AI has removed the constraint that made minimum viable products necessary. What it has not removed is the risk of being 99% wrong — it has just moved the discipline from engineering capacity to product design.
Washington is treating AI as a weapon to be controlled. Beijing is treating it as infrastructure to be spread. The world runs on what it can depend on, not on what is most powerful, and 2026 has already shown which strategy wins.
A new platform launches with a wager: that AI should work inside a company's own systems, not beside them — turning customer operations from a room full of people doing lookups into something a small team supervises rather than staffs.
AI transformed software development almost overnight. So why hasn't it done the same for the rest of the enterprise world? The answer isn't about AI capability — it's about prerequisites.
When MCP connects knowledge across silos in large healthcare and medtech companies, something more significant than efficiency emerges. A look at what these organisations become when the full picture is always present.
The MCP architecture for large organisations is shaped by what already exists. A look at the gateway model, federated domain ownership, the legacy interface pattern, and the phased rollout that builds organisational trust.
Large healthcare and medtech companies are not short of data. They are short of the organisational structure to let it move. MCP is the technical unlock — but the harder question is structural.
How we built a production computer vision service on a Chinese 8B parameter model that costs almost nothing — and why that was the right engineering decision.
What changed our output quality when building a production platform with Claude was not better prompting. It was explicit verification, built into every task definition.
The /insights command in Claude Code holds up a mirror. What you see reflected back is more interesting than you might expect.
A look at how large organisations are adopting AI at the infrastructure level, not just the application layer.
How AI changes the rhythm of building products — and what that means for teams, processes, and the definition of done.
Why platform teams build for everyone and end up serving no one — and how to escape the trap.