Life & WorkFor Jim Collins.
A clear head on one thing that matters, with the AI you already use. Yours to keep, change or share.
On a phone: open the app, then use “Add to Home Screen” in the browser menu, and it sits on your phone like an app.
A real example
A manager typed one line in the app: “Promote Alex or hire from outside?”, expecting Alex to do well, 70% sure, check-in in a month. A tap on Claude copied this text and opened Claude; pasted there and sent, it got the answer below.
My own forecast: Alex will do well · How sure I am: 70% · Check-in: November 27
Help me with this, one step at a time… Work with these ideas from Jim Collins, and name the idea when you use it: Confront the brutal facts, the Stockdale Paradox, First Who, then what (the right people on the bus, the wrong people off it and the right people in the right seats, before settling where to drive), Fire bullets, then cannonballs; where the question fits them, also the Hedgehog Concept, the 20 Mile March, the SMaC recipe, Leading above the Death Line, the Flywheel, the Genius of the AND, Preserve the core, stimulate progress and Return on Luck, each with its core in a line. Keep facts apart from guesses. End with one small next step and a check day.
Your starting point: You think Alex will do well, and you're 70% sure. That's a real lean, but it leaves a 3-in-10 chance you're wrong. We'll keep that number in view throughout.
The question, as I'd frame it:
Is Alex the right person for this specific seat, and can I know that well enough, by when I need to, to say yes over the best outside candidate I could realistically get?
I've framed it this way because of First Who, then what. The question isn't "Alex or an outsider?" in the abstract. It's whether Alex fits this seat, the one the role will need over the next year or two. Collins also says: when in doubt about a person, keep looking. So a 70% case asks whether that doubt is the kind that should stop you, or just ordinary uncertainty about any hire.
A few things are hidden inside your question. I'd like to make them visible before we go on:
• What is the seat? What does this role have to deliver that it doesn't today?
• What happens to Alex if you hire from outside? Do they stay, leave, or stay but disengage?
• Is it truly either/or? There may be middle paths, such as an acting or trial period. (That's Fire bullets, then cannonballs, and we'll come back to it.)
The one hard question for this step:
In one or two sentences, what must the person in this role do well in the first year that Alex has not yet shown they can do?
If the answer is "nothing, they've done it all," your 70% may be too low. If there's a clear gap, that gap is probably where the remaining 30% lives.
Your answer and any corrections to how I've framed the question will take us to Step 2: the facts and the biggest risk.
On November 27 the app asks: what really happened? Over time you see how often your forecasts held up.
Put it inside your AI
Claude Desktop (no terminal)Four simple steps, no commands
On your computer (Mac or Windows):
- Download the plugin (life-work-collins-plugin.zip, 1.1 MB). If your Mac unzipped it automatically, no problem: the original .zip is still in Downloads.
- In Claude Desktop open Customize (left sidebar), then Plugins.
- Choose Add, then Upload plugin, and pick the .zip file you downloaded.
- Start a new chat and say: “Help me think through a decision.”
SHA-256 fingerprint of the zip:
2bd67db661a1d42c05fda135af19059dd58a0869d66017b18276a172c75db7f2
Add it to Claude Code, Codex or another agentFor technical users · three steps, one ready-made message
- Download the kit (life-work-collins.zip, about 1.8 MB) and keep it in Downloads.
- Open Claude Code, Codex or your own agent and paste this message:
Please install the Life & Work kit for me. I saved the kit on this computer as a file named life-work-collins.zip (look in Downloads, then on the Desktop). Before opening it, check that its SHA-256 is 797bf60c883f91b639d64343520afa38a442c832a047d6fe3425c240a726bdc8 (Mac/Linux: shasum -a 256 <file>; Windows: Get-FileHash <file>); if it differs, stop and tell me. Unzip it into a new empty temporary folder and run the installer that matches this computer (you can tell which one it is; I do not have to choose): on a Mac or Linux, sh <folder>/life-work/install.sh ; on Windows, powershell -ExecutionPolicy Bypass -File <folder>\life-work\install.ps1 . If it asks for anything, finish those steps. Then tell me in two lines what was installed and start the first conversation (the map skill) with me right away.
On Windows, “-ExecutionPolicy Bypass” lets PowerShell run this one installer script, this one time; it changes no setting on your computer. Before anything runs, your agent checks the zip’s SHA-256 fingerprint against the one below.
- Answer its questions. It tells you what it installed and starts the first conversation.
SHA-256 fingerprint of the zip:
797bf60c883f91b639d64343520afa38a442c832a047d6fe3425c240a726bdc8
What it does
On your phone, tap Start. Pick the situation (a decision, a conversation you keep putting off, too much on your plate, or something else) and describe it in a sentence or two. For a decision you also say what you expect, how sure you are (50, 70 or 90%) and when to check back (2 weeks, 1 month or 3 months).
Then tap ChatGPT, Claude or Gemini. Your text, with good questions built in, is copied and your AI opens (on a computer, with one more click): paste it and send.
For a decision, the check-in day goes into your calendar with one tap, so it reminds you. On that day the app asks: what really happened? Your check-in history shows how often your forecasts matched what happened. You can also ask a second AI to argue against your plan.
No account, no laptop needed. Everything stays on your phone. (In the browser version the check-in date waits on the home screen; a web page cannot ring your phone.)
On your computer, if you want more. The kit lives inside the AI you already use: Claude (desktop, web or Code), Codex, Gemini and others. It learns your goals once, helps with real work and decisions in your own files, suggests a relevant idea from your published work, and on the day you set brings you back to what you decided.
Your files stay on your computer and your notes on your phone. What you send to an AI is handled by the AI provider you choose.
Who made this: Vadym Chernets, PhD, AI researcher; two papers at NeurIPS 2026 workshops: Verification Theater and Self-Confidence in LLM Evaluation.