ENFR
8news

Tech • IA • Crypto

TodayMy briefingVideosTop articles 24hArchivesFavoritesMy topics

Anthropic Profit Debate, Google Genie 3, France €10B AI Gigafactory

AIThursday, May 28, 2026· 17 videos

Briefing

Audio player
0:00 / 0:00

Anthropic profit claims face scrutiny

Anthropic reported $10.9 billion in quarterly revenue and $559 million in operating profit, marking its first profitable period. Growth of 130% was driven largely by a shift to usage-based pricing, with enterprise bills rising to $1,000–$2,000+ monthly. Սակայն analysts point to massive infrastructure costs, including a reported $1.25 billion per month deal with SpaceX, which could erase margins. The results raise questions about whether profitability reflects sustainable economics or temporary advantages.

Google Genie 3 builds world models

Google is testing Genie 3, an AI system that converts images into fully navigable, real-time environments. By integrating Street View data, the model enables exploration of realistic global locations with spatial consistency. Users can modify environments dynamically, transforming real-world scenes into alternate simulations. The advance signals a shift from static generation to interactive world modeling, with implications for gaming, mapping, and training simulations.

France pushes €10B AI gigafactory

France and partners including EDF, Orange, and Capgemini are backing a €10 billion AI project named ION. The facility could scale to 1 gigawatt of compute, aiming to reduce Europe’s reliance on U.S. and Chinese infrastructure. Europe currently holds just 5% of global AI compute despite 20% of GDP, highlighting the gap. The പദ്ധ emphasizes private funding and low-carbon energy, positioning sovereignty and sustainability as strategic priorities.

xAI scales Grok V9 to 1.5T

xAI has trained Grok V9, a 1.5 trillion parameter model, tripling its previous size. The system leverages data from Cursor, used by 67% of Fortune 500 companies, to improve real-world coding performance. Alongside it, Grok Build introduces agentic programming with parallel execution and a $300/month price point. Despite scale, benchmarks suggest the model still trails leading competitors in enterprise adoption and performance.

Mythos AI finds 10,000 vulnerabilities

Mythos, an autonomous cybersecurity AI, identified around 10,000 critical vulnerabilities across 50 organizations. Human review of 1,700 cases confirmed roughly 90% accuracy, indicating high reliability. The findings prompted warnings from the European Central Bank بشأن systemic financial risk. The system represents a shift toward automated, large-scale security auditing with minimal human intervention.

Claude Opus 4.8 adds effort control

Anthropic released Claude Opus 4.8 with incremental improvements and unchanged pricing at $5/$25 per million tokens. A new effort mode lets users control compute levels from low to max, balancing cost and performance. The model emphasizes improved honesty and broader usability rather than major capability leaps. კონკurrent comparisons show gains over Opus 4.7, though GPT-5.5 still leads in agentic coding tasks.

OpenAI Agents SDK targets production

OpenAI updated its Agents SDK to support long-running, autonomous systems. A Codex-style harness automates tool use, context management, and task loops for agents operating over days. New architecture separates execution environments from orchestration, improving reliability and security. The release reflects a broader shift toward deployable AI agents capable of handling complex workflows continuously.

AI film 'Hell Grind' stirs Cannes

Hell Grind, a feature-length film created entirely with AI, surfaced around the Cannes festival. Produced by a team of about 15 people, it used tools like Xfield and DiaCens 2.0 to maintain narrative and visual consistency. The project highlights how AI can replace traditional film crews across production stages. Its emergence has intensified debate over disruption in creative industries and the future of filmmaking.

Videos covered

Previous briefings · AI