Is Claude dumb today?

The card above is today’s community verdict: a 0–100 score built from every vote cast since midnight UTC, blended with yesterday’s finalized score until enough people have weighed in. It is the same number that sits on the home page, minus the live needle.

If you are here because Claude just did something baffling, the fastest useful thing you can do is vote. One number, optional comment, done. Your vote nudges the score for everyone who searches this exact question after you, and the comments feed is where you find out whether it is just you.

High is good. The dial runs from clown to genius in five zones:

Why it does not swing wildly at 00:01 UTC. A fresh day starts with yesterday’s finalized score. Each new vote pulls the number toward today’s average, and once today has roughly as many votes as a typical recent day the average takes over completely. One angry vote at breakfast cannot tank the day; fifty can.

What the little numbers mean. avg is the raw mean of today’s votes before blending; vs yesterday is the change against the last finalized day. A low avg with a high score just means the day is young.

1. Check the official status page. status.claude.com lists claude.ai, the API, Claude Code and Cowork separately. An active incident explains a lot of “dumb” in about four seconds.
2. Start a new conversation. Long chats degrade; Anthropic calls it context rot. If the fresh chat is sharp, the model was fine and the conversation was the problem. In Claude Code that is /clear, or /compact if you need the history.
3. Look at the effort and model you are actually using. Defaults change. In April 2026 a default effort change made Claude Code feel dumber for weeks before Anthropic reverted it; the developer guide has the details and the /effort fix.
4. Check your usage limits. A throttled session and a dumb model feel identical from the outside.
5. Then vote. Especially if all four checks passed. That is exactly the data this site exists to collect.

“Is Claude dumb today?” and “Has Claude been getting dumber?” are different questions with different evidence. Today is the dial. Lately is the score history: one finalized verdict per UTC day, so a single bad afternoon shows up as a dip and a genuine regression shows up as a slope.

If the slope is what worries you, the nerfed guide walks through what Anthropic has actually admitted to, and why Claude seems dumb covers the nine documented reasons a model that did not change can still feel worse this week.