TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Judgment · AGE 18—22 · POSITIONAL · METHOD
Build a personal failure catalogue
A designer's eye grows fastest on its own written-down failures.
When to use
Every time an AI output got past you: the image you believed, the brief that misfired, the polished answer you pasted before checking. Especially the ones that embarrassed you, because those are the ones you will otherwise forget.
The method
One entry per failure, three lines. What fooled you: the output, kept as a thumbnail or a quote. The tell you missed: what gave it away in hindsight. The shadow, the missing source, the too-neat table. How you now catch it: the check you added, in one sentence. Re-read the catalogue before every jury and every client meeting. Over a term, sort the entries. The tells that repeat are your blind spots, with names on them.
Watch for this
Recording the machine's mistake instead of your own. "The model hallucinated a byelaw" is not an entry. "I pasted a byelaw table without asking for its edition" is. The catalogue is about your eye, not the tool's faults.
The Lab's note
The catalogue is the Lab's method for turning a private thing, being fooled, into something that can be taught. It records what you did not yet know to distrust. We hold that a designer's eye grows fastest on its own written-down failures, and more slowly on anyone else's warnings, including ours.
Try it
Start the catalogue this week with the last three outputs that got past you. For each, write the tell you missed. Then pick the one you suspect you repeat most and design one check that would have caught it.
Prove it
Name one failure you repeat, straight from the catalogue, and show the check you now run for it. The answer should come from a record, not a guess.
How it works
Knowing what you know, and noticing when you do not, is what psychologists call metacognition: monitoring your own thinking. Research on working with generative systems finds that much of the difficulty is metacognitive. The hard part is not the tool. It is judging the tool's output against your intent, and knowing how much to trust that judgment. A failure catalogue is a metacognitive aid. It keeps the evidence of where your confidence was wrong in one place, where it can correct you the next time the polish arrives.
What this idea builds on
What this idea opens up
Sources
- collaborative-red-pen
- disciplined-iteration-log
- Tankelevitch et al., The Metacognitive Demands and Opportunities of Generative AI
- The Seduction of the Machine
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.