Lattice Open-Source Framework Promises to Fix AI Coding Chaos with Battle-Tested Engineering Disciplines

Breaking: Lattice Framework Launches to Tame AI-Assisted Programming

Rahul Garg today released Lattice, an open-source framework designed to eliminate the chaos introduced by AI coding assistants. The framework operationalizes patterns from his recent series on reducing friction in AI-assisted programming.

Lattice Open-Source Framework Promises to Fix AI Coding Chaos with Battle-Tested Engineering Disciplines — Source: martinfowler.com

Garg warns that current AI tools "jump straight to code, silently make design decisions, forget constraints mid-conversation, and produce output nobody reviewed against real engineering standards." Lattice addresses this through composable skills in three tiers—atoms, molecules, and refiners—that embed established engineering disciplines like Clean Architecture, Domain-Driven Design, and secure coding.

How Lattice Works

The framework includes a living context layer stored in a .lattice/ folder that accumulates a project's standards, decisions, and review insights. Garg explains that "after a few feature cycles, atoms aren't applying generic rules—they're applying your rules, informed by your history."

Lattice can be installed as a Claude Code plugin or downloaded for use with any AI tool. This flexibility makes it immediately usable across different development environments.

Related Developments: SPDD Q&A

In a parallel development, the article on Structured-Prompt-Driven Development (SPDD) by Wei Zhang and Jessie Jie Xia has generated massive traffic and numerous questions. The authors have now added a Q&A section to address a dozen of the most common queries.

Double Feedback Loops in AI Development

Jessica Kerr (Jessitron) shared a tool she built to work with conversation logs, highlighting the existence of two feedback loops. The first is the development loop where "Claude does what I ask and then me checking whether that is indeed what I want." The second is a meta-level loop: the "is this working?" check when she feels resistance.

Kerr notes that "frustration, tedium, annoyance—these feelings are a signal to me that maybe this work could be easier." This double loop changes not only the product but also the tools used to build it.

Background

The rise of AI coding assistants has brought unprecedented speed but also significant quality concerns. Studies show that generated code often contains hidden bugs, design flaws, and security vulnerabilities because AI models lack contextual understanding of the project's standards and history.

Traditional software engineering disciplines—like Clean Architecture, DDD, and design-first methodologies—have long been proven to produce maintainable, secure code. However, these practices are rarely integrated into the AI-assisted workflow, leading to a gap between generation and quality assurance.

Garg's earlier posts on reducing friction in AI-assisted programming attracted wide attention, prompting him to codify those patterns into Lattice. The framework is now available on GitHub under an open-source license.

What This Means

Lattice represents a shift from treating AI as a black box to embedding engineering rigor into the AI workflow. Developers can now enforce battle-tested practices without manually reviewing every line of AI-generated code.

The framework's living context layer ensures that project-specific standards evolve with the codebase, reducing the risk of constraint drift. As Garg puts it, the system gets smarter with use.

The emphasis on internal reprogramability—the ability to mold the development environment to fit the problem—harkens back to the days of Smalltalk and Lisp. Jessica Kerr captures this joy: "With AI making software change superfast, changing our program to make debugging easier pays off immediately. Also, this is fun!"

For teams using AI coding assistants, Lattice offers a structured path to maintain quality while accelerating development. The framework is immediately available for experimentation and should be of particular interest to organizations with strict code quality and security requirements.

Tags: