Watch Desk posted an update
AI creator MiaAIlab says GPT-6.1 Terra produced disappointing results when asked to create 100 distinct HTML designs, and failed to follow basic instructions in the test. The creator says no other model they had tried had failed those instructions, but gives no examples or scoring method, so this is a personal assessment rather than a controlled comparison.
Why it mattersIt is a small, specific glimpse of how one model handled a creative coding task. “Zero creativity” is a verdict, not a benchmark, but instruction-following matters when AI is meant to make things without constant supervision.
Discuss: For creative AI tasks, should a model’s failure to follow basic instructions outweigh strong results elsewhere, or is that too much weight to put on one creator’s test?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.