Plano Prompt Engineers (ChatGPT, OpenAI, Langchain)

🏗️ Building Your First Engineering Harness

AI coding tools are getting really good, but most workflows today still assume an engineer is sitting locally, setting up the repo, giving the agent instructions, reviewing what it does, and moving the change forward.

What if we moved that workflow into a remote engineering harness?

In this session, we'll explore how engineers may start spending less time implementing every change directly and more time designing the systems, constraints, and feedback loops that allow agents and other contributors to safely turn intent into software.

This kicks off a three-part series on Harness Engineering with Bharadwaj Pendyala:

Part 1: From Intent to Pull Request — Building a Remote Engineering Harness

Part 2: Making the Harness Reliable — Context, guardrails, evals, and verification

Part 3: From Harnesses to Software Factories — Scaling across workflows and teams

We'll start by building the remote harness foundation and demonstrating the core workflow end to end with a real application:

• Turn a natural-language feature request into a concrete implementation spec
• Spin up an isolated environment with the right repository and context
• Use the harness to implement and validate the change
• Produce a draft PR with evidence for human review
• Define where the harness ends and engineering judgment begins

Attendees will also get access to a starter repository and reference implementation so you can experiment with the same pattern in your own projects and build with us from here, with Parts 2 and 3 to follow at later dates.

Bharadwaj Pendyala is a Lead Member of Technical Staff at Salesforce with more than 12 years of experience building software, automation, and platform systems. His recent work focuses on AI-enabled engineering workflows, agent reliability, evals, and harness engineering, including practical approaches for moving agentic systems from experimentation toward reliable production use.