DESIGN.md: The Markdown File Helping AI Generate Consistent UI Designs

·1 views

DESIGN.md is a lightweight Markdown-based document format that combines design tokens with written intent, enabling AI tools to generate UI components more consistently. Google Labs has described it as a structured way to persistently communicate design systems to AI agents, using YAML front matter for machine-readable values and plain text for human-readable rationale. Unlike full design systems such as Figma libraries, DESIGN.md acts as a quick-reference guide that AI can consult before starting any UI task, reducing drift across sessions. The format's official repository currently labels it as alpha-stage, meaning teams are encouraged to treat it as an evolving shared reference rather than a finalized industry standard. Multiple gallery sites, including designmd.app and designmd.sh, already host hundreds of community-created examples, and tools like Design MD Generator can help extract design information from existing webpages to auto-generate the file.

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)

How to Convert Any Photo Into a Cross Stitch Pattern Without Expensive Software

Converting a personal photo into a stitchable cross stitch chart involves three core steps: resizing the image to a grid, reducing its colors to a real DMC floss palette, and generating a symbol chart with a shopping list. The quality of conversion depends heavily on accurate color matching, ideally using perceptual color distance calculations against the full 454-shade DMC catalogue. Grid size determines the finished dimensions of the piece, with a 140-stitch-wide chart on 14-count Aida fabric producing roughly a 10-inch result. Beginners are advised to start with clear, evenly lit photos of a single subject to achieve the cleanest output. A common pitfall is allowing the converter to use too many colors, which can result in an unmanageable floss list and a fragmented, difficult-to-stitch pattern.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Builds AI-Powered Debugging Journal With Security-First Prompting on Cloud Run

A developer participating in Google's Cloud Run AI Challenge built DevLog, a structured debugging journal for developers, instead of the prescribed mood-reflection journal. The project centered on configuring Google AI Studio with a security-first directive that required the AI to produce a threat model — covering input surfaces, memory, tool execution, and inter-system communication — before generating any code. Each new feature request triggered a threat summary table mapping risks to countermeasures, acting as a binding contract against insecure code proposals. The app uses two separate Gemini API calls: one for free-form conversational rubber-ducking and a second dedicated call to extract structured, schema-constrained JSON data about the bug being discussed. This dual-call architecture was a deliberate design choice to prevent fragile regex-parsing of mixed conversational and structured outputs.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Roblox LiveOps Events Need Server-Side Rollback Plans Before Launch

Roblox LiveOps events can fail due to configuration errors — such as incorrect reward tables, overly generous promo codes, or wrong time zone settings — even without any code deployment. Developers are advised to follow a four-step incident response: contain, inspect, repair, then re-enable, rather than defaulting immediately to a hotfix. All event validation logic, including whether a promotion is active or what reward it grants, should live on the server rather than the client to allow consistent enforcement. Each event definition should carry a version identifier so that a specific version can be disabled during an incident without overwriting it, preserving data needed to assess player impact. A claim ledger keyed by player, event, and reward ID is also recommended to prevent duplicate grants during network retries or server restarts.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Built an AI Website Builder for Webround, Then Shelved It

A developer building the Webround headless website platform spent quieter summer weeks creating an AI-powered site builder that could generate entire websites from a single prompt. The tool split its workflow into a planning phase and an execution phase to keep context manageable and let users approve changes before the AI acted. However, the developer ultimately chose not to release it after identifying fundamental issues, including poor visual output quality from no-code component generation. A key concern was that Webround's public APIs and React environment already allow users to get working results by prompting mainstream tools like Claude or ChatGPT directly, making a proprietary integration hard to justify. With no clear advantage over free existing options and unresolved quality problems, the feature was put on hold indefinitely.

0 comments Read more at DEV Community