Today in AI — 10 July 2026

Digest

OpenAI shutting down Atlas is the sharper product signal this week. The centre of gravity is shifting from chat as a destination to AI as a work surface: embedded, measurable and judged by whether it completes useful tasks.

The work surface wins

The strongest product signal is consolidation. OpenAI’s new model and ChatGPT Work point towards a broader productivity push, while the end of ChatGPT Atlas suggests standalone AI browsers may be too narrow a bet.

• OpenAI rolls out GPT-5.6 after government greenlight — and announces ‘ChatGPT Work’ — The Verge
• The ChatGPT browser is already dead — The Verge
• Meta says its new AI model is ready to compete on coding — The Verge
• Say hello to Claude Wrapped — The Verge

For builders, the lesson is blunt: distribution and workflow placement matter as much as model capability. Claude Wrapped also hints at a product layer where AI usage becomes something users can review, compare and change.

Agents move into creative tools

Creative AI is becoming more practical. The shift is from asking a chatbot for advice to letting an assistant act closer to the tool where the work already happens.

• Character.AI wants a piece of the microdrama pie — The Verge
• FL Studio 2026 turns its AI chatbot into your assistant engineer — The Verge
• CanvasAgent: Enabling Complex Image Creation and Editing via Visual Tool Orchestration — arXiv

This is a familiar software pattern: once a feature becomes useful, it stops living in a side panel and starts acting on the canvas, timeline or project file. The product question becomes permission, control and reversibility, rather than novelty.

Inference economics and enterprise AI

The infrastructure stories are about cost, control and optionality. ZML is pushing free inference software across chips, SambaNova has raised another large round as an AI chip maker, and Skello shows funding still flows to AI packaged inside operational software.

• Hot French startup ZML releases free product to speed inference across lots of AI chips — TechCrunch
• AI chip maker SambaNova raises $1B at $11B valuation, 5 months after last mega round — TechCrunch
• Paris-based Skello, which provides AI-powered HR tools for frontline workforce management, raised €200M led by Bridgepoint to support its European expansion — Techmeme

Skello is the useful counterweight: AI funding is not only chasing model labs. Plain operational categories still have room for AI when there is a clear buyer and a clear workflow.

Trust, communication and physical deployment

AI is also becoming a governance and communication problem. Labels on AI-made ads, warnings about false scientific interpretations, advice on announcing “AI-first” plans and AI in vehicles all point to the same issue: users and workers need to know what the system is doing and why.

• Google will now tell you if an ad was made with AI — The Verge
• Scientists say AI is falling for 'alien hoaxes' too easily — and that's a problem for research — TechRadar
• Don’t become a meme: How to tell your team you’re going AI-first — Fast Company
• “It's never been a more exciting time to imagine what's possible with software”: Xero CEO outlines how AI could transform the finance industry as we know it — TechRadar
• The SpaceXAI logo is so bad that even Grok hates it — Fast Company
• Cadillac says it will be first to mass-produce Momenta’s R7 world model — IT Home

The takeaway for product teams: stop treating AI as a feature announcement. The winners will place it where work happens, explain it clearly, and make its costs and limits visible.

---