Agents Building Agents - Alfonso Graziano, Nearform
www.youtube.com- Category
- AI Agents
- Type
- ARTICLE
- Added
- Jul 24, 2026
About
Building an AI agent for a real team is not a prompt problem, it is a systems problem. In this session we walk through a practical, production-minded workflow for building an agent using a coding agent, and designing the codebase so that this loop stays reliable as complexity grows. The core pattern is two agents with different jobs. The coding agent is the builder: it writes and changes the agent’s codebase. The agent you are building is the product agent. It is the custom agent you ship for a
What it can do
Build and modify a product agent's codebase using a coding agent
Natural language requirements or development instructions → Written or updated agent codebase
Run an eval suite that exercises the product agent across representative tasks
Product agent and eval test suite → Eval results with pass/fail status and failure artifacts
Inspect failure artifacts to diagnose eval failures
Failed eval artifacts → Diagnosis identifying the responsible layer (context, tool contract, or code)
Propose targeted fixes to the correct layer of the codebase
Eval failure diagnosis → A targeted code/context/tool-contract fix
Open a pull request with a change report
Proposed fix and eval context → PR containing the fix plus a short report of what changed and what is still missing
Escalate unresolvable failures to a human
A failure the agent cannot safely resolve → A request for specific human input with an explanation of the blocker
Why it made the leaderboard
If you're shipping custom agents for teams, this lays out a practical loop for keeping them reliable as complexity grows — using a builder agent to run evals, diagnose failures across context/tool/code layers, and open PRs, while escalating to humans only when it can't safely resolve a failure.
Media
Comments (0)
No comments yet
Indexed by a proprietary survey. Corrections welcome.