AI Newsletter

Mistral has released the weights of its 1-billion-parameter model Large 4 and now leads open-weight benchmarks in the coding space. According to benchmarks, the model outperforms Kimi, DeepSeek, and GLM on Terminal Bench 4 and is considered the strongest open coding agent worldwide. For the market, this represents a massive boost for sovereign, European AI infrastructure and significantly increases pressure on proprietary providers like OpenAI.

Despite technical catching up, the revenue gap between Mistral ($400M run-rate) and OpenAI (~$70B) and Anthropic ($65B+) is dramatically widening. Mistral raised $830M in external capital in 2026 to purchase 13,800 Nvidia chips – a risky growth betting model. The numbers show that technical excellence alone is insufficient: sales, ecosystem lock-in, and enterprise trust remain the decisive competitive moats.

Polymarket prices an Anthropic IPO by year-end at 87% probability – Anthropic has filed S-1 documents and plans a Nasdaq listing under the ticker 'ANT'. New compared to earlier reports: market capitalization expectations have fallen 21.7% over the course of the month, signaling investor skepticism following initial hype. The IPO will become a stress test for the entire AI valuation logic of the industry.

Goldman Sachs forecasts AI investments of $1 billion in 2026 alone, but according to BBC, political change following the US midterms could significantly slow capital flow into the sector. The sector has absorbed so much capital that external capital costs for other industries have risen – a systemic risk. The question of whether the AI investment boom generates fundamental returns or creates a bubble becomes the main political-economic topic of Q4 2026.

According to current tool reviews, the era of manual prompt entry is officially over: leading productivity tools like ChatGPT Work, Claude's Background Mode, and Notion's Custom Agents run multi-hour autonomous tasks without human intervention. According to Coworker.ai, companies deploying enterprise AI productivity tools complete 126% more projects per week. This structural shift in human-machine interaction accelerates questions about liability, control, and compliance in regulated industries.