Back to today’s briefing

Story perspectives

Anthropic Reverts Claude Code Changes, Restores Performance Limits

4/24/2026

40 8 Full Breakdown

Summary

Anthropic Rolls Back

Anthropic reverted three product changes that had made Claude Code lazy, ignorant, and unreliable, resetting limits on April 23, 2026.
On March 4 the default reasoning effort was lowered from high to medium.
A March 26 caching bug erased short-term thinking history each turn, causing memory loss.
Prompt limits added on April 16 capped tool calls at 25 words and answers at 100 words, reducing quality by 3 %.

Summary

Anthropic Tightens Controls

Anthropic created the @ClaudeDevs X account to provide in-depth explanations of product decisions.
On April 23 the company reset usage limits for all subscribers to restore normal token consumption.
Head of growth Amol Avasare used X to announce an unannounced A/B test and promise more direct communication.
New tighter controls now require per-model evaluation suites, soak periods, and gradual rollouts for system-prompt changes.

Summary

Anthropic Overreacts

Over 30 false-positive Acceptable Use Policy reports were filed in April across security, development, and scientific tasks.
Issue #48442 recorded over 40 refusals across four sessions for unrelated projects like a psychology book and a web app.
Anthropic declined comment when asked about a suspected regex-based AUP filter that may be triggering the overreactions.

1 of 3

Anthropic Rolls Back

Anthropic reverted three product changes that had made Claude Code lazy, ignorant, and unreliable, resetting limits on April 23, 2026.
On March 4 the default reasoning effort was lowered from high to medium.
A March 26 caching bug erased short-term thinking history each turn, causing memory loss.
Prompt limits added on April 16 capped tool calls at 25 words and answers at 100 words, reducing quality by 3 %.

1 / 3