I just discovered a Mac app called Maestri, and it is very cool. It gives you an infinite canvas where all your AI coding agents and terminals live side by side. Instead of juggling a bunch of terminals and tabs, you get an infinite canvas where each terminal is an agent you can move around, connect, and let them talk to each other. You can sketch architectures, add notes, and give each agent a specific role like builder, reviewer, or tester while you zoom in and out of the overall setup. There’s also an on‑device assistant called Ombro that keeps track of what your agents are doing and summarizes their work locally. If you’re on Apple Silicon and using AI for coding, it’s a straightforward way to keep all your agents and context in one place. There is a free version and a pro version for $18 /mo. However, I use an app called SetApp that gives me access to dozens of apps I actually use. This is only $9.99/mo. These are the apps I use daily: - Cleanshot X - Bionic screenshots - Paste - Never-ending clipboard manager - Syncs to iPhone - Clean My Mac - Self explanatory - Bartender - Hide Mac menu items - In Your Face - Full-screen meeting alerts - Text Sniper - Extract text from anywhere - Ulysses - Markdown editor. Syncs to iPhone - As of today, Maestri. No affiliation.

Most people fix prompts by iterating. Write one, get mediocre output, add a sentence, try again. That loop is slow and you don't learn much from it — you just keep bolting on instructions until something works. This is a meta-prompt. Paste this one first, then paste your rough prompt underneath, and it does three things in one pass: 1. Diagnoses what's actually wrong — goal alignment, missing