Thank you all for the thoughtful discussion and responses.
Is the Dash Framework in trouble?
It is not. In fact, quite the opposite. Dash’s use is growing thanks to coding agents (downloaded 59M times last month) and we’ve merged more PRs this YTD (209!) with more contributors (50!) than any other year in the last 5 years.
That being said, AI has had a huge impact on Q&A and community building. There’s been a serious decline on community forums across the board, including ours.
Plotly and Dash are at the forefront of a broad shift in analytics from point-and-click GUIs to code-based analytics built with AI. There’s a whole new class of users that can now build with code, and a whole new set of apps that are getting built because it’s cheaper than ever. Across our customer base, we’re seeing instances of customers creating 10-100x the number of applications and users.
But AI is also decimating search and online Q&A communities across the board. Users don’t ask each other questions, they just ask AI. More people are working on their own (with their AI agents) rather than collaborating and working with each other. In open source communities, users are re-inventing the wheel vibing and inlining utility libraries rather than publishing or importing shared utilities to package managers.
In the era of AI, we believe in strong frameworks. A vast amount of enterprise software and analytics will be rewritten over the next decade and a good framework gives you more intelligence / dollar than any model upgrade. A good framework abstracts away the bits that you don’t need to care about and exposes the bits that you do, and that’s what continues to drive us forward with Dash. Just this year, we’ve evolved Dash into an MCP framework, built a cloud service for community members to share apps (for free! filling the gap for the community that Heroku left), shipped our 275th release of plotly.js, shipped websocket support for Dash, enabled users to add chatbots to their app with a single click, shipped an agentic analytics product for creating Dash apps (Plotly Studio), and shipped a major overhaul to our enterprise platform to support 10x the number apps getting built in this new AI era.
We’re shipping more than ever because Plotly the company is in a great place. We’re unlike most other startups in that we don’t rely on precarious “raise and burn” funding cycles; we operate on the income from our customers and haven’t needed to raise money in several years. I really can’t emphasize enough how rare this is in the industry. And as open source stewards, we’re coming into our 10th year of maintaining Dash and our 14th year of Plotly.js. This longevity in support and funding is also rare among open source projects; so many projects are either maintained by a couple of solo developers who are stretched very thin or corporations where the library commodifies their complement (e.g. Kubernetes). And at the end of the day, all of our open source work is MIT-licensed, which is the most permissive open source license out there. We chose this license 13 years ago - very early on in Plotly’s journey - and we were very intentional about this decision. We wanted this software to be able to live on forever.
I mention both the open source offerings alongside our enterprise offerings because we value them equally. I want Dash and Plotly to be ubiquitous and available to anyone in the world with an internet connection. Before Plotly, technical data visualization was only available to users who could pay thousands of dollars for a MATLAB or SAS license and were hopelessly tied to the desktop, impossible to share. But to use this technology to be actually used in any major organization, it needs to be accompanied by a platform with enterprise functionality. We do both. And from a business perspective, we’ve designed the incentives of these product suites to be aligned. Here is an essay I wrote 6 years ago about how I think about what should be open source vs what should be enterprise, and it’s remained true since then: 📔 Dash Enterprise Fieldnotes #1 - Why Dash Enterprise?.
Our governance and development structure in open source has remained pretty constant over the last 13 years. We hold the roadmap with a loose grip, and try not to make too many public commitments and over promise. This is a tricky balance to find, and I recognize that it can come at the expense of predictability.
As we all know, AI is remarkable at writing Python, Plotly, and Dash code. The content from our documentation and this forum has likely been scraped by every model provider. In the early days of ChatGPT 3, I remember feeling so shocked to see the AI repeat back to me words that I originally wrote in our documentation. In today’s RLVR world, the models are likely getting better at building Dash and Plotly code by writing and training on examples in their own autonomous loops, independently of the forum. We didn’t have any explicit data sharing arrangements for this to happen, it just happened through their scraping; the forum isn’t a walled garden.
Despite all of this, community in the AI era will look different. After I released Dash in 2017, I made a point to read and answer every single question in the community forum. Many others started paying it forward and answering each other’s questions once they got help the first time. Q&A was the entry point to connection and this forum was the meeting place of Dash. This took several hours each day, and we eventually hired Adam to take over. He did an amazing job running this and recently moved abroad to new ventures. As we look for someone to fill his role, I’m stepping back into the community. I’m also working on ways for others at Plotly (the people that build the software) to step in and share more as well.
But now that AI has taken over Search and Q&A, we’re missing the entry point to this meeting place. We’ve been putting time into other channels: IRL events (I co-hosted an event in San Francisco for a few hundred people in the spring, and we’ve done a dozen smaller roadshows around the world this year), web events hosted by the engineers of the features (we run these about once a month), social media (although this always veers more promotional). Personally, I’ve started writing again (chris-parmer.com) and sharing some of that writing here in the forum.
Perhaps the forum will have its second life as a place for discourse, sharing, and connection rather than Q&A. This will be smaller, but it could be a nice reprieve from our days spent with our agents.