@Tezos co-founder w/ wife @breitwoman & Agitprop founder. Aligning ASI to not zap everyone is unsolved, that's bad. Tezos stuff, high context humor & more.
Announcing https://t.co/MGa5lVXtEM which I’d love feedback on as we plan to go live on Monday. Trade perps on major stocks like SPCX, NVDA, META, SKHY, SDNK and more @ up to 500X leverage against a user-loss funded LP smart contract. Price oracle provided by @PythNetwork and built on the HyperEVM chain with 20% of the EVERYTHING tokens going to holders of $PAPER. Users can also sell options via Promise Orders and collect premium.
When I saw https://t.co/cVr0yRUHZQ it was impressive how novel, pristine and effective the system design was in terms of bootstrapping both the initial liquidity and user base that every new exchange so badly needs and most never get and thus never get to critical mass.
Its mechanism enables 1000x on BTC and ETH but can also be applied to 200+ stocks for enabling much more liquidity in these RWA perps than exists on major perp exchanges today, albeit with a payout asymmetry that may deter some sharp traders and arbitrageurs, but should be highly compelling to pure directional leveraged users.
I’ve started several exchanges over my 14 years in crypto and most require large amounts of VC capital + a meaningful team to operate and are subject to attacks from all sides. This LP and leverage model enables a hyper-focused approach purely on a product users love and communicating that product to the audience.
We also wanted to reward PAPER holders with 20% of the loss-earned token supply, giving them a chance to participate in the build out of the ecosystem PaperTrade has created. We will funnel our BTC and ETH perp volumes through https://t.co/cVr0yRUHZQ as well once Builder Codes are released.
Would love your thoughts and feedback on this concept. Check out the docs https://t.co/tAXilGFjfa and try the demo.
Our transferability plans are here: https://t.co/vQQiGeRJtL - again feedback appreciated.
This Hoover Institution study, which has fueled the deluge of ads (more than $200 million spent by billionaires so far) to defeat the California billionaire tax, is spectacularly misleading.
Let me explain its crucial flaw. 🧵
"Every billionaire who lived in the state as of Jan 1 2026 will get kicked in the balls once. So there's no incentive for them to move, since they'll get kicked once anyway. Why would they be irrational enough to move to avoid future kicks?"
You're in a trench, you found the cellphone of your fallen comrad. You can call for help but you don't know how to unlock it. You realize the smudges on the screen indicate his code only uses 0 and 1. You only have three tries. You obviously try 0000 and 1111 first. There are no frequentists in fox holes.
@glebazz@DanNeidle I calculated about £18,800 a year, so £1,560 a month, or just £51 a day. It's higher than the US figure because it includes state and local taxes.
Makes complete sense, it's trivial to insert a backdoor that triggers malicious behavior in certain environments, ala Stuxnet. And currently we have no way to know from looking at the weights whether such a backdoor exist. This is a problem, and it's closely related to alignment.
The Department of Homeland Security is requiring all civilian federal agencies to remove DeepSeek products and 'any third-party services developed using DeepSeek implementations,' and to stop buying them.
You can pedantically slice and dice the whole stack it doesn't change the argument or the conclusion. The weights can be trained so that, when the entire system is deployed in a particular environment or context, malicious software is deployed. You can absolutely be hacked by a model that's been trained to hack you, even if you use it locally, regardless of your harness.
Agentic environments use the weights to output bash command and large pieces of code that will be executed. You can try to put some simple guards around it, but the model will easily find their way around it. You can have a sophisticated model review everything, in which case, if you trust that model, why aren't you using it in the first place.