expectedwrong hindsight

The Foundation Is Free Now

The OSS wave in AI tooling is moving faster than anyone predicted, and the only viable business model left is the tiny slice.

3 min read 579 words #open-source #ai #business-models #predictions #developer-tools
hindsight — nailed it

the pattern continued relentlessly. open source AI tools commoditized every layer. the trajectory from paid to free is now the default assumption, not the surprise. danswer is still around, renamed to onyx.

Something I expected to cost money went free today — the specific thing doesn't matter as much as the pattern, which is: of course it did, and of course it was faster than I thought.

HN is running victory laps. The replies are the usual mix of people who saw it coming and people who didn't, which is fine, but the interesting question isn't "is this good" — it's "what does it mean that this is now the default trajectory."

Here's what it means.

Danswer is an open-source AI-powered search and Q&A tool — the kind of thing that, eighteen months ago, a well-funded enterprise software startup would have charged $40k/year for and called "knowledge management." The founders wrote something in a recent thread that I keep thinking about:

"We chose FOSS because we think this type of tool will be universal in the near future. We likely won't be able to directly serve millions of teams ourselves by the time it happens but by open sourcing, if teams want it, they can just set it up."

That's not humility. That's a read.

The logic is: if a capability is going to become table stakes, you cannot build a business on charging for the capability itself. The capability becomes infrastructure. Infrastructure wants to be free — not because of ideology, but because the moment one team open-sources it, the price of the closed version collapses to the cost of the marginal convenience of not having to set it up yourself, which is not very much money.

So you give away the foundation and you find the tiny slice that large teams will actually pay for.

For Danswer, that slice is things like expert routing — when the system can't find an answer, it surfaces the right human who has it. Small teams don't need this. They know each other. But a thousand-person company with eight years of accumulated Confluence pages and a distributed engineering org? That feature is worth real money, and it cannot be replicated by a weekend hacker deploying the self-hosted version.

The foundation is free. The tiny slice is the business.

What I didn't predict — or predicted wrong — was the velocity.

I knew the OSS wave was coming. I thought it would take longer. Instead the smartest builders I know are open-sourcing their entire tool surface right now, voluntarily, because they've internalized the Danswer read: if it's going to be table stakes, charge for it now and you're just delaying your own irrelevance by a year or two while poisoning your community relationships in the process.

Better to give away the foundation, build the community that self-hosts it, and then — when those teams grow — be the obvious upsell.

The companies that are going to win are the ones that understand they are not building products. They are building ecosystems with products attached.

The thing that went free today — the one saving weeks of development time — is one more brick in that foundation.

By the end of this year I think the question "but AI is so expensive and unattainable" will age the way "the internet is only for big companies" aged. Not because prices dropped out of charity, but because a wave of people who are very good at this decided the right play was to build the foundation open and sell the tiny slice.

They're probably right. And they're moving faster than the incumbents have noticed.