What We've Learned
Lessons from hundreds of n8n workflows. No theory. Just what actually works when you need AI automation to ship.
Featured Posts
AGENTS.md vs CLAUDE.md: which file your agent actually reads
Both files in one repo, measured: load order, token cost, and which one wins when they disagree. Every guide explains the difference; this one runs the test.
Compound engineering vs Superpowers: what each one installs
Both Claude Code plugins installed and measured: 14 skills against 33, one SessionStart hook against none, and the six skills that collide when you run both.
Claude Code Video Editing: How I Stopped Touching the Timeline
Claude Code can drive a real video timeline over MCP – reading project state, placing clips, styling captions – on a project file you can open and inspect. Here's the setup that turns an agent into an editing team, the four-layer specification that makes it work, and the decisions that stay human.
AI readiness checklist: five checks before you automate a workflow
Readiness gets assessed at two levels. Organization readiness plans the program; workflow readiness picks the next build. This is the second one: five checks, scored out of 25, in about five minutes, with your weakest check named so you know what to strengthen first.
Claude for Small Business (2026): all 15 workflows, the real cost, and when to skip it
Claude for Small Business is a plugin for Claude Cowork with 15 ready-to-run workflows and 15 skills. Here's what's in it, how it compares to ChatGPT Team, and a decision tree for picking the right setup.
The Gauntlet Loop, tested: four runs of Matt Shumer's method
Matt Shumer's Gauntlet Loop, run four times on real projects: three browser games and an SEO overhaul. What the viral prompt leaves out, and what actually decides success.
How to Automate Invoice Processing: An Operator's Guide
The five steps every invoice automation runs, the extraction step where projects actually fail, real 2026 pricing from $0 to $89 per user, and the invoices you should not automate at all.
Your CLAUDE.md File Is Too Long: The Context Diet That Cut Mine 87%
Anthropic removed ~80% of Claude Code's system prompt for Claude 5, and the internet decided that means delete your CLAUDE.md. Don't. Here's the four-bucket diet that took mine from 1,204 lines to 151 without losing a single incident-earned warning.
AI for property management: what it handles, what it breaks, and what it costs
Property management is mostly relay work, and that is exactly the shape automation is good at. What it handles reliably, where it creates legal exposure, and what it costs.
AI receptionist for small business: what it handles, what it ruins, and what it costs
An AI receptionist answers the calls you are currently missing. The decision is not whether it works, it is which calls it is allowed to take, and what a missed call is worth to you.
AI agents for small business (2026): what they actually do, what they cost, and when not to buy one
AI agents complete multi-step jobs across your tools without someone driving them. Here is what the categories are, what they cost, and the three questions that tell you whether to buy one or leave the process alone.
Claude Opus 5: Fable 5 intelligence at half the price, with a dial for how hard it thinks
Anthropic shipped Claude Opus 5 on July 24 at exactly Opus 4.8's price, landing within 0.5% of Fable 5 on coding for half the cost. A builder's read on the effort dial, the two breaking API changes, and the prompting habit you now need to delete.
Ready to put AI to work?
Book a free discovery call and we'll tell you what we'd automate first – and what we wouldn't touch yet.
30 minutes. No pitch. You keep the plan.
Prefer to build it yourself? Try Flow free