The AI startup drama that's damaging Y Combinator's reputation

This past weekend a Y Combinator startup sparked a huge controversy almost immediately after launching. The startup was PearAI, an open-source AI code editor. When people looked at its code they found that it was a clone of an existing open-source project called Continue.dev.

As with many controversies in startup land, it was less about the crime and more about the cover-up. The way the co-founders talked about their product and defended themselves is what really put them in hot water.

The controversy also grew to encompass Y Combinator itself, and now people are asking if the famous startup accelerator's best days are behind it.

Here's everything you need to know:

How NOT to launch a product

It all started with this launch thread on X from Duke Pan, one of Pear’s co-founders:

The thread went viral, but not in the way the founders wanted. People on X immediately slapped a community note on the original post explaining that Pear had gotten its code from a different open-source project:

“Pear is a fork of Continue.dev, an open-source AI code editor. PearAI used Continue.dev’s code and mass-replaced all references to ‘Continue’ to ‘PearAI’ to mislead people into believing that they built this product on their own.”

Turns out, the community note was inaccurate, and eventually got taken down. Pear had credited Continue in their GitHub README file and in a few of their YouTube videos prior to launching.

But this failed to quiet the outrage for a couple reasons:

First, the founders didn’t exactly make it easy to find their references to Continue. Pan himself went on to admit as much in a follow-up apology:

“What we screwed up, critically, was not being clear enough about this. We … wanted to launch as fast as possible. But doing so upon a fork of others’ work without many new features, and talking about it so publicly online, made it look like we were stealing the work of others as our own.”

But the second reason is more important. Whereas Continue was covered by the open-source Apache license, Pear created a custom closed license for their product. Which is a great way to piss off the open-source community.

Here’s why:

How NOT to piss off the open-source community

There are three things a startup founder should know before loudly announcing a new open-source project:

1. How Apache licenses work. (They let you freely use, modify, and distribute software as long as you give credit to the original creators and include the same license with any changes you make.)

2. How strongly the open-source community feels about values like transparency, collaboration, and proper attribution. (Very strongly!)

3. How intimately connected points one and two are!

Enter Pear, which not only changed the Apache license to a closed “Pear Enterprise License,” and not only used ChatGPT to create said license, but also defended these actions in the most “IDGAF” way imaginable:

Again, it’s not the “crime” (technically it’s not illegal to relicense an Apache license), but the cover-up.

Once criticism started to mount, Pear quietly capitulated by reverting the license to Apache.

Continue responds with class. The community responds with sass.

It didn't take long for Continue to jump into the fray with its response:

Classy! Though with a bit of a veiled threat in the line that "open source can't be taken for granted."

The rest of the tech world wasn't as polite. Just scroll through the replies to Pear's original launch post. Or check out this roast post from Jason Levin, the former head of growth at Product Hunt. Or feast your eyes on this masterpiece:

(Yes, someone actually built and and launched an app just to roast Pear.)

The reaction from the community seems to have emboldened Continue, because yesterday they upped the ante by issuing a much stronger statement than the one they started with.

This time they accused Pear of making "repeated attempts to misrepresent Continue" and even recommended a set of steps that Pear could take to restore its reputation:

“Our view is that PearAI should start over from scratch:

1. Make a true fork of the Continue repository rather than a copy

2. Retain the Apache 2.0 license at all times

3. Re-apply whatever changes they wish to make to Continue’s code in compliance with the license, making sure to correct any code attribution problems caused by PearAI’s initial conduct

Continue believes that anything less risks permanently tainting PearAI’s project and continuing to harm our community.”

Pretty extreme! An obvious PR nightmare for Pear.

And what's bad for business is bad for investors. So it was only a matter of time before Y Combinator's CEO, Garry Tan, entered the chat.

Y Combinator defends Pear, as well as itself

Just as the criticism was reaching a fever pitch, YC's CEO stepped in to defend Pear. Though tellingly, he left out the detail about Pear's initial decision to do away with the Apache license:

This didn't seem to do Pear any favors, but it was clear that Pear wasn't Tan's only concern. Because Y Combinator's investment strategy was also being called into question.

In the wake of all this drama, a blog post titled "Y Combinator Traded Prestige for Growth" went viral and hit the top of Hacker News. Which you might have missed, because Hacker News — which is owned by Y Combinator — seems to have manually dropped the post lower in the rankings to suppress its visibility.

In the post, the author argues that Y Combinator’s success as a startup accelerator came from its exclusivity, backing companies like Airbnb, Stripe, and Dropbox. But by accepting more and more companies recently, it's been losing the prestige that made it valuable.

For clear evidence of this decline, the post goes on to argues, we should look no further than the case of Pear, which:

“shows that (1) YC is willing to fund just about anything, (2) they’re not doing any real due diligence, and (3) they don't particularly care about their existing portfolio companies.”

Versions of this sentiment were echoed by people on X, and Tan's response shed some light on the inner workings of Y Combinator: