AI attempt
You ask AI to make, solve, find, or do something.
prompt: what AI could not do yet
Record what AI can’t do. Yet.
CantYet turns anything you wanted AI to do — but it still couldn’t — into a first-hand record: what you wanted, what you tried, and what happened. Keep it private, or publish it to find others waiting for the same possibility — and see when it finally becomes possible.
Connect once. Then say, “Log this to CantYet.” New Yets start private.
Private Yet (recorded)
CantYet captures what you wanted, what you tried, and what happened.
Saved as a private Yet
Published signal (optional)
Publish to find others waiting for the same possibility.
Published Yet
What AI still can’t do
Open Yets indexFrom attempt to proof
Capture the edge
When AI gets stuck, log it to CantYet. We timestamp the attempt and preserve exactly what you saw.
Private Yet saved
Attempt
Make demand visible
Publish your Yet to find others hitting the same wall. The record becomes a signal for the ecosystem.
Published Yet
Waiting
Watch the frontier move
When a model or tool update solves it, CantYet marks the change so your record becomes proof.
Solved by an update
Solved
Anatomy of a Yet
Converting GPT ImageGen 2 logos to SVG with GPT-5.5 does not work reliably
What I wanted
I wanted to turn good-looking images or logos from GPT ImageGen 2 into practical SVG assets using GPT-5.5.
What I tried
I generated a good-looking image or logo with GPT ImageGen 2, then asked GPT-5.5 to recreate it as SVG.
What happened
The raster image looked good, but the SVG produced by GPT-5.5 lost shape fidelity and did not become a usable logo asset. It felt like SVG creation needs a model or workflow specialized for vector graphics.
New Yets start private.
Only the Yet you log is sent to CantYet. It starts private; you decide what becomes public.
Connect CantYet to your AI
Connect once. Then say, “Log this to CantYet.” New Yets start private.