Skip to content
Every Sunday from my repo to yours

What I shipped this weekand exactly how I built it.

The prompts, skills, and architecture I used that week to ship real systems with Claude Code, Xano, and MCP. Pulled straight from client repos. Nothing I haven't run myself.

Free. One click to unsubscribe. I never share your email.

2,400+ builders subscribed42 Sundays straight, no misses

Four things. Every issue. No filler.

No think pieces. No tool roundups. No AI news cycle. Just the work, pulled from that week's repos.

One playbook, end to end

The exact prompts, skills, and commits from a real client build that week. Copy them into your own repo Monday morning.

One tool I actually use

I test tools so you don't have to. One tool per issue — only the ones that earned a permanent spot in my stack.

One architecture, drawn out

A real Claude Code + Xano + MCP build, mapped end to end. See how the pieces connect before you spend a week guessing.

The actual code

Every issue ships with the repo, the prompt, or the diff. If I can't show you working code, I don't send the email.

“I've unsubscribed from every AI newsletter except this one. Robert ships the actual prompts and diffs. I copy them into my repo on Monday and they work.”
Scott R. — Technical founder, Snappy client

The last three Sundays.

Read these first. If they don't earn your inbox, the signup form is wasted on you.

Browse the full archive
Issue 042Mar 30
Agents

Build and control agents that actually ship code

The control loop I use to keep Claude Code on rails across a 40-file refactor. Skills, checkpoints, and the one rule that stops the agent drifting.

Read issue
Issue 041Mar 23
Xano MCP

Turn your Xano backend into an MCP server in an afternoon

Auth, rate limits, and tool descriptions an LLM can actually call. The same setup running in production for three Snappy clients right now.

Read issue
Issue 040Mar 16
Claude Code

The skill structure I copy into every client repo

Why most Claude Code skills rot after two weeks. The five-file pattern I use so skills get sharper the more the agent runs them.

Read issue

Read it if this sounds like you.

Sharp on purpose. If you're not in the left column, the emails won't land for you.

This is for you if
  • You've shipped real software and you want agents doing more of the work.
  • You're already in Claude Code or an IDE with an agent — and you want to go deeper.
  • You run a Xano backend, an MCP server, or a skill system — or you're about to.
  • You read code faster than you read hot takes.
Skip it if
  • You want a weekly AI news roundup. I don't cover the news cycle.
  • You want a curated list of 47 new tools. I cover one per issue, max.
  • You've never opened a terminal. The emails won't land yet.
  • You want somebody to hand you a finished product. I write for people building their own.

Get the next one.

2,400+ builders get this every Sunday morning.

Free. One click to unsubscribe.

Next one ships Sunday.

Subscribe and it hits your inbox the morning I send it.

Latest issueIssue 042

The control loop: keeping Claude Code on rails across 12 hours of agent work

The checkpoint pattern, the skill that catches drift before it lands a bad commit, and the one rule I never let the agent break. From a live client session this week.

Free. Unsubscribe in one click.

Book your session

A free session on your project.

We'll open your codebase together and start working on it.

What you can expect:

Architecture mapped, bottlenecks identified

Live building on your codebase, screen to screen

A clear plan for what to build next