What Work Should We Actually Give Coding Agents?
As coding agents become increasingly capable, the important question isn't simply whether an agent can implement something. It's whether we should give the agent the autonomy to do it.
SOFTWARE, SYSTEMS & THE WORK AHEAD
Engineering leader exploring how we build and secure software in an agentic-first SDLC.
I've spent my career building software systems, engineering organizations, and technology products. My current work explores agentic software development, secure software engineering, and how AI changes the architecture and practices of the SDLC.
01
Currently exploring
01 /
Agentic-first SDLC
How development changes when agents become active participants in planning, implementation, testing, and delivery.
02 /
Harness engineering
The infrastructure, constraints, and feedback systems that make coding agents reliable.
03 /
Context engineering
Designing the information, tools, and environment agents need to do useful engineering work.
04 /
Secure agentic development
Security architecture for systems that can modify code, invoke tools, and make engineering decisions.
02
All writing ↗Field notes
AN ONGOING SERIES / BUILDING AN AGENTIC-FIRST SDLC
03
All projects ↗Work in progress
Agentic Software Development Pipeline ↗
An internal software-development system in which users can request product changes conversationally and a coding agent implements the requested changes.
Explore the project →- 01Request
- 02Agent
- 03Pull request
- 04Human review
AUTONOMOUS IMPLEMENTATION → HUMAN JUDGMENT
BUILT ON EXPERIENCE. DRIVEN BY QUESTIONS.A little more about me ↗