Anthropic product team sees personnel change, Claude Code lead departs

It is reported that Cat Wu, the product lead for Anthropic's Claude Code, has left the company. She recently posted on X that Friday was her last day at Anthropic, but did not disclose the reason for her departure or where she will go next.

Cat Wu was previously responsible for two products, Claude Code and Cowork. Before joining Anthropic, she worked as an engineer at Scale AI and Dagster, and later moved into venture capital at Index Ventures.

According to reports, she joined Anthropic in August 2024, initially serving as a product manager for research, and later took part in product management for Claude Code and Cowork. Her departure means a change in leadership for two of Anthropic's important AI products.

Putting Digital Fingerprints on AI Outputs: Anthropic Develops FlashProve to Leave Model Swapping Nowhere to Hide

Anthropic is developing an open-source verification system called FlashProve, aimed at addressing whether the AI answers users receive are truly generated by the model they claim to be. In a Tuesday blog post, the company said its 'provable inference' will use cryptography to provide a mathematical guarantee for every output, confirming that it came from a specified model, similar to how cryptographic email signatures verify the sender. The system can prevent model A from masquerading as model B, close a hidden backdoor for cost cheating, and expose inflated benchmark results, striking directly at a long-standing unspoken practice in the industry.

Anthropic's Killer Verification System FlashProve Officially Open-Sourced: Precisely Exposes Rebranded Models in 17 Seconds, Verification Speed Soars 5,000x

The prosperity of the open-source AI ecosystem has given rise to the chaos of 'rebranded models': some developers slightly repackage open-source models, swap weights, or tamper with outputs, then claim them as self-developed new models. To tackle this chaos at its root, leading AI company Anthropic has launched and open-sourced FlashProve, a provable inference system for large models. Relying on underlying cryptographic technology, the system can accurately verify a model's true identity and curb rebranding fraud.

Claude Issues Itself an ID Card: Anthropic Uses Zero-Knowledge Proofs to Cut Off the Gray Market of Model Wrapping

In the large-model supply chain, there is a gray area where users pay for top-tier configurations but run downgraded models. Anthropic, together with zero-knowledge proof researchers, has open-sourced FlashProve: before launch, a single H100 generates and publishes a "weight fingerprint" ID in about 39 minutes; after each conversation, the GPU generates a proof of about 19 MB; users can run cp_verify on an ordinary CPU to verify authenticity in 17 seconds, without needing weights or a GPU. The underlying zkSNARK uses about 2,000 sumchecks per request, with security relying on BLAKE2. AI writes code, Lea

Recorder Hidden in Toilet Paper Exposes OpenAI's Long-Standing Security Lapses

In October of this year, a recording device was secretly planted in OpenAI's executive meeting room. On October 13, the U.S. Department of Justice reported that the suspect was a 35-year-old Saudi citizen and then-OpenAI engineer Al-Fawzan, and the device was hidden in toilet paper. Several former employees were not surprised, saying the company had long-standing security lapses: as early as before ChatGPT's launch in late 2022, people had warned about security management issues, but they had not been taken seriously for years. Reuters disclosed this, citing five people familiar with the matter.

Anthropic product team sees personnel change, Claude Code lead departs

Related Recommendations

Using Its Own Large Model to Build Its Own Chips: Anthropic Forms 15-Person Chip Design Team, Claude Helps Accelerate R&D

Putting Digital Fingerprints on AI Outputs: Anthropic Develops FlashProve to Leave Model Swapping Nowhere to Hide

Anthropic's Killer Verification System FlashProve Officially Open-Sourced: Precisely Exposes Rebranded Models in 17 Seconds, Verification Speed Soars 5,000x

Claude Issues Itself an ID Card: Anthropic Uses Zero-Knowledge Proofs to Cut Off the Gray Market of Model Wrapping

Recorder Hidden in Toilet Paper Exposes OpenAI's Long-Standing Security Lapses