Practical AI, in writing.
Workflows, prompts, tools, and field notes from building production AI systems.Page 3 of 22
Claude Cowork Isn't Chat. Stop Using It Like Chat
Copy, paste, repeat is a chatbot habit. Cowork actually touches your files and finishes the work, if you stop treating it like a Q&A machine.
Stop Starting Over When Your AI Chat Breaks
When an AI conversation goes sideways, you don't have to abandon it. Edit the last good message and erase everything after.
Stop Prompting Before You've Planned Anything
Most AI projects fail before the first prompt gets typed. Here's the free Claude skill I built to fix that, and why five minutes of planning beats hours of rebuilding.
The Phone Call That Doesn't Need a Human Anymore
Local businesses lose up to 30% of bookings to voicemail and after-hours calls. Here's what it sounds like when an AI answers instead.
Your Prompts Keep Failing Because You're Guessing
Rambling output, wrong format, and hallucinations are three different problems. Stacking the same tricks on all of them is why your prompts don't work.
The Skill Type Anthropic Trusts Most (Most Teams Skip It)
Anthropic runs hundreds of skills internally across nine types, but most teams only use two. Here's the one that actually keeps Claude honest.
Claude Now Builds Its Own Team of Agents
Ask Claude for 50 things and it gives you 35, then quits. Dynamic workflows fix that by having Claude build its own fleet of agents.
Claude Design and Claude Code Finally Talk to Each Other
One command now syncs your real design system into your codebase, so Claude builds with actual components instead of guessing what your button looks like.
Claude Sonnet 5 Basically Closed the Opus Gap
Anthropic's new Sonnet lands near Opus on coding and reasoning at $2 per million tokens, and it now checks its own work without being asked.
Anthropic's AI Now Writes 65% of Its Own Team's Code
Anthropic put Claude inside Slack as a coworker you @ mention, and it's already writing 65% of their product team's code. Here's what that actually changes.
AI Doesn't Get Less Confident When It's Wrong
AI sounds exactly as sure about a quasar as it does about a taco recipe. One of those it actually knows. Here's how to tell which.
One Idea, Ten Posts, Fifteen Minutes
The blank page isn't your problem. Your process is. Here's how one good idea becomes a week of content without five hours of staring at a cursor.