Goods and Bads of Being Open Source AI Company

machine-learning, startup, open-source

Having dedicated the last half-dozen years to an open-source AI, here are my insights on the pros and cons of operating as an open-source organization.

This article is a reflection of my experience. I welcome you to participate in the conversation, regardless of your agreement or disagreement with any points. Feel free to reach out at any time.

Goods #

Easier to reach your users

Cool models and projects attract a ton of attention easily. It helps you find your early users, get user feedback fast, and improve your models & product. Nonetheless, in this feedback loop, it’s crucial to separate the meaningful signals from the irrelevant noise and accurately prioritize to align the users’ requests with the business requirements.

Community is everything and the greatest asset.

Actively listen to them and try to address their pain points. It helps you a lot in finding your product market fit. Honestly, we did a terrible job here initially and realized how important it is afterward.

Your community is there to help with things that are impossible otherwise. Getting things done with the community is the best way to make sure that you are on the right track.

Contributions of contributors

Contributions are great in especially the early stages. They help a lot with fixing bugs, you are oblivious to, adding features you never thought of, and helping with the design & structure of the code.

Open source is a great hiring engine.

Contributors create a great talent pool for hiring. And since they already know the code base, it is really easy to onboard them.

Github stars ~= VC interest

Open-source repos with many stars attract investors. It is especially helpful at the early stages. Later on, it is not enough.

A successful open-source project does its marketing and PR on social media

Open source is highly appreciated by the social media. The release of a good model receives an overwhelming response. While it tends to gain extensive recognition overnight, the fast-paced nature of AI means that the hype is usually short-lived, giving way to new models the very next day.

It is easier to see the impact of your work with a healthy dopamine loop.

Open-source enables people to create and develop crazy things that help other people create crazier things and so on. Seeing this would give you a dopamine rush and satisfaction that is harder to achieve otherwise. That’s probably why most open-source devs cannot give up on open-source going forward.

Open source introduces you to great people

You meet a ton of unique people it is impossible to meet in different settings! This is probably the best part.

Bads #

Making revenue is hard.

Revenue is what makes a project a company. But it is hard. There is yet to be found the right business model for open-source AI companies, especially for the model companies.

It is harder to gateway your models.

It creates two main problems. First, it is harder to make users pay even with a non-commercial license. Many don’t care about it, deploy your models and like your tweets.

Second, people use your models in different use cases and you have no control. You try to inform them with ToS or license but you know…

Making changes is hard.

When you need to change something, you have a community to convince and the backlash can be rough.

I think it is necessary to be transparent and keep your community in the loop when you need to do that.

Haters hate

Some people hate everything. And because you are open-source, you are more susceptible to them. Don’t care about everything people say.

Managing a community is a challenging task.

An open-source company should spend enough resources for community management, and advocacy. Until you have such resources, it takes a significant amount of well-deserved time.

Engineering overhead in the long run

There are the needs for your business and the issues & feature requests from your open-source users, but even in the best scenarios, these tasks bring additional overhead. Therefore, it is advised to anticipate this and plan accordingly.

This is actually why we commonly see some mature open-source companies stop accepting PR requests.

That’s all for this one. I think I’ll follow up with another post soon expanding on the business side and addressing what the current challenges are, how today’s companies try to address those, and what should be done in my opinion. Until then 👋