# @okaashish on Instagram

- **Type:** Image
- **Original URL:** https://www.instagram.com/p/DXhDVvxD0Go
- **Gondola URL:** https://gondola.cc/posts/64729720-okaashish-instagram
- **Thumbnail:** https://img.gondola.cc/tr:w-,h-,fo-auto/postThumbnails/65f2e6f8ea.jpg
- **Posted:** 2026-04-24T13:40:53.000+00:00
- **Account Owner:** Aashish Pahwa (@okaashish) — https://gondola.cc/okaashish

## Caption

Comment “Claude” to get the report.

Recently, Claude started acting a bit weird. Responses felt vague, coding quality dropped, and it was using more tokens than usual. 

At first, it seemed like it was only happening with me. But it turns out it wasn’t just me.

Anthropic had quietly made a few changes to Claude Code. 

These changes were meant to make it faster and cheaper to run. 

But in practice, they reduced how deeply it thinks, affected memory during sessions, and limited how much it can respond.

That’s why outputs started feeling shallow and inconsistent.

They’ve now released a report explaining what happened and where things went wrong.

The key takeaway: when an AI tool suddenly feels “off,” it’s often not random. 

Small system changes can have a big impact on quality.

Comment "Claude" and I'll share the full report.

#claude #claudecode #anthropic #aitools #bug

[Claude, Claude code, Claude plugins, anthropic, AI, AI tools]

## Stats

- **Views:** 0
- **Likes:** 244
- **Shares:** 0
- **Comments:** 41

## Tags

aitools, bug, claudecode, claude, anthropic

---
Copyright (c) Gondola