2020: robots, from a distance

• COVID. I was in Toronto, working with a robotics company in Vancouver
• I was remote anyway; in Vancouver, the space was split into three sections, one person in each at a time
• My work: simulation environments and teleoperation, operating robots remotely

A lab, and a university, in a browser

• A simulation environment can run in a web browser: much of the science and engineering of physical systems can be learned there, with nothing to install
• The universities were closed
• The idea: an online university where the only thing a student needs is a computer with a browser
• That is how Collegica started

Then, and now

• I never got to finish the idea; with AI coding assistants, it is much more feasible today
• A few weeks ago I started updating collegica.org
• Robotics, software development, AI, finance, and aging well

Three kinds of capital

• Mind, muscle and money: all three compound, all three decay
• All three are far cheaper to build early than to repair late

Know where you stand

Measure what a year costs you.

• Every financial plan starts with your current financial status
• Twelve months of actual spending, not a guess: your number
• Most people are wrong by 20–30% on their first guess

Let your AI do the reading

• I wrote collegica.org for AI agents as well as for people
• You don’t have to read the article
• Give it to your AI: it reads the method and does the work
• You are interested in the results, and in checking them

Install the desktop app

• ChatGPT, Claude or Gemini: the desktop app, not the website
• The desktop app can open files and folders on your computer and work on them
• It is safer: it writes code that processes your statements
• Your statements stay in one folder; the results come back to the same folder

Step 1 — One folder of exports

• Export from online banking for every account you spend from: chequing, each credit card, any line of credit
• The cards, not just chequing: chequing alone shows card payments and nothing about where the money went
• Six months if you have them; one month works
• Nothing else in the folder

Step 2 — The folder and the prompt

• Open your desktop AI app and start a new conversation
• Let it open your folder (the app asks for permission)
• Paste the prompt from the event page:
  collegica.org/events/monthly-budget
• Send it

Step 3 — Check what it found

It lists the files, the accounts, the date range and any problems, then waits.

• Is every account there, including every card?
• Do the dates cover the months you expect?
• Then tell it: go

Step 4 — Answer its questions

Largest amounts first. The kind of thing only you know:

• Was that cheque or e-transfer a loan? Has any come back?
• Was that large payment one-time, yearly, or the start of a monthly one?
• What do you pay yearly that these months don’t show: insurance, property tax, a trip?

Each answer is saved in budget-output/rules.md, with your reason.

Step 5 — Read the result, and check it

• Planning figure = recurring spending (measured) + your stated yearly items
• Spread: 90%–109% of the median is a number to plan on; 57%–134% is a question
• Card payments, savings and loans are not in spending; card purchases are
• Looks wrong? Ask it to show the ten biggest lines

What you get

An illustrative Toronto family of four, in an advisor’s form: collegica.org/events/monthly-budget/example

Your planning figure

• Write down your number, and whether each part is measured or stated
• Keep the budget-output folder; next quarter, new exports and the same prompt
• Then take it to a human advisor: it is where the conversation starts

AI does not replace people

• AI can tell you where you stand; what to do next is a human conversation
• We still need to talk to a human advisor, and it is better to
• That is what I did, and I am very happy about it
• Even after many successful investments, when I thought I knew everything

Quick questions

• The prompt and the practice statements: collegica.org/events/monthly-budget
• The method: A Year of Spending, from Your Own Statements
• The in-browser tool: collegica.org/owl