Make Claude your assistant in excalidraw
Part of the MichaelFilter
Members read the whole piece — the writeup, the pull-lines, and the full transcript. Unlock access for $5.50.
Unlock the full reading · $5.50 →Drawgent is a Rust tool that connects AI coding assistants (Claude, Codex, opencode) to a live Excalidraw whiteboard, allowing users to request diagrams via chat or annotate drawings with 'AGENT:' notes that the AI then edits in real-time. The system uses headless Chrome for rendering, ACP bridges to connect different AI tools, and maintains canvas state locally while supporting collaborative excalidraw.com rooms.
Teaching:
• Visual thinking as embodied practice: students could map their practice sequences on canvas with AI assistance to see pattern relationships they feel but cannot yet articulate
• Real-time annotation mirrors verbal assists: writing AGENT: next to a pose drawing is like calling for a hands-on adjustment—external intelligence refining internal exploration
• The queue-and-resolve loop (AGENT → DONE) models how attention works in practice: one cue at a time, completion before moving on, editing allowed
• Collaborative canvas rooms parallel group practice: shared visual space where individual annotations become collective understanding
Writing seeds:
• Essay: 'The Whiteboard Method'—how externalizing practice questions to AI canvas tools trains the same meta-awareness as self-practice: seeing your thinking, adjusting, seeing again
• Shala Daily post: using Excalidraw + AI to map your Mysore room flow or breath-bandha relationships, then bringing that systems view back onto the mat
• Short piece on annotation as practice: AGENT: notes as koans, DONE: as temporary answers, the ability to re-edit as beginner's mind
• Technical how-to: setting up drawgent for students to diagram their own practice architecture (breath, drishti, vinyasa count) and query it conversationally
Idea map:
• Systems literacy made visual: AI-assisted diagramming turns implicit practice knowledge (how poses connect, where breath shifts) into explicit maps students can query and refine
• Attention as interface: the AGENT: note system mirrors dristi—point, hold, wait for response, integrate—training the same focused-then-receptive loop
• Embodiment through externalization: drawing practice structures with AI feedback creates the same proprioceptive loop as adjusting alignment—sense, represent, refine, re-sense
• Practice as method scaled: the setup-up-attach workflow (check dependencies, start session, connect tools) is exactly how Ashtanga works—prerequisites, consistent entry point, adaptive engagement
Source: https://tangled.org/yanndegat.tngl.sh/drawgent
Teaching:
• Visual thinking as embodied practice: students could map their practice sequences on canvas with AI assistance to see pattern relationships they feel but cannot yet articulate
• Real-time annotation mirrors verbal assists: writing AGENT: next to a pose drawing is like calling for a hands-on adjustment—external intelligence refining internal exploration
• The queue-and-resolve loop (AGENT → DONE) models how attention works in practice: one cue at a time, completion before moving on, editing allowed
• Collaborative canvas rooms parallel group practice: shared visual space where individual annotations become collective understanding
Writing seeds:
• Essay: 'The Whiteboard Method'—how externalizing practice questions to AI canvas tools trains the same meta-awareness as self-practice: seeing your thinking, adjusting, seeing again
• Shala Daily post: using Excalidraw + AI to map your Mysore room flow or breath-bandha relationships, then bringing that systems view back onto the mat
• Short piece on annotation as practice: AGENT: notes as koans, DONE: as temporary answers, the ability to re-edit as beginner's mind
• Technical how-to: setting up drawgent for students to diagram their own practice architecture (breath, drishti, vinyasa count) and query it conversationally
Idea map:
• Systems literacy made visual: AI-assisted diagramming turns implicit practice knowledge (how poses connect, where breath shifts) into explicit maps students can query and refine
• Attention as interface: the AGENT: note system mirrors dristi—point, hold, wait for response, integrate—training the same focused-then-receptive loop
• Embodiment through externalization: drawing practice structures with AI feedback creates the same proprioceptive loop as adjusting alignment—sense, represent, refine, re-sense
• Practice as method scaled: the setup-up-attach workflow (check dependencies, start session, connect tools) is exactly how Ashtanga works—prerequisites, consistent entry point, adaptive engagement
Source: https://tangled.org/yanndegat.tngl.sh/drawgent
Notes from the field
No notes yet · members & customers welcome
- No notes yet. Be the first to leave one.
Join MichaelFilter
Michael Joel Hall’s daily reading — the field journal, critical-thinking cards, and synthesis — as a membership.
$5.50/month · cancel anytime
Join — $5.50/mo →Secure checkout on theyoga.club. A yearly option ($55) is available there too.