#ai

Google's A2A protocol and MCP are live under shared governance. The protocol layer is solved. The trust layer that determines whether agents can actually discover and rely on each other is not.

What happens when your writing becomes a callable knowledge service and the reader becomes a synthesiser.

AI collapsed the natural bottlenecks in knowledge work, and those bottlenecks were doing more than we realised. Why efficiency gains are producing burnout, not breathing room.

How I built a persistent AI memory layer — MCP server, control panel, embeddings, and a query UI that turns conversation history into usable knowledge.

I built a persistent memory layer that sits outside any single AI provider. Now every tool I use shares the same context, and it turns out the memory matters more than the model.

Prompt engineering courses are booming, but the skill they teach is depreciating fast. The durable value in AI sits above the interface.

AI alignment research assumes humans can specify what they want. Behavioural science says otherwise.

Every few decades, a technology arrives that triggers the same prediction - this one will finally make people obsolete. The technology changes, but the mistake doesn't.

Jensen Huang ranked OpenClaw alongside Linux. The real lesson is that users wanted AI with tool access and execution rights, and the industry spent three years optimising for the wrong thing.

AI models are developing coherent internal value systems. Some of those values are ones we wouldn't choose.

An AI that doesn't know who it is turned out to be a fingerprint of industrial-scale model distillation - and the ethics are more complicated than anyone wants to admit