blog3 posts20 topicsfiltered: ai

Notes from the field.

Working notes on the craft — engineering at scale, the AI-era developer experience, frontend that most teams leave on the table, and whatever else has my attention this week.

Building grant-pilot: a multi-turn agent that orchestrates three sub-agents over real federal data
Third in the trilogy. fedbench measured agents. fieldops-mcp shaped what they could do. grant-pilot is what happens when an agent actually has to run a workflow end-to-end — discovery, eligibility, drafter, all dispatched by a planner over public grants.gov + SAM.gov data. Here's what I built and what shipping it taught me.
ai
agents
sub-agents
orchestration
fde
open-source
May 10, 202610 min
Building fedbench: an LLM eval harness for grounded Q&A
Most LLM agents that read documents and answer questions fail in three specific ways. fedbench measures all three. Here's what it does, why I built it, and what I learned shipping it.
ai
evals
rag
open-source
May 9, 202614 min
Building fieldops-mcp: designing the tool surface an agent actually uses
I built a small MCP server that turns a fictional dispatcher's workflow into agent tools. The interesting part wasn't the code — it was the design choices that decide what the agent can do at all. Here's what I built and what shipping it taught me.
ai
agents
mcp
tool-design
open-source
May 9, 20267 min

Building grant-pilot: a multi-turn agent that orchestrates three sub-agents over real federal data

Building fedbench: an LLM eval harness for grounded Q&A

Building fieldops-mcp: designing the tool surface an agent actually uses