After getting access to @typesafeai's JEV this weekend, it was obvious I had to build a weekend project with it. I first played around with JEV to strengthen the token moderation layer on the @rwagmicom frontend. Then I decided to work on a fun onchain experiment. I created a pot and seeded it with 0.1 ETH. Guarding that pot is JEV: a very stingy judge who will only release everything to someone who submits an exceptionally good plea. You have to burn 10 million BEGJEV tokens every time you submit a plea. I just launched BEGJEV on RWAGMI with a 3% trading fee. Because RWAGMI distributes trading fees in both assets rather than only ETH, I wired the tokenomics like this: 2% of the ETH-side trading fees feed the pot. All BEGJEV collected from trading fees gets immediately burned. The remaining 1% ETH-side fee goes to my wallet. Half of that will be used to buy RWAGMI and the other half will be sent to @panik_eth, who has always been a huge help so that he can figure out ways to feed the BAES ecosystem. Not to mention that the BEGJEV I will be collecting will also be burned. I basically designed this thing to become a token sink. A crazy one at that. If you want to play with Typesafe's new frontier model and see how far you can push it: begjev.com Shoot your shot. The pot keeps growing for as long as people find the token interesting enough to trade. And please do not top blast your entire net worth at the pico top and then come here crying. This is a fun experiment. I would also prefer not receiving death threats for building something people can have fun while trying to break a frontier model. Trust assumptions: I retain a 1-day time-locked power to rescue funds from the contract holding the pot. I added the rescue function because even I don't know whether JEV can be broken under the rules I gave the agent, and this is an experiment. Other than that, go play. Try to prompt-engineer JEV. Try to break the game or the contract. Think of it as a bounty hunting game where the best beggar wins. And if someone completely breaks it, I can always start another round with stricter rules. PS: I am not going to pay dexscreener, use basedbot or gmgn or llamaswap anyways. CA: 0xb200000000000000000000ad822dc9282ebe66ed

Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family. It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.

It’s possible to write bug free code. In ~2 years it’ll be impossible not to.

Jev by @typesafeai might change how we control robots. We compared Jev, GPT-6 Astra and GPT-4.1 mini in MuJoCo. One apple. One plate. Each model chooses intent → X/Y/Z direction + gripper open/hold/close. 🧵

B20, explained from beginner to expert Beginner: B20 is Base’s native token standard Think of it as a faster, cheaper, safer way to create and manage tokens on Base, built right into the chain itself