Language-model applications and tools for building software people can use.
I build model-specific tools with a concrete job: compare research, help draft accessible content, or examine an interface before someone has to struggle through it. This organization brings that work together.
| Project | What it does |
|---|---|
| Accessibility Devkit | Code-level accessibility tools, with optional language-model assistance and Model Context Protocol tools. |
| Intentional UX | Examines product task paths with accessibility built in. |
| Consensus | Compares research and arguments from multiple language models. |
| Universal Chat | A shared interface for chat and media across model providers. |
| Humanize | A writing plugin for clearer documentation and code comments. |
Give the model a specific task. Keep the person in control of what gets sent, changed, or published. Check the result against something outside the model's own answer.
Accessibility work needs both deterministic checks and human judgment. A generated suggestion is a draft; it does not establish compliance or replace testing with people who use assistive technology.
Start with the README in the relevant repository for installation, supported clients, and release details. Provider access and credentials depend on the tool you choose. Report reproducible bugs through that project's issue tracker, and check its license before reusing code.
Luke Steuber · Data Poems · Ambient Time · Actually Useful AI · One Impossible Thing
Made by Luke Steuber. Questions or collaboration: luke@lukesteuber.com.