A growing practice of creativity, taste, and making.
AtomCraft is part of the AtomRearch matrix: a public collection of learning notes, reusable AI skills, and workflows for creative work. It connects humanities, art, design, app making, writing, and creative scientific practice.
The aim is to turn encounters with good work into better decisions and things worth making. Taste develops through close observation, explicit choices, practice, and revision. A useful method should preserve room for the creator's judgment.
| Material | Purpose | First entry |
|---|---|---|
| Learnings | Record sources, observations, interpretations, and reading depth | Source index |
| Skills | Give a person or AI actionable guidance for a particular task | Human-centered release notes · Reference-led design |
| Workflows | Connect learning, making, feedback, and revision | Learn to craft · Typography in context |
| Templates | Make the next contribution easier | Learning note · Skill proposal · Design decision |
- Humanities and expression: reading, interpretation, narrative, cultural context, and writing.
- Art and design: visual language, composition, typography, interaction, and aesthetic judgment.
- Apps and products: concepts, prototypes, interfaces, onboarding, and product communication.
- Taste: comparative critique, contextual judgment, tradeoffs, and learning from references.
- Creative research: question framing, conceptual models, explanation, and scientific storytelling, with evidence tracked separately from speculation.
- Workflows: practical combinations of tools, people, and AI across these activities.
These are directions for growth. The current library contains four source notes (including one directory-level reading map), two draft skills, and two workflows. The broader directions will grow through concrete tasks and documented feedback.
The source index distinguishes articles read from reading queues. It also links SkyWT's existing Chinese typography skill for upstream reuse; that external skill is not bundled or installed here.
Open a skill's SKILL.md and give it to your assistant together with the actual task and materials. For an assistant with repository access:
Read
skills/human-centered-release-notes/SKILL.md, then draft release notes from these verified changes: …
Read linked references only when needed. Repository skills are ordinary Markdown files; cloning this repository does not automatically install or activate them in an assistant. You can also follow their instructions yourself.
Start with a concrete work or problem. Record what you observed, identify a transferable choice, try it on an actual task, and keep the evidence that changes your judgment. The workflow explains this loop. Contributing describes how to add and revise entries.
Skills progress from draft to tried to refined through documented use. A convincing example alone is not evidence of real-world validation.
Original contributions are available under the MIT License. Linked articles, artwork, screenshots, quotations, and other third-party materials retain their own rights. Source notes provide attribution; a link does not relicense its target. The initial source note contains original commentary and a link, without reproducing the article or its images.