market’s a mess, hacks keep happening, and @Lighter_xyz team just keeps adding things. 23 stock tokens now available as collateral, with the tokenized asset lineup still growing. pretty impressive how consistently they keep building while everyone’s glued to the charts and the latest drama. real builders win eventually, and i have no doubts about that
OpenAI just released lifeval because academic benchmarks only tell you so much about whether ai can actually make your life easier it says a year of work made its models faster and better at using browsers, computers and connected apps, with Gpt-6 astra posting best recorded scores of 100% on selected tasks like booking car inspections or ordering groceries within budget that’s kinda interesting data because people will trust ai with more of their lives if it reliably gets things done. benchmarks like this push development toward the messy, everyday tasks that actually earn that trust
gpt-6.2 astra 👀
an ai agent is now driving... a SETI telescope Nvidia says AI agents are now helping run the Allen Telescope Array in California, from controlling its dishes to processing up to 2.4 terabits of radio data per second, all through plain English prompts basically, the telescope hears a flood of radio noise every second, and a filter decides what’s worth a second look and what gets thrown away. now a researcher types one request and the agent rewrites that filter, so a human gives the order but the AI’s code decides what could even count as an alien signal (!) (so far its big win is a real-time catch of the crab nebula pulsar, a dead star astronomers already knew about)
back in 2022, when everyone started rushing into ai, half the startups were basically glorified gpt wrappers. but comparing ai back then to what we have today is almost ridiculous crypto is sort of already becoming part of the ai stack itself, from agent wallets to stablecoin payments. these aren’t two competing trades (even if the investment cycles differ)
ai agents are taking security proofs from handmade to mass production v12 and Zellic say it’s the first published proof covering a whole x86 program run, checked on the raw machine code the CPU runs, like inspecting the built bridge instead of the blueprint, and a second similar program took just 739 new lines. would be nice to get a checkable security guarantee about the compiled program itself, with fewer layers taken on trust. reusing the hardest parts means the next similar program takes much less work to verify, which could make rigorous security proofs practical at scale! it’s important to note that these are their own results, and they acknowledge their x86 model is still in development and hasn’t been audited by a human yet, but it’d be pretty cool if this scales https://t.co/CS5gv6j3uz