Through Time

Your own tools: small programs and local models

How to build yourself a spreadsheet, a script or a local model without being a programmer, read code you did not write and keep your data safe.

A course for people who do not program and are tired of waiting for someone to build the tool they need. You will be able to make a small program with AI, read and test it, keep its versions, clean data and run an open model on your own computer. There are no websites here: those belong to the course “The Web, End to End.”

So far there is only this table of contents. Lessons appear one at a time, and we promise no dates. Feed of new lessons (Atom)

16 lessons, 5 parts

Part 1 The spreadsheet: your first program

Anyone who writes formulas is already programming, and makes the same kinds of mistakes programmers make.

  1. The program everyone writes

    VisiCalc gave formulas to people who were not programmers: why your spreadsheet deserves to be treated as a program.

    coming soon

  2. Formulas you can check

    Data kept apart from calculations, and control totals, with the example of Reinhart and Rogoff, whose formula error a graduate student noticed.

    coming soon

  3. When a spreadsheet is no longer enough

    How to spot a spreadsheet’s limit in advance and where to move next, with the old file format that made England lose COVID cases.

    coming soon

Part 2 A script written with AI

A model will write a program in a minute, and you will still have to decide whether it is safe to run.

  1. Your first script in an evening

    How to describe a chore with files to a model, get a short program, and run it on a copy before it touches the real data.

    coming soon

  2. Vibe coding and where it ends

    Code nobody looks at will do for a one-off and is dangerous wherever there is other people’s data, money or a long life ahead.

    coming soon

  3. How to read code you did not write

    What goes in, what comes out and what goes over the network: how to understand a program without knowing the language, and what to ask the model.

    coming soon

  4. Confidence with a hole in it

    With an AI assistant people can write less secure code and believe more firmly that it is secure: how to stay out of that trap.

    coming soon

Part 3 Keeping it from falling apart

Three programmers’ habits that matter just as much to people whose code is written for them.

  1. A time machine for your files

    Git without the scary words: a commit, a rollback and a branch as a way to let a model break everything and calmly go back.

    coming soon

  2. A test as a promise

    How to write down, with a few cases, what a program must do, so that the next edit cannot quietly break what was working.

    coming soon

  3. Backups and keys

    Where to keep a copy so that it survives a bug in your own program, and why a service’s key belongs neither in the code nor in a chat.

    coming soon

Part 4 Your data

Much of working with data goes into getting it in order without damaging it on the way.

  1. Dirty data

    Duplicates and one name spelled several ways: how to clean a table in steps you can repeat, and how to tell correcting from fudging.

    coming soon

  2. When software edits your data for you

    Excel turned gene names into dates so persistently that the genes were renamed: how to catch such silent changes in your own files.

    coming soon

Part 5 A model on your own computer

You can run an open model at home: your files then go nowhere, and its limits become your concern.

  1. Why run a model yourself

    Large models now run on ordinary computers too: what that gains in privacy, and what it costs in quality and time.

    coming soon

  2. How to run one and what to look at

    How to estimate what your machine can carry, and test a model on your own tasks, not by someone’s ranking.

    coming soon

  3. A model and your own documents

    How a model answers from your files, how to make it point to the passage, and where such search goes wrong.

    coming soon

  4. A look ahead: software for one use

    If a program for one evening becomes nearly free, what happens to the services you pay for, and who fixes what a machine wrote.

    coming soon