3 ms·
wow what a candid, and humble take! Really impressive. Lots to be learned here. However, one thing I am curious about is how their VC investor, SignalFire, is
by data_ders 2y ago
wow what a candid, and humble take! Really impressive. Lots to be learned here.
However, one thing I am curious about is how their VC investor, SignalFire, is ok with this "back to it's roots" terminus. Do they still own the same amount of the company as before?
- ayhanfuat 2y agoThere is some info on Honnibal's Twitter (https://x.com/honnibal/status/1813650728222880157 https://x.com/honnibal/status/1813650728222880157): Delip Rao: Curious how do you go back from VC funded to bootstrapped? Did you return money to investors? And were they okay with just that? Regardless, I am happy you are doing more lib dev. You have very good taste. Matthew Honnibal: There's no change in ownership, it's just that we don't have any more VC money to spend --- so the operating reality is like it was before. We think it's helpful to explain this. Delip Rao: So if you consult and do other stuff as Explosion to bring money in, you have to pay distribution of the profits to the shareholders on your cap table? Matthew Honnibal: If we ever pay a dividend, all shareholders will get some, (and the VCs have liquidation preferences that mean they'd take a larger share than their ownership until the money is paid back). But we're some distance away from dividends -- we'd pay ourselves a salary from consulting Delip Rao: That's right. I am surprised your VCs were okay with this arrangement. Usually, they nudge the company towards an acquihire in such situations.
- teruakohatu 2y agoThe title is a little misleading. They raised $6 million from SignalFire (who also invested in Grammarly) for 5% of the company [1]. As SignalFire retains that shareholding, they are not less independent (nor more) than that have been in the past couple of years. They are just not looking for more investment and so are more "independent-minded". I used SpaCy quite expensively pre-2020. It still has a lot of great uses, and more predictable than LLM models. But now many NLP tasks that seemed near impossible, or required a lot of expensive annotated training data (which their product Prodigy is used for) can now be coded in very little time with LLMs. [1] https://explosion.ai/blog/weve-sold-5-percent-of-explosion https://explosion.ai/blog/weve-sold-5-percent-of-explosion
- codelion 2y agoYeah I was hoping there will be some bit about how LLMs make many of the tasks that people used SpaCy for a lot easier and cheaper. That is a bigger threat to the project going forward I don’t see how it can exist in the same form as today.