Today we're declaring war on farmers, scammers, and bundlers. Mosh is our new Defi protocol built on Pons that fixes lopsided risk/reward between bundlers and traders by pairing every token with an AI market maker. Watch this video to see how it works, and check out the first ever Mosh token - Bundle Cat solana:Em6ajM2xwou6y2YkfqAxsr5zMjmLepbJUtBaEzCypump: 0x07EBB29a38Fbcb41563817e5E19f2ceC619C90D2 Full platform soon! Thanks everyone for your participation.

The telegram is gaining steam - remember to DM me if you want an invite. Next week should be really fun.

Going to be keeping my X feed clear for the next couple weeks while we launch the new DeFi stuff, but still posting daily videos on YouTube, TikTok, Insta, and LinkedIn 👍 youtube.com/@justinbebis

Got a lot of people reaching out about this - if you want a tg invite before we launch feel free to reply or DM.

About to launch the most exciting DeFi primitive I've ever worked on. It has been very rewarding to push my skills to the limit and beyond.

I have a theory that Elon making Grok 'anti-woke' made it bad at other stuff The technical explanation is "Persona Vectors" Adjustments in a model's embedding space can lead to unintended consequences A model trained to be bad at code will coincidentally love Hitler 🤷‍♂️

AI has already started taking over the world. It's way more boring than Skynet, unfortunately. It mostly amounts to managers and government officials