Human signal over synthetic scoreboards
Which model feels strongest in your hands?
Instead of looking at endless sponsored synthetic benchmarks that do not reflect how a model really behaves in actual work, tell us how you personally feel about it and see what others feel. If any black-box model or subscription has been silently downgraded, see it here live on the community-contributed chart. Let's unite and protect our rights ourselves.
The silent-downgrade radar
How does it feel today?
Model weights are a black box. A provider can quietly swap, quantize, or throttle a model and never ship a changelog. Log how each one feels today; the weekly trend lines below expose a model that has silently gone duller — or sharper — over the last two weeks, when no release note admits it.
Loading trends…
Community field notes
Field notes
Longer observations a single vote cannot hold: a regression you noticed, a prompt that stopped working, a model that quietly improved. Not a chat — a slow, public notebook.
Join the community forum
Field notes here are quick takes — full threaded discussion, categories, and history live on the forum.
Forum — coming soonStart a thread
- Loading field notes…