Hand Aldo a job the way you’d hand it to a capable assistant. It works on computers of its own, checks the result, asks before anything that matters and tells you when it’s done.
In a thread, by text or on a call. Attach the files, or say where they are. A sentence is enough.
The AP aging macro takes 20 minutes and freezes Excel. Make it fast.
Desktop Excel on Windows, a Chrome signed in as you, your code. Yours stays free, or shut.
It reads the 48,212 open items one cell at a time. I’m rewriting it to read them once and age them in memory.
It reruns the macro, ties the deck to the model, clicks through the app. The evidence comes back with the result.
Done. BuildAging runs in 3.1 seconds now, and its output matches the old macro’s on all 48,212 rows.
Paying, booking, cancelling and applying: asked every time, with the total and the terms. Email and invites wait for your yes, or a rule you made. This card’s buttons work.
File the Q3 sales and use tax return?
Taxable sales of $201,402 after resale and out-of-state exemptions, from the Q3 sales export. The return is filled in on the portal and stopped at Review and submit.
On your phone or your desktop. Every job lands on your Home screen with what it made, so you can check it in a minute.
Q3 board deck
Lark
Your Q3 deck is ready. All 20 figures match. Revenue: $3.3M. EBITDA: $600,000.
October month-end close
Cedar Workshop
October is closed. Profit: $753.92. The $1,093.00 bank balance matches.
The macro is an example. Lark and Cedar Workshop are sample companies, made for Aldo’s demonstrations; their jobs, and the checks on them, are real.
This is Aldo’s Home screen, replaying seven real runs, sped up. In the app, open any job to watch its computer, or take the mouse.
Chatbots hand you text to paste. Aldo takes the work: the model that won’t tie out, the deck due at nine, the macro nobody wants to touch.
| A chatbot | Aldo | |
|---|---|---|
| Where it works | In a chat window | On computers of its own: desktop Excel, PowerPoint and Word, a Chrome signed in as you, your code |
| What you get back | Text to paste somewhere | The finished work: the model, the deck, the merged pull request |
| Who checks it | You do | It does, and the evidence comes back with the result |
| Paying and sending | Left to you | It asks first. Payments wait for your yes every time |
| While it works | You wait in the tab | You go do something else. It tells you when it’s done |
Each thread gets computers of its own, set up for the job, and works in the real apps.
Chrome with your saved logins, typed in for it and never shown to it.
Real Excel, PowerPoint and Word, so macros, queries and formatting survive.
An RTX 4090 for Blender renders and other heavy work.
GitHub, GitLab, Bitbucket and Azure DevOps. You review before it merges.
Aldo’s agents run on your Claude, ChatGPT, Grok or Muse plan, or on Aldo’s own models. Pick the model for the job, and switch when a better one ships.
Real access, on your terms. What it does on its own, it learned from you.
It ties the deck to the model, reruns the macro and clicks through the app. The evidence comes back with the result.
Saved logins are typed in for it and API keys are added on the way out, so the agent never sees either.
You approve each one there, and the card it gets works once, for that amount, at that store.
Threads never share a machine. Watch any job beside its thread, and take the mouse when you want.
