Tech • IA • Crypto
Musk vs OpenAI trial, Gemini 750M users, Mistral 3.5
Briefing
Musk vs OpenAI trial escalates
The Elon Musk–OpenAI trial entered a pivotal second week in Oakland, with over seven hours of testimony. Evidence centered on Greg Brockman’s $30 billion stake and early governance disputes. Testimony described clashes over control structures and leadership authority. The case is exposing foundational tensions that shaped OpenAI’s trajectory.
OpenAI faces growth and cost pressure
OpenAI has reached roughly 1 billion weekly users but is struggling to meet revenue expectations. Internal concerns, including from CFO Sarah Friar, highlight sustainability risks. The company’s massive $600 billion infrastructure commitment is raising scrutiny. Analysts question whether demand will justify such unprecedented spending.
Google Gemini hits 750M users
Google Gemini has rapidly scaled to about 750 million users, signaling major competitive momentum. Its growth is powered by Google’s vertically integrated infrastructure and capital resources. This advantage reduces dependence on external compute markets. The surge positions Gemini as a leading challenger to OpenAI.
Compute bottleneck reshapes AI race
Access to compute power is emerging as the defining constraint in AI competition. OpenAI has pursued aggressive capacity acquisition, while rivals like Anthropic face shortages. This imbalance limits deployment despite strong demand. Infrastructure strategy is now as critical as model quality.
CASI expands pre-release model oversight
The U.S. Department of Commerce’s CASI has secured agreements with OpenAI, Google, Microsoft, Anthropic, and xAI. The agency has conducted over 40 model evaluations, including on unreleased systems. Reviews occur before public deployment, increasing federal visibility into AI development. However, most findings remain undisclosed.
Regulatory bottleneck fears grow
Experts warn that pre-release AI screening could slow innovation cycles. Companies may face delays bringing models to market under CASI oversight. This could shift competitive dynamics toward firms with greater regulatory resources. The balance between safety and speed is becoming a central policy debate.
Claude 4.7 rewires agentic prompting
Claude 4.7 introduces a structured prompt system spanning up to 50 pages of instructions. It enforces a rule to search before answering factual queries, increasing token usage but improving accuracy. The system revolves around context retrieval, action selection, and verification. This marks a shift toward fully agentic AI behavior.
Mistral Medium 3.5 challenges incumbents
Mistral Medium 3.5 launches as a European alternative with open-weight access and strong performance gains. It excels in coding, function calling, and agentic workflows. The model is freely usable up to $20 million monthly revenue, after which licensing applies. Its release strengthens Europe’s push for AI sovereignty.
Videos covered
Ce que Sam Altman vous cache derrière les vrais chiffres d'OpenAI !
- •OpenAI’s mounting challenges
- •The compute bottleneck reshapes competition
- •Google Gemini’s rapid rise
Sam vs Elon Trial, Coinbase Cuts 14% in AI Pivot, Digesting Meta Earnings | Diet TBPN
- •Musk vs OpenAI trial intensifies
- •Disputes over control and AI expertise
- •Government expands AI model oversight
Claude 4.7 le prompt system enfin accessible !
- •A structured and dense prompt architecture
- •Three fundamental pillars of reasoning
- •A deeply modified search logic in 4.7
Mistral 3.5 : l’IA qui code seule
Read full article →Actus, débats, tutos, IA, web, tech : on décode et on forge le progrès !
- •New flagship AI model rollout
- •Prompting paradigms are बदलing
- •Anthropic and OpenAI escalate competition