This week’s curated links show AI moving deeper into the systems that run companies, assets, regulation and physical work. The common thread is operational maturity: more capable agents, more expensive infrastructure, more pressure on governance, and more evidence that adoption is spreading faster than organisational design.
Enterprise agents and work redesign
Snowflake CIO says he used layoffs to convince staff to use AI: The story is uncomfortable but important because it shows how AI adoption can become entangled with workforce pressure, culture and trust.
Meta targets the business AI market with agents that close sales: Meta’s business agents point toward AI systems that book appointments, take payments and complete customer interactions, moving automation closer to revenue operations.
Microsoft launches Scout, an OpenClaw-inspired personal assistant: Microsoft’s agent work shows how personal assistants are being positioned as always-available workflow layers, which raises practical questions about audit trails and organisational control.
Mira Murati steps back into the spotlight, carefully: Murati’s return is worth tracking because leadership moves in AI often signal where talent, capital and new product bets may gather next.
Ahead of its IPO, Anthropic’s Daniela Amodei shrugs off doubts about AI’s returns: The interview gives a useful view of how a leading AI company is framing the gap between current spending and the long-term commercial case.
Reid Hoffman is leaving Microsoft’s board to go founder mode with startup Manus: Hoffman’s move underlines how experienced operators are still choosing direct company building in the agent market.
Microsoft Build 2026: Build’s AI focus matters because Windows, developer tooling and enterprise cloud services are being aligned around agent-based workflows that many organisations will inherit by default.
OpenAI’s GPT-5.5 in Microsoft Foundry: Microsoft’s Foundry positioning shows how frontier models are being packaged for enterprise environments where deployment, governance and integration matter as much as raw capability.
GitHub Copilot is moving to usage-based billing: Copilot’s billing change is an early sign that AI productivity tools are moving from fixed seat assumptions toward consumption management and internal cost discipline.
Infrastructure, compute and the economics of AI
Alphabet plans to raise $80 billion from stock sales to fund AI build-out: Alphabet’s capital plans show the scale of funding required to support AI infrastructure and the pressure this creates for investor narratives.
NVIDIA and Microsoft reinvent Windows PCs for the age of personal AI: NVIDIA’s RTX Spark work signals the continuing push to bring more AI capability onto personal computing hardware.
DeepSeek nears US$7 billion haul in first-ever funding round: DeepSeek’s reported raise shows that model competition is still attracting very large pools of capital outside the US market.
Anthropic confidentially submits draft S-1 to the SEC: Anthropic’s filing marks a step toward public market scrutiny of frontier model economics, revenue quality and long-term infrastructure costs.
Governance, safety and legal pressure
Trump signs AI executive order asking companies to give the government early access to models: The order highlights the growing policy interest in pre-deployment review, cybersecurity coordination and earlier government visibility into advanced models.
Open letter in support of mandatory nucleic acid synthesis screening and recordkeeping: The biosecurity letter signed by leading AI figures shows how model risk debates are expanding into adjacent scientific and supply chain controls.
CNN sues Perplexity over alleged AI copyright theft: The lawsuit adds to the unresolved question of how AI companies use publisher material and what evidence will be needed to prove misuse.
Anthropic Institute on recursive self-improvement: Anthropic’s research note is a useful primary source for readers following how labs are thinking about the feedback loops that could make advanced systems improve faster.
Introducing new capabilities to GPT-Rosalind: OpenAI’s scientific AI work is a useful marker for how specialist reasoning systems are being shaped around high-value research workflows.
Robotics and autonomous systems
This Trump-linked startup plans to put humanoid robots in the military: Foundation Future Industries’ military robotics work shows how quickly humanoid systems are being tested in high-risk operational settings.
Exclusive: Uber to deploy 500 data collection vehicles this year: Uber’s fleet plan shows the continued importance of real-world data collection for autonomous vehicle development and mapping quality.
AI in the built environment
Six contech startups raise a combined $121m: The funding round-up gives a useful snapshot of where investors see near-term value in robotics, predictive scheduling, drawing review and construction automation.
Veras included with Enscape, V-Ray and Corona: Chaos embedding Veras into familiar visualisation products makes AI ideation more likely to appear inside routine design workflows.
We’ll start our deep-dive for next week, as you enjoy this read…
All content reflects our personal views and is not intended as professional advice or to represent any organisation.
