Skip to main content

Current Work & Thinking

Investigating agent reasoning, orchestration, knowledge and context, and adaptive decision-making—by testing assumptions and examining why an approach succeeds or fails.

Questions worth investigating


Coordinated engineering agents. I’m exploring how engineering work can be decomposed, planned, executed, tested, and evaluated by coordinated intelligent systems. The practical questions include how a system preserves task state, selects useful context, coordinates tools and workers, and determines whether a result meets its objective. For an engineering team, the relevant decision is not simply whether an agent can produce code; it is what can be verified, where human approval belongs, and whether the workflow improves the work reliably.

Knowledge, context, and model state. I’m experimenting with semantic knowledge organization and techniques for reusing compiled context within state-space and hybrid language models. Early investigations into model-state transformations and selective context injection are promising, but remain research rather than a commercially validated product. I share methods, evidence, and limitations selectively when relevant to an engagement or collaboration.

From hypothesis to useful decision. Agent reasoning and orchestration, knowledge and context management, and adaptive decision policies are areas of inquiry, not a list of proven client outcomes. A useful assessment identifies the assumption at issue, defines what an experiment should reveal, and examines the result—including failures—before recommending a larger investment. Some questions may be better answered with a simpler approach.

Explore how I help

Working through a consequential AI question? See how focused architecture and hands-on engineering support can help test the assumptions before committing to a larger build. Research directions are exploratory; scope and success criteria should be agreed for each engagement.