Coding Agents on Tech Foundations

Agents Md

Mon, 01 Jan 0001 00:00:00 +0000

Operating Model

A practical mental model is to treat a coding agent as a capable colleague who starts each session with strong general knowledge but limited project-specific context.

Ca Dict

Mon, 01 Jan 0001 00:00:00 +0000

internal harness

Comprehension

Mon, 01 Jan 0001 00:00:00 +0000

Project understanding

The engineer must maintain a working mental model of the system. Coding agents can generate, modify, and explain code quickly, but they do not remove the need for human comprehension. In fact, they increase the need for it, because implementation velocity can exceed the engineer’s ability to evaluate whether a change is coherent.

Harness Engineering 1

Mon, 01 Jan 0001 00:00:00 +0000

Idiosyncrasy

Mon, 01 Jan 0001 00:00:00 +0000

Guide: Working with Idiosyncrasy in Agent-Ready Target Systems

A target system intended for coding agents should be evaluated not only by its runtime behavior, but also by its architectural legibility. In this context, architectural legibility means that a competent newcomer – human or agent – can form reasonable expectations about where functionality is located, how changes should be made, and how those changes can be validated.

In Effective Context Engineering

Mon, 01 Jan 0001 00:00:00 +0000

After a few years of prompt engineering being the focus of attention in applied AI, a new term has come to prominence: context engineering. Building with language models is becoming less about finding the right words and phrases for your prompts, and more about answering the broader question of “what configuration of context is most likely to generate our model’s desired behavior?”

In Harnes Building

Mon, 01 Jan 0001 00:00:00 +0000

Harness-Building for Coding Agents

Executive summary

Recent writing by Martin Fowler, LangChain, OpenAI, GitHub, Anthropic, and open-source coding-agent projects converges on the same broad intuition: a coding agent is not just a model. It is a model embedded in a system of prompts, tools, files, plans, approvals, execution environments, and feedback loops.¹²³

Part 0

Mon, 01 Jan 0001 00:00:00 +0000

Part 0: What I understand about working with coding agents after ~60B token events in 9 months

Over the last 9 months, my agentic engineering workflow has consumed roughly:

Perception And Context Management

Mon, 01 Jan 0001 00:00:00 +0000

Perception and Context Management in Coding Agents

Executive summary

In the language of Martin Fowler and recent writing from LangChain, a harness is often treated as “everything around the model”: state, tool execution, feedback loops, constraints, memory, and orchestration.¹² That is operationally useful, but analytically too coarse for coding agents. Classical AI kept several of these functions separate: sensing and acting in a task environment, maintaining an estimated state under partial observability, controlling what remains in working memory, and closing the loop with monitoring and replanning.³⁴⁵

Project Ownership

Mon, 01 Jan 0001 00:00:00 +0000

Project understanding

We need to discuss a basic question of project understanding and manageability. This may sound obvious, but it becomes easy to forget when working with LLM-based coding agents.