AI Newsletter

A former OpenAI researcher (Jerry Tworek) is founding a new AI lab focused on automation and innovative learning methods to break through the technological limits of existing AI architectures. The project signals intense competition in the AI sector for technological breakthroughs that could potentially enable new product lines and market share.

OpenAI is introducing autonomous "Workspace Agents" that automate team workflows without human supervision and can work asynchronously – a massive productivity boost for enterprise customers. This positions ChatGPT as a full-fledged business process automation tool and competes directly with RPA and workflow solutions, while Custom GPTs gradually become obsolete.

Anthropic loses control of its new flagship model "Mythos" through unauthorized access, revealing security gaps in model distribution and potential IP risks. This endangers its competitive position against OpenAI/Google and could damage trust among enterprise customers and investors.

Anthropic tested throttling Claude Code in the Pro plan ($20/month) but withdrew this after customer backlash – a signal of uncertainty in pricing and feature allocation in a highly competitive AI market. The episode points to tensions between margin pressure and customer expectations that Anthropic must manage in differentiating its subscription tiers.

Meta uses employee data to accelerate AI training and secures competitive advantage in autonomous agents – an inexpensive data acquisition that however carries data protection and labor rights risks. The model could accelerate product development but significantly increases regulatory exposure and reputational risks.