Weekly

The planning machinery behind an AI infrastructure company. A useful reminder that technological growth still depends on coordinating physical supply.

Swiss technology for recycling PET and polyester into raw materials. Circularity becomes more tangible when you look at the chemistry and the industrial process.

An early-access family-office platform built around local data, persistent context and traceable agent actions. An interesting example of AI designed around a specific operating environment.

Turning written material into something you can listen to. Sometimes the useful improvement is changing how information reaches you.

Simulated populations for testing products, pricing and decisions. The important question is how well simulated behaviour predicts what real people subsequently do.

A control layer for enterprise AI agents. Visibility, permissions and runtime enforcement become more important as software gains the ability to act.

A practical look at whether agents can discover, understand and use a website. The findings underneath the score are the useful part.

Creative technology education through making things. The student work is a good place to start.

A school model organised around adaptive learning and time for practical skills. Worth examining both the educational design and the evidence behind its claims.

An agent that proposes algorithms, tests them and improves the promising ones. The evaluator is as important as the model.

An AI workspace built around the idea that company knowledge should compound. The useful question is what the organisation remembers after an individual finishes a task.

Developing chip-patterning technology using atoms instead of light. A reminder that the future of computing also depends on difficult advances in physics.

Agent-assisted investigation of applications and binaries. Interesting tooling for understanding how existing software actually works.

A thought experiment about agents on both sides of commerce. What happens to marketplaces when finding and comparing alternatives costs almost nothing?

Watermarking generated content gives provenance something concrete to work with. It answers a narrower question than whether something ‘sounds like AI’.

One of cinema's most recognisable images of technology began with Japanese cookbooks and hand-painted characters. A lovely piece of design history.

AI applied to formulation and physical product development. An interesting connection between proprietary data, scientific experimentation and commercial constraints.

Turning a document into a whiteboard explainer. Making a video is getting easier; deciding what needs explaining still takes work.

Expressive speech generation and voice tools. Another part of the production process becoming accessible without a studio.

Portable satellite connectivity, and what changes when reliable internet becomes something you can carry.

A reminder that the value of AI is increasingly measured by judgement and prediction, not only automation.

A sharp example of why capable agents require security boundaries, observability and clear accountability.

A new agentic development environment moving from a controlled waitlist to general availability.

Why most AI requests can run on self-hosted open-weight models while frontier models are reserved for the hardest tasks and higher-level review.

A detailed account of sophisticated AI misuse and the safeguards used to detect, disrupt and learn from it.

A three-day attempt to build a company with an AI agent across planning, product, engineering, sales, support and marketing.

On's move into football, and how a new sport creates a new expression of the brand across athletes, product, community and technology.

Why treating AI as an autonomous collaborator rather than a software tool fundamentally alters organizational delegation and management.

Enterprise engineering patterns for context compression and developer experience in large-scale agentic codebases.

A rigorous framework distinguishing simple prompt chaining, routing workflows, and autonomous evaluator-optimizer loops.

Bridging foundation models and physical embodiment to create generalized robot control policies.

Why iterative reflection and tool-augmented agent loops consistently beat next-generation raw foundation models on complex reasoning.

Structural shifts in global sports business, athlete-led media IP, and modern DTC fan monetization.

Emerging discovery and payment primitives designed for agent-to-agent transactional systems.

Local, private Apple Silicon speech-to-text that removes friction from drafting thoughts and meeting notes.