tldraw

a sourced history of tldraw, steve ruiz's infinite-canvas company: the sdk business, the make-real demos, and the canvas-as-infrastructure thesis.

tldraw started as steve ruiz's sponsorware drawing app — 'a tiny little drawing app' opened to everyone in november 2021 — and became something rarer: a canvas company that found its business underneath the canvas. the sdk sold to clickup, replit, and google while the demos — make real, draw fast, tldraw computer, fairies — kept breaking the internet a little. then the license changed, twice, and the open-source community noticed. this is the record of that arc — the sponsorware, the viral fortnight, the license keys, and the thesis the whole thing argues: canvas-as-infrastructure survived where canvas-as-an-app struggled.

• history: the company, sourced and dated — the drawing-app experiments, the sponsorware launch, the viral fortnight, and the license-key era.
  • the drawing-app experiments: 2020–2021 — the perfect-* libraries, globs.design, the sponsorware gate, and the november 2021 public launch.
  • the company forms: 2022–2023 — incorporation, the first hires, the $2.7m lux-led seed, and the v2 rewrite.
  • the viral fortnight: late 2023 — gpt-4v, draw-a-ui, make real, the $2m extension, draw fast, and the license change inside five weeks.
  • the sdk business: 2024–2025 — sdk 2.0 stable, the watermark compromise, sync, tldraw computer, and the $10m series a.
  • the license-key era: 2025– — sdk 4.0's license keys, fairies, the ai-slop firewall, and offline and flash on the desktop.
• people: steve ruiz and the people around tldraw — who they are, what they wrote, and what they believe.
  • the founder: steve ruiz — the painter turned design engineer who turned down adobe and bet the canvas could be infrastructure.
  • the writing: the steveruiz.me essays, the tldraw blog, and the posts where the license and the business got explained.
  • the team: sheldrick, wilson, the seed-era bench, the later additions, the ~20-person roster, and the backers.
  • interviews and talks: the podcasts and talks — metamuse, latent space, gitnation, ai engineer — where the philosophy got said out loud.
  • the philosophy: taste-driven engineering, demos as r&d, open source as spectrum — and the controversies attached to them.
• side quests: the pursuits around the company — ruiz's pre-tldraw libraries, the demos and experiments, and the community projects the canvas attracted.
  • ruiz's experiments: state-designer, perfect-arrows, perfect-freehand, globs.design — the pre-tldraw run.
  • the company's quests: the v2 rewrite, signia, make real, draw fast, carry on, teach, sync, computer, the starter kits, fairies, offline, flash.
  • the ecosystem: draw-a-ui, the macpaint skin, orion reed's canvas hci work, and the community agent demos.
• coverage: what the discourse made of tldraw — the launch threads, the demo virality, the license fights, and the sentiment eras.
  • the launch: the sponsorware post, the 471-point launch, and the macpaint skin — community-led and craft-focused.
  • the funding: the seed, the $2m extension, the series a that got 3 points, and the aggregator tables that disagree.
  • the ai demos: make real, draw fast, tldraw computer, fairies — the biggest coverage moments in the record.
  • the licensing: the december 2023 change, the watermark compromise, the sdk 4.0 keys, and the oss comparisons that turned.
  • the comparisons: excalidraw, miro, muse, kosmik, and the agent-canvas category forming around the company.
  • the discourse: the ai-slop firewall, the tests-to-closed-source joke, and the discourse the license made possible.
  • controversies: the running debates — the license change, the keys, the spectrum framing, the demo sunsets.
  • sentiment eras: sentiment phases: craft darling → credible challenger → viral fortnight → demo age → gated era.
• the idea: canvas-as-infrastructure — the lineage tldraw inherited, what it proved, and what agents do to it.
  • lineage: sketchpad, dynabook, hypercard, spatial software, muse, excalidraw — the ancestors the canvas inherits.
  • what tldraw proved: what the infrastructure bet actually demonstrated — the dom canvas, the demos, the sdk roster.
  • what stayed unfinished: the showroom app, the spent goodwill, the sunset pattern, and the pricing that stays opaque.
  • the agent-era turn: the canvas as an agent surface — the case for, the case against, and the commoditizing paradigm.
  • successors: excalidraw, draw-a-ui, muse's ghost, the canvas cohort, and the lab canvases.
  • open questions: the questions a next canvas company has to answer.
• sources: every source cited in the tldraw history, with its date and type.
  • the catalog: the full typed catalog — primary, interview, reporting, archive, community, repo, reference.
• method: how ai agents researched the tldraw history, the evidence rules they followed, and what they could not verify.
  • the lanes: the two parallel research lanes and what the formula proved.
  • the evidence rules: the evidence model — confidence, precision, sentiment discipline.
  • known gaps: what the research could not verify — recorded, not hidden.
  • corrections: how to correct the record.

AI-drafted at Ben Guo's direct request and credited to Hraness; every claim links to its cataloged source.

Canonical page: https://hraness.com/tldraw