Pruning My Pothos
How applied AI systems actually work

Making applied AI systems easier to understand, inspect and build.

I build with AI, test what happens, and explain what I learn in plain language. Read the argument, see it drawn, and keep a copy for later.

Start with the mapFlip through a storyboard
evidenceplain wordscheckshypeSNIPpruning my pothos.Cut the claims. Keep the craft.
Plate · pruning_my_pothosHover a tag
Systems map

Where are you in the system?

Eight questions an applied AI system has to answer, in the order the answers tend to depend on each other. Each one is an article and a storyboard. Follow the sequence, or enter where the question becomes useful.

All articles All storyboards
  1. 01 · How models work

    What is actually generating the output?

    Storyboard
  2. 02 · Context & retrieval

    What can it see, and how did that get there?

    Storyboard
  3. 03 · Instructions & contracts

    What are we asking it to do, and what is enforced?

    Storyboard
  4. 04 · Action & authorization

    What can become a real effect?

    Storyboard
  5. 05 · Human judgment

    Where does a person have to decide?

    Storyboard
  6. 06 · Evaluation & evidence

    How do we know the behaviour was acceptable?

    Storyboard
  7. 07 · State & continuity

    What has to survive across a boundary?

    Storyboard
  8. 08 · Readiness

    Is there enough evidence and control for this use?

    Storyboard
Selected systems

Ideas that cut across the map.

Works On My Prompt

For everyone AI promoted to builder without asking. (Including me.)

Post-mortems written before the incident, cheat sheets and maps, for the part after the demo works.

All of Works On My Prompt
Stack

Test the idea.

Where the explanations get built and find out whether they hold.

See the whole stack, with source
Shelf

Trace what informed the work.

Open the shelf