
The wider AI landscape this week is splitting into two tracks. One is about interfaces getting faster, cheaper and more capable. The other is about the real-world constraints catching up, whether that is safety, regulation, pricing, trust or infrastructure.
For Project Flux readers, that matters because the signal is no longer confined to model launches. It is showing up in procurement, planning, budgeting, staffing, software consolidation and the practical question of where AI fits into delivery work. The stories below are the ones worth tracking.
Agentic AI, Tools, and Interfaces
Microsoft Is Removing Copilot Buttons From These Windows 11 Apps. Microsoft is stripping Copilot out of built-in utilities where usage never really stuck, which is a useful reminder that AI earns its place when it sits inside real workflows rather than beside them. Source
Meta's New 'Personal Superintelligence' AI Is Coming to Its Smart Glasses. Meta is pushing its next-generation assistant into smart glasses, signalling a very different strategy from the desktop-first race between OpenAI and Anthropic. Source
Microsoft Copilot Studio Gets MCP Server Support. MCP is increasingly looking like the shared connector layer for enterprise agents, and Microsoft's support matters because it brings that pattern closer to mainstream business users. Source
Anthropic Rebuilds Claude Code Desktop App Around Parallel Sessions. Running multiple agent sessions side by side pushes AI coding another step away from single-chat assistance and closer to actual team-like orchestration. Source
Anthropic Debuts Claude Design for Prototypes, Pitch Decks, and Mockups. Claude Design looks especially relevant for consultancy and bid work because it compresses the journey from rough concept to presentable output. Source
OpenAI Takes Aim at Anthropic With a Beefed-Up Codex. Codex moving further onto the desktop reinforces the point that the next battle is not just model quality, but control of the working environment around the user. Source
Safety, Governance, and AI Economics
Trump Officials May Be Encouraging Banks to Test Anthropic's Mythos Model. This is one of the clearest signs yet that safety-withheld frontier models may still be pushed back into the market through political and institutional demand. Source
Finance Ministers and Top Bankers Raise Serious Concerns About Mythos. The Mythos debate is no longer a lab story. It is moving into finance, regulation and state capacity, which is where frontier AI governance gets materially harder. Source
AI and Human Behaviour. The Behavioural Insights Team makes a simple but important point: AI changes behaviour before it changes outputs, which is exactly why rollout and governance cannot be treated as separate conversations. Source
Reid Hoffman Weighs In on the Tokenmaxxing Debate. The argument here is that the real economic prize may sit with smaller, cheaper models rather than endless compute escalation, which is highly relevant for firms trying to budget AI use sensibly. Source
AI in the Built Environment
OpenAI Pulls Out of UK Investment Project Over Energy Costs and Regulation. This is a meaningful built-environment story because it links AI ambition directly to energy pricing, regulation and the real economics of infrastructure delivery in the UK. Source
Virginia Tech's MARIO Robot System Could Cut Construction Costs by 30%. MARIO is worth watching because it brings robots, drones and digital twins together around a live inspection problem rather than a polished demo. Source
Graitec Earns Autodesk Gold Partner Status. Graitec's move is another sign that AI implementation in AEC is shifting into the established channel ecosystem, not staying with niche specialists. Source
McKinsey and ALICE Technologies Form Alliance. AI scheduling and optimisation are becoming advisory-grade topics now, not just specialist software topics, and that broadens the audience that needs to understand them. Source
OpenAI Signs Lease for First Permanent London Office. The London move matters because AI's growth is now shaping office demand, talent concentration and the geography of the sectors building around it. Source
That is all for this week’s round-up. We’ll come back with more next week.
All content reflects our personal views and is not intended as professional advice or to represent any organisation.

