1.1
- plan mode: the plan before the typingfirst published, claims validated against the official permission-modes and common-workflows docsv1.0
- worktrees and parallel sessionsfirst published, claims validated against the official worktrees and common-workflows docsv1.0
- review: an agent checks the agentfirst published, claims validated against the official code-review and ultrareview docsv1.0
- permissions: how much ropefirst published, claims validated against the official permissions, permission-modes, and sandboxing docsv1.0
- scheduled work: the agent without youfirst published, claims validated against the official scheduled-tasks, routines, and GitHub Actions docsv1.0
- the living project: memory as leveragefirst published, claims validated against the official memory docsv1.0
- the anatomy of AI slopcopyable prompt card added to the exercisev1.1
- the anatomy of AI slopfirst published, claims sourced from Simon Willison's slop definition and GitClear's 2025/2026 code-quality researchv1.1
- tests: the check the agent can runcopyable prompt card added to the exercisev1.1
- tests: the check the agent can runfirst published, claims validated against the official best-practices doc and Anthropic's team-usage reportv1.1
- design law: a system the agent can't drift fromcopyable prompt card added to the exercisev1.1
- design law: a system the agent can't drift fromfirst published, token claims sourced from the DTCG stable specification and Tailwind v4 theme docsv1.1
- anti-slop copy: prose that survives a suspicious readercopyable prompt card added to the exercisev1.1
- anti-slop copy: prose that survives a suspicious readerfirst published, tell patterns sourced from Wikipedia's Signs of AI writing field guidev1.1
- accessibility and speed: the gates most sites failcopyable prompt card added to the exercisev1.1
- accessibility and speed: the gates most sites failfirst published, claims sourced from the WebAIM Million 2026 report, web.dev vitals thresholds, and Lighthouse CI docsv1.1
- your own quality gatecopyable prompt card added to the exercisev1.1
- your own quality gatefirst published, gate mechanics validated against the official hooks and best-practices docsv1.1
2026-07-14
- what is a coding agent?first publishedv1.1
- system requirementsfirst publishedv1.1
- accounts and costsfirst publishedv1.1
- installing claude codefirst publishedv1.1
- installing codexfirst publishedv1.1
- the first mini-buildfirst publishedv1.1
- the good PRD: what should the agent build?first publishedv1.2
- CLAUDE.md: the project's constitutionfirst publishedv1.2