The Agentic AI Engineer
summary
Claims an Eval-Driven Developement Loop that automates spec → build → evaluate → diagnose → monitor → optimise for agents, replacing slow human review with a multi-agent team steered by an orchestrator. Demonstrates two research‑preview agents: an evaluator agent that generates eval suites and binary criteria, and a diagnostics agent that ingests traces (LangFuse or JSONL), clusters failure modes, emits an HTML report and suggested fixes, and produces markdown tasks or PRs for deployment; connectors, local/cloud runtimes and a future managed service complete the pipeline.
AI-written summary · about
chapters (17)
- 0:00Welcome to Agentic AI Engineer
- 1:04Manual loops slow agent dev
- 2:47Agentic loops scale better
- 3:10Offline loop spec to deploy
- 6:01Cold start vs optimizing agents
- 7:06Spec-driven agent dev
- 8:54Building agents from specs
- 11:21Eval-driven agent dev
- 13:52Evaluator agents essential
- 15:15Effective agent evaluations
- 19:17Diagnosing agent failures
- 22:26Autonomous optimization loops
- 23:28Full agent lifecycle
- 25:16Mutagent’s agentic platform
- 27:01Demo of diagnostics agent
- 31:43Diagnostics agents analyze traces
- 34:15Closing and product overview
related talks
46mAgentic Engineering: From Writing Code to Reviewing It
30mArchitecting Integrity: Responsible Software Evolution in the AI Era
45mAI agents in practice - beyond the prototype
02h 04mShip Real Agents: Hands-On Evals for Agentic Applications
28mAgentic AI
27mAgentic Engineering: Working With AI, Not Just Using It
25mWho Let the Bots Out? A Guide to Evaluating AI Agents
30mBuilding Tools with Tools: Modern AI-Driven Development