Skip to main content

D2 is non-profit

· 4 min read

Hello, I wanted to share some news about the future of D2.

The creation and development of this project was funded by a for-profit company, Terrastruct. We made money through selling an IDE and a closed-source layout engine for D2.

Terrastruct is winding down, and D2 has been transferred to Hack Club to continue running as a non-profit, open-source project. The previously paid components, the layout engine (TALA) and the IDE (D2 Studio), will be open-sourced.

I will continue to maintain and improve D2. But my time commitment will be limited. If you have interest in becoming a maintainer, please email me at alex@d2lang.com. High-quality maintenance and contributions may be invited to enter into a paid contributor contract, funded by donations to the non-profit.

Direction​

D2 has accumulated a large backlog and roadmap, and I have some pretty cool ideas for features. But first, a disclaimer:

I am very pro-AI usage. My contributions towards D2's development will all be made with AI. I invite AI contributions, and will use AI to help me review your AI.

Up until now, D2 has been a project of no AI. Not by choice; it just wasn't good enough until recently. Pre-good-AI days, I embedded a JavaScript runner in D2's Go runtime to run some critical dependencies that were only available in JavaScript. It was massive bloat, it was ugly, and slow, and there was no other way. Post-good-AI, I just converted all of those JavaScript libraries to Go; it runs 9x to 50x faster.

In this manner, D2 will evolve at a rate that is orders of magnitude faster than it has. I of course know this is controversial and has downsides, and I'm not recommending anything to anyone; I just wanted to be transparent about my plans to the community and those who rely on D2. It's all open-source, so you may pin or fork v0.7.1, which is untainted by AI.

What's next​

Directionally, there's 2 things I'll prioritize:

  1. Get rendering support from popular platforms
    1. Natively integrated onto GitHub, Notion, Linear, etc.
  2. Be the most powerful harness for agents
    1. They're really good at making almost-perfect diagrams, but it's still too non-deterministic, uncustomizable, imprecise, etc. I think that'll last the foreseeable future, so you still want some text-based source to back the artifact. Kind of like HTML for web pages instead of drawing every pixel.

But I want to have fun with it along the way. Squeeze out every millisecond of performance. Add an isometric renderer. Rewrite some subset to make a native JavaScript version instead of having to pay the cost of Go WASM runtime.

Non-goals​

I want to avoid bloat. To be clear, just because AI will allow me to make a Swiss Army knife of a diagramming tool, I'd much rather spend the tokens on depth and quality of a small, curated feature set. I'll want to add some plugin/module system to easily add on things that don't belong in D2 core. For example, I still plan for D2 to focus on the software architecture diagram use case and won't be adding e.g. Venn diagrams, even if it'd be trivial to do so.

I still don't see a place for LLMs integrated into diagramming tools. I think models are smart enough to just look at some examples and then take your specs or code or whatever you want and turn it into a .d2 file. There doesn't need to be yet another thinly tuned LLM interface for D2 to provide.

Since D2 is non-profit, we won't be supporting things that need servers. No MCP, no API, no user accounts, no server rendering. D2 Studio will be purely client-side, usable completely offline.

More soon​

That's all for now. I'll be making posts as TALA and D2 Studio are released.

Thank you to everyone who has engaged with us over the years, and especially those who paid for the closed-source parts of it to fund its development. I'm glad D2 has resonated with so many people and I still stumble upon it in blogs and projects and mentions on the public internet and internal company channels. It's always a leap of faith to adopt some new language with unknown motives and continuity. I'm really happy about this direction for D2 and that the value of this excellent IDE and layout algorithm can be shared with more people. I'm excited for what's next.

If you'd like to donate to D2, please see this page on how to do so and what the fund is used for.

Contents