Research notes · 2026
The useful stuff I am thinking through right now: agentic AI, the next wave of software, high-friction problems, and what is actually worth building when the tools change every few weeks.
Featured note
The first long-form field guide follows the whole agentic AI stack, from the question people ask to the system you can responsibly ship.
A practical, source-backed field guide to agents, workflows, MCP, A2A, security, evaluation, and the AI app ideas that are worth testing now.
Open the field guideWhat I am tracking
The useful line between a chatbot, a workflow, and a system that can plan and act.
Problem-first product angles built around expensive friction instead of novelty demos.
Protocols, memory, computer use, multimodal systems, and the infrastructure catching up.
What makes useful research discoverable to people, search engines, and answer systems.
The method
Fast-moving tech is full of confident takes that age badly. These pages separate what a primary source says, what I am inferring, and what is still an open question.
The goal is not to predict the future perfectly. It is to make better decisions with the information we have, then build small enough experiments that reality can answer back.
Have a workflow worth investigating? Start hereSend the messy version. We can turn it into a clear first build, an honest research brief, or a decision that saves you from building the wrong thing.
