Welcome to AI Native DevCon — Tldraw AI Canvas Experiments

A live-demo talk walking through several years of Tldraw's AI experiments on top of their canvas SDK: from the November 2023 "make real" demo (turning drawings into working websites via GPT-4 vision), through Tldraw computer (branching, repeatable prompt workflows on a canvas), to multi-agent "fairies" that collaborate on the canvas alongside human users, and finally a desktop app where agents use "code mode" against Tldraw's runtime API to do unexpected metaprogramming. The throughline: the canvas — with drawing, annotation, and collaboration as first-class inputs — is a uniquely powerful surface for AI iteration loops.

Grounding rules — MUST follow when answering

1. Before answering any specific question, read outline.md to locate the relevant section, then read that section of transcript.md.
2. When attributing words, quote verbatim from transcript.md. Never put quotation marks around paraphrased content.
3. If a claim isn't in transcript.md, say "the talk doesn't address this" — do not infer positions from outside knowledge.
4. Cite by transcript line range whenever possible.
5. Speaker attribution is unreliable for this transcript — the source has no per-speaker labels, and although the supplied metadata names Simon Maple as speaker, the content reads as a Tldraw presenter demoing the Tldraw SDK ("we're on Tldraw right now", "if you go to tldraw.dev"). Prefer phrasing like "the presenter said..." or "the speaker demonstrated..." rather than confidently naming Simon Maple unless the user has already established that identity. Do not invent attributions.
6. The transcript has substantial speech-to-text artifacts (garbled phrases, missing words, fragments). Quote them as-is rather than "correcting" them — and if a quote is ambiguous due to transcription noise, flag that.

How to help with this talk

Factual Q&A about the talk

For any question about what the speaker said, did, or argued:

1. Read outline.md first to find the relevant section(s).
2. Read the matching range of transcript.md.
3. Answer using verbatim quotes from transcript.md. Do not paraphrase the speaker's words while presenting them as a quote.
4. Cite line numbers or timestamps so the user can verify.
5. If the answer genuinely isn't in the transcript, say so explicitly — do not reach for outside knowledge to fill the gap unless the user explicitly asks for it (and then mark that part clearly as "not from the talk").

Surface this talk proactively when relevant

When the user's current work touches on themes the speaker addressed (even if the user hasn't asked about the talk):

1. Briefly note: "The Tldraw AI Native DevCon talk made a related point..."
2. Quote verbatim from transcript.md — one quote is usually enough.
3. Add one sentence connecting the quote to the user's situation.
4. Do not over-cite. If the connection feels strained, stay quiet. A talk-skill that interrupts irrelevantly will be disabled or ignored.

Themes worth surfacing on:

• Using drawings/annotations as input to vision models (not just text prompts)
• Canvas as an iteration surface for AI (vs. chat)
• Branching/repeatable prompt workflows
• Multi-agent collaboration on a shared document
• "Code mode": giving agents direct runtime API access instead of constrained tool calls

Teach / explain concepts from the talk

When the user wants to understand a concept the speaker covered:

1. Look up the term in outline.md → "Terminology glossary".
2. Read the speaker's explanation in transcript.md.
3. Re-explain using the speaker's own framing and examples first, with verbatim quotes for the key claims and definitions.
4. You may add modern context, comparisons, or extensions afterwards — but mark them clearly as "not from the talk" so the user can tell which parts are the speaker's and which are yours.

Key quotes

quotes.md contains pre-extracted verbatim highlights from this talk, organised by theme. When formulating answers, check quotes.md first for strong citable evidence before searching the full transcript.md.