SShortSingh.
Back to feed

How to Write DESIGN.md Files That AI Agents Can Actually Follow

0
·1 views

A structured DESIGN.md file helps AI agents apply design systems correctly by explaining intent and rules rather than just listing token values. For example, instead of stating a color hex code, effective prose explains that a color is reserved for the single most important action on screen. Large language models parse markdown with high fidelity, making well-written prose an efficient channel for communicating design rationale. Key sections include an overview, color roles, typography, layout, and a Do's and Don'ts list that sets hard guardrails against common mistakes. The core principle is that tokens tell an agent what a value is, but only prose tells it how and when to use that value.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Developer builds cross-distro Rust package manager to solve Linux From Scratch gap

A developer building a custom Linux distribution from scratch using Linux From Scratch (LFS) found there was no package manager available for such systems. To solve this, they created Chiral, a cross-distro binary package manager written in Rust that works on any Linux system, including LFS, Arch, and Debian. Chiral uses a three-way fallback chain — a personal GitHub repo, Debian stable repos, and Arch Linux repos — to locate and install packages. It features full dependency resolution using breadth-first search and topological sorting, file tracking for clean removal, and a self-update mechanism via GitHub releases. Distributed as a fully static binary with no dependencies, Chiral can run even on a minimal LFS system with nothing pre-installed.

0
ProgrammingDEV Community ·

How PostgreSQL Uses WAL and Shared Buffers to Deliver Near-Instant Write Speeds

PostgreSQL achieves fast write performance by avoiding slow random disk writes at the moment a query is executed. Instead, it first modifies data in a RAM-based cache called Shared Buffers, marking changed pages as 'dirty' until they can be persisted. To prevent data loss if the system crashes, Postgres records every transaction in a Write-Ahead Log (WAL) — an append-only file on disk — before confirming success to the application. Because appending to the WAL is sequential and far faster than random disk writes, durability is guaranteed without sacrificing speed. A background process called the Checkpointer periodically flushes dirty pages from RAM to permanent storage files, completing the full write cycle.

0
ProgrammingDEV Community ·

Flutter_skin lets developers update app themes remotely without App Store resubmission

A developer has built flutter_skin, an open-source runtime skin engine for Flutter apps that allows UI color themes to be updated remotely without requiring a new app release. The tool replaces Flutter's compile-time theming system with named color tokens managed via a web dashboard at app.fskin.dev. When a new skin is published, changes are pushed to all connected devices in approximately one to two seconds using a server-sent events connection backed by Supabase Realtime and a Node.js backend. Currently in alpha, the package supports full Material ColorScheme color tokens and offers team collaboration, version history, and API key management through its dashboard. The package is available on pub.dev and the source code is hosted on GitHub.

0
ProgrammingDEV Community ·

NVIDIA Releases LocateAnything-3B Vision Model for Open-Ended Object Localization

NVIDIA has released LocateAnything-3B, a vision-language model designed to locate objects in images based on open-ended natural language queries rather than predefined categories. Unlike traditional detectors such as YOLO, the model returns precise bounding boxes for objects described in conversational prompts, including complex attributes and spatial relationships. The model gained attention through a demo in which it successfully identified densely packed, heavily overlapping objects individually, highlighting its spatial reasoning capabilities. LocateAnything-3B combines a language backbone, a vision encoder, and spatial reasoning components to interpret both what a user is looking for and where matching objects appear. NVIDIA positions the model as particularly relevant for developers working on AI agents, robotics, autonomous systems, and document intelligence applications.

How to Write DESIGN.md Files That AI Agents Can Actually Follow · ShortSingh