Former Anthropic engineer warns AI agents may create false productivity
Thariq Shihipar, a former Anthropic engineer who worked on Claude Code, criticized the growing reliance on AI agents in a post on X. He described a ‘manic loop’ where developers feel productive from constant interaction with agents—prompting, testing, and rebuilding—without achieving meaningful progress. Shihipar suggested this activity can become a distraction, replacing strategic…
Key points
- Former Anthropic engineer Thariq Shihipar calls AI agent use a ‘manic loop’ that misrepresents productivity
- Workers spend **6.4 hours/week** supervising AI outputs, per Glean’s Work AI Index
- Only **13%** of digital workers say AI significantly improves their organization’s performance
Data from Glean’s Work AI Index supports his concern. Workers spend 6.4 hours a week supervising AI outputs, correcting errors, and refining results—a figure that exceeds the 36% of time spent using AI to actually produce work. Despite 75% of digital workers reporting AI boosts their productivity, only 13% say it significantly improves their organization’s performance. Critics like Andrej Karpathy and Sam Altman have also noted the pitfalls of over-relying on AI tools, emphasizing the need for intentional use over sheer volume.
The story so far
10 episodes →- Former Anthropic engineer warns AI agents may create false productivitythis story
Former Anthropic dev Thariq calls out the manic loop of AI agents
yourstory.com · 29 October 2026
Loading the full article…
This text was published by yourstory.com and written by Nucleus_AI. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Society & Work
All →- Theranos whistleblower warns AI boom risks repeat of fraud like Theranos · 1 src
- Cursor finds SSH backdoor in fake Claude Code extension · 3 src
- Epoch AI finds GLM 5.3 exploits Firefox patch gaps at $1,000 per version · 2 src
- Musk and Zuckerberg adopt Trump’s ‘superintelligence’ term on earnings calls · 1 src
- OpenAI outlines four-layer system to detect AI misalignment in advanced models · 4 src
Comments
via GitHub Discussions