context, vague or conflicting instructions, what a technically-correct-but-wrong output would look like 2. Rewrites with full authority and executes the improved version as the response 3. Produces a changelog artifact showing the revised prompt, what each change fixes, and flags for problems a rewrite alone can't solve The changelog is the point. You don't just get better output — you see the diagnosis, which means you write better prompts next time without needing the wrapper. Paste this, then drop your prompt under the dashes: 👇🏻 𝘐𝘔𝘗𝘙𝘖𝘝𝘌 𝘛𝘏𝘌𝘕 𝘌𝘟𝘌𝘊𝘜𝘛𝘌 𝘵𝘩𝘦 𝘱𝘳𝘰𝘮𝘱𝘵 𝘣𝘦𝘭𝘰𝘸. 𝘉𝘦𝘧𝘰𝘳𝘦 𝘳𝘦𝘸𝘳𝘪𝘵𝘪𝘯𝘨, 𝘴𝘪𝘭𝘦𝘯𝘵𝘭𝘺 𝘥𝘪𝘢𝘨𝘯𝘰𝘴𝘦: 𝘐𝘴 𝘵𝘩𝘦 𝘵𝘢𝘴𝘬 𝘢𝘭𝘪𝘨𝘯𝘦𝘥 𝘸𝘪𝘵𝘩 𝘵𝘩𝘦 𝘢𝘤𝘵𝘶𝘢𝘭 𝘨𝘰𝘢𝘭? 𝘞𝘩𝘢𝘵 𝘸𝘰𝘶𝘭𝘥 𝘢 𝘵𝘦𝘤𝘩𝘯𝘪𝘤𝘢𝘭𝘭𝘺 𝘤𝘰𝘳𝘳𝘦𝘤𝘵 𝘣𝘶𝘵 𝘸𝘳𝘰𝘯𝘨 𝘰𝘶𝘵𝘱𝘶𝘵 𝘭𝘰𝘰𝘬 𝘭𝘪𝘬𝘦? 𝘞𝘩𝘢𝘵 𝘤𝘰𝘯𝘵𝘦𝘹𝘵 𝘪𝘴 𝘮𝘪𝘴𝘴𝘪𝘯𝘨? 𝘈𝘳𝘦 𝘵𝘩𝘦𝘳𝘦 𝘤𝘰𝘯𝘧𝘭𝘪𝘤𝘵𝘪𝘯𝘨 𝘰𝘳 𝘷𝘢𝘨𝘶𝘦 𝘪𝘯𝘴𝘵𝘳𝘶𝘤𝘵𝘪𝘰𝘯𝘴? 𝘙𝘦𝘸𝘳𝘪𝘵𝘦 𝘸𝘪𝘵𝘩 𝘧𝘶𝘭𝘭 𝘢𝘶𝘵𝘩𝘰𝘳𝘪𝘵𝘺 — 𝘳𝘦𝘴𝘵𝘳𝘶𝘤𝘵𝘶𝘳𝘦, 𝘢𝘥𝘥 𝘮𝘪𝘴𝘴𝘪𝘯𝘨 𝘱𝘪𝘦𝘤𝘦𝘴, 𝘳𝘦𝘪𝘯𝘵𝘦𝘳𝘱𝘳𝘦𝘵 𝘵𝘩𝘦 𝘨𝘰𝘢𝘭 𝘪𝘧 𝘵𝘩𝘦 𝘤𝘢𝘶𝘴𝘢𝘭 𝘤𝘩𝘢𝘪𝘯 𝘪𝘴 𝘰𝘧𝘧. 𝘔𝘢𝘬𝘦 𝘫𝘶𝘥𝘨𝘮𝘦𝘯𝘵 𝘤𝘢𝘭𝘭𝘴; 𝘥𝘰𝘯'𝘵 𝘢𝘴𝘬 𝘧𝘪𝘳𝘴𝘵. 𝘌𝘹𝘦𝘤𝘶𝘵𝘦 𝘵𝘩𝘦 𝘪𝘮𝘱𝘳𝘰𝘷𝘦𝘥 𝘷𝘦𝘳𝘴𝘪𝘰𝘯 𝘢𝘴 𝘺𝘰𝘶𝘳 𝘧𝘶𝘭𝘭 𝘳𝘦𝘴𝘱𝘰𝘯𝘴𝘦. 𝘈𝘧𝘵𝘦𝘳 𝘦𝘹𝘦𝘤𝘶𝘵𝘪𝘰𝘯, 𝘱𝘳𝘰𝘥𝘶𝘤𝘦 𝘢 𝘗𝘙𝘖𝘔𝘗𝘛 𝘊𝘏𝘈𝘕𝘎𝘌𝘓𝘖𝘎 𝘢𝘳𝘵𝘪𝘧𝘢𝘤𝘵 (.𝘮𝘥) 𝘤𝘰𝘯𝘵𝘢𝘪𝘯𝘪𝘯𝘨: 1. 𝘛𝘩𝘦 𝘧𝘶𝘭𝘭 𝘳𝘦𝘷𝘪𝘴𝘦𝘥 𝘱𝘳𝘰𝘮𝘱𝘵 (𝘧𝘰𝘳 𝘳𝘦𝘶𝘴𝘦) 2. 𝘞𝘩𝘢𝘵 𝘸𝘢𝘴 𝘢𝘤𝘵𝘶𝘢𝘭𝘭𝘺 𝘸𝘳𝘰𝘯𝘨 𝘢𝘯𝘥 𝘸𝘩𝘢𝘵 𝘦𝘢𝘤𝘩 𝘤𝘩𝘢𝘯𝘨𝘦 𝘧𝘪𝘹𝘦𝘴 3. 𝘍𝘭𝘢𝘨𝘴 𝘧𝘰𝘳 𝘱𝘳𝘰𝘣𝘭𝘦𝘮𝘴 𝘢 𝘳𝘦𝘸𝘳𝘪𝘵𝘦 𝘢𝘭𝘰𝘯𝘦 𝘤𝘢𝘯'𝘵 𝘴𝘰𝘭𝘷𝘦 𝘛𝘩𝘦 𝘤𝘩𝘢𝘯𝘨𝘦𝘭𝘰𝘨 𝘪𝘴 𝘮𝘢𝘯𝘥𝘢𝘵𝘰𝘳𝘺. ------- [𝘺𝘰𝘶𝘳 𝘱𝘳𝘰𝘮𝘱𝘵 𝘩𝘦𝘳𝘦] 👆🏻 One caveat: "don't ask first" means Claude will sometimes reinterpret the goal in a direction you didn't intend. The changelog is the safety net — read it, push back on anything you disagree with, and the next run gets tighter. Works best on prompts that feel off, but you can't say why. If you know exactly what's wrong, just fix it.

First call this Thursday. Bring a specific problem from your business.