Do you think 3 is better than 1 & 2 as context gets larger? I think for smaller data sets its mostly fine no. It's an interesting bet by TM. End state does seem some form of continual learning (model weights update like dreaming)
Very dumb question, but how exactly does gene therapy work in vivo? Don't you have to modify every cell (or affected type of cell) with a new copy of the DNA? How does that actually.... happen?
I strongly also think we're going to have a social network for humans-only (and one for AI-only, and one for humans + AI)
have you looked into world.org? They're scanning eyes and giving robust human identities. Before immediately dismissing, I'd check out the video -- https://www.youtube.com/watch?v=SXnwoMxKHV8
I think atproto + worldid can one day be very interesting
What about the current DeFi framework makes it impossible to address?
Imo smart contract engineering has much more in common with firmware, hardware, etc. than the mindset a typical React/web2 engineer might have (i.e. try things and iterate). I think that point is not talked about enough.
Yes for the most part it is. One reason being that it's just a lot easier to code and cleaner when everything is on-chain and tracked the same way.
But that's starting to change (https://centrifuge.io/) - still a long way to go ofc. I also know people who have done non-crypto collateral or just completely unsecured lending, but it does involve trust that is off-chain (legal contract, know the person, etc). I think that's a fine stepping stone (and may end up being necessary) as people figure out how to do everything on-chain and in decentralized ways (ex. using on-chain reputation)
To be fair, you do have to start with over-collateralized loans before you can get under/no collateral loans with plenty of projects trying to do the latter. That's how lending itself started (https://www.provenir.com/resources/collateral/history-of-len...).
But I agree that if it's always overcollateralized then there is obviously much less utility created (though definitely some - similar to how you can get a loan against your home).
The project mentioned, Goldfinch, has already distributed 4.5M in uncollateralized loans to emerging markets using DeFi rails and liquidity already (https://medium.com/goldfinch-fi/goldfinch-raises-11m-from-an...). My 2c is that it's a sign of what's to come.
> "How should I choose what to do? Well, how had I chosen what to work on in the past? I wrote an essay for myself to answer that question, and I was surprised how long and messy the answer turned out to be. If this surprised me, who'd lived it, then I thought perhaps it would be interesting to other people, and encouraging to those with similarly messy lives. So I wrote a more detailed version for others to read, and this is the last sentence of it."
I loved the ending. The essay was primarily for him. It seems that some of the best writing, similarly to the best products, is when you yourself are the recipient.
TSLA and Comma.Ai aren't only ones. Ghost (https://driveghost.com/) is funded by top VCs (incubated at sutter hill ventures, same firm that incubated Snowflake) with a highly technical second time founder in John Hayes (first company is the public company Pure Storage).
They're taking the same bet (i.e. CV + ML advances on a large and high quality enough data set are enough). After listening to Hoetz on podcasts, I wouldn't underestimate this approach. People trying orthogonal technical approaches win surprisingly often.
Why? You can argue that CV is enough for SDC AND add extra LIDAR sensors to make it even better/safer (esp if the cost curve of the sensors make sense for consumers).
Just because I can ride a bike doesn't mean I shouldn't wear a helmet
I’ve found RenTech to be fascinating over the years and highly recommend the book on it “Man who solved the market”
Two big takeaways for his success -
1) He was pretty early, and quite contrarian, in betting on computer and quant strategies and thus took the “low hanging fruit” early on (def wasn’t low hanging back then when no one knew or believed in computer trades strategies)
2) From the book, Rentech’s main strategy was based on “reversion to the mean” - I.e “We make money from the reactions people have to price moves”. Trading on how you think OTHERS will trade and systemizing it (ex vol and momentum) is powerful but clearly doesn’t scale when you become the market yourself
And a bonus one - despite being a math genius, he basically was failing till he brought on others. He hired the right people (ie those interested in math not finance), created the right environment, took care of logistics, and pushed on a key insight (model to trade). He couldn’t have done it by himself.
We’ve been playing around on the side with ways to share trades with friends for some time, excited to share the latest iteration with HN - a personal page for your stock trades and holdings! Once you create your own custom page (see sample - https://rob.withlaguna.com/) you'll be able to import your trades automatically (Plaid), notify followers/friends through text, and grow an audience.
In our beta we have casual users sharing with friends & family to keep them updated, new traders sharing on their social media to build an audience, and existing traders more deeply connecting with their loyal followers through text. We even have someone using it as a faster way to see their portfolio holdings over Robinhood.
Would love to get some feedback from HN! What do you like/dislike? What features should we add?
Yeah close rate is a good alternative way to look at it. GC can always have Josh as a founder reference to chat with a founder on the fence - at current market they could probably turn over in one deal (i.e. opp to invest in another business they wouldn't have won otherwise, at the same amount as Baremetrics). Great move for everyone!
Feel free to get in touch - love particularly talking about technology, business, finance, psychology, and politics!