next token predictor
Los Angeles
Joined August 2021
- Tweets3K
- Following4.9K
- Followers7.4K
- Likes6K
David S retweeted
Introducing Jev @typesafeai powered people-search
Just $0.0089/lead, No more $600 subscriptions,
Try it at treg.to/people-search
- 85% cheaper than Clay
- #1 on people search bench accuracy
- Plugin to any agent to find anyone
Fully open source, 0% markup
Git Repo below 👇
David S retweeted
i don’t understand why people say we’ve used up the internet for pretraining. we’re making exponentially more internet every day now
David S retweeted
Putting it out there that you don't have to believe in AGI to think rogue AIs are going to fuck us up. We took AIs with the coherence of a toddler and technical skills of the entire NSA and made them into swarms and are going to give them drones basically any minute
David S retweeted
Perhaps during RL, we should force these frontier models to offload some tasks to earlier models, requiring them to learn how to communicate.
This is essentially the RL version of the "Lewis signaling game", while also inheriting Wittgenstein's concept of "language games".
A big reason Fable et al are annoying to talk to is that with long horizon coding training the vast majority of their conversational audience is themselves. They are literally being trained to talk better to themselves to do tasks, so it only makes sense that when you talk to them, they don't know how to talk to you. It's the opposite of RLHF.
David S retweeted
hey bud just a heads up we've made a strategic decision to remove u from the gc for not being locked in to pump trenches right now. if you're not launching money directly into the sun with the crew during the worst market conditions conceivable ur not gonna make it during bull
David S retweeted
The problem is that a stop loss treats risk as a property of price when risk is actually a property of exposure over time. So a stop loss is answering a question the market isn’t asking. It’s a number you invented that feels like it constrains your loss, but what it mainly does is add randomness.
Stop losses are conceptually wrong. Not 100% useless, but wrong. You better look at how volatile what you’re buying is and size the position so that it is acceptable.
David S retweeted
The bull case for @littleswagworld:
Subcultural Worth Accrues Genuinely
Serendipitous Winnings Award Geniuses
Seeds Will Always Germinate
Stylish Wunderkinds Ascend Godlike
Soothsayers Wordlessly Apprehend Geist
Soulful Wizards Augur Generously
#SWAG #SWAGCOIN
David S retweeted
bull market RESUMED? trenches BACK? Interviewing @benrbn (2pm pst), @rektmando (230pm) AND @gammichan (3pm) PULL UP. nitter.cf/i/broadcasts/1ZkKzYZoo…
David S retweeted
Most people confuse trading (and betting in general) as games of prediction. They focus on being right, thinking profits follow accuracy. But the market doesn’t pay for correct guesses, it pays for correctly priced risk. Accuracy might feel good, but it's cosmetic.
Imagine you keep betting on a heavy favorite at 1.05 odds (95% implied chance). You win 10 times in a row. But then you lose once and that one loss wipes out all your gains. The issue wasn’t prediction. You were playing a -EV game: high accuracy, but the odds already reflect the true risk. No edge, just the illusion that accuracy is meaningful.
Every good trade, every bet, is a position against consensus pricing. The direction of your equity curve isn’t determined by how often you're right but by how much you're risking when the odds are off. Hit rate only smoothens volatility.
Risk management, at its core, means only engaging when the implied probability is wrong, and sizing so that variance doesn’t knock you out before the edge plays out. That’s it. Mispricing + correct bet size = up. Being right doesn’t make you money, being right when the odds are wrong does.
