Now
Now
I’m spending a lot of time on the harness around agents.
The models are getting good. The bigger opportunity now is the system around them.
I’m focused on how agentic systems run with confidence across a real workflow: clear stages, tight feedback loops, strong guardrails, and enough context to keep moving with low touch. That matters in code. It matters just as much in product, planning, security, testing, and non-functional requirements.
The interesting work is not one agent doing one clever thing. It is shaping a sequence that can keep progressing, recover when it needs to, and generate outcomes the enterprise can trust.
What I’m focused on
Right now, the center of gravity is agentic harnesses for software and product development.
A lot of energy is flowing into code generation. That makes sense. I’m more interested in the larger cycle around it: defining the work, breaking it into stages, tightening the loop between steps, and reducing the amount of human coordination required to get from intent to completion.
That is where leverage starts to compound. A well-structured system can standardize good practice, reduce variance, and let teams spend more of their energy on judgment, direction, and context.
Where this goes
I think a meaningful share of the product and software lifecycle can run autonomously when the system is shaped well.
That starts with clear outcomes. Then sequencing. Then the right checks at the right stages. Then retry paths, escalation paths, and feedback loops that keep the work moving without drifting off course.
People still define the destination. People still shape context. People still matter at the edges where ambiguity, judgment, and change show up. The system takes on more of the repeatable execution. That is the path toward higher-confidence delivery with less friction.
What keeps pulling me forward
I keep coming back to the same pattern: new capability becomes powerful when it is wrapped in the right structure.
That is why the harness matters so much.
The value is in directing the system well. Encode experience. Tighten the loop. Shape reusable foundations. Turn speed into outcomes. That is the part of the problem I find most interesting right now.
Who I like building with
I do my best work with people who are fun, creative, smart, and guided by good instincts.
People who know what good looks like. People who move with energy. People who care about the craft and like figuring things out together. The kind of team that can stay with a problem, sharpen the idea, and keep going until the thing feels right.
That mix matters to me as much as the domain.
What I tend to see early
I’m usually early on use cases.
I tend to see where a capability becomes useful before the packaging catches up. That showed up for me in GenAI with ideas like computer use and block-based orchestration before they were baked into the mainstream conversation.
I also spend a lot of time translating between people. In technical and business conversations, I can usually hear when two people mean more than they are saying, or when they are using the same words for different things. Clearing that up changes the work. Teams move faster once they are operating from the same reality.
My attention is on systems that take new capability and make it dependable.
Agentic harnesses. Better sequencing. Tighter loops. Stronger guardrails. More autonomous execution with good judgment built into the flow.
If you’re building in that direction, I’d love to talk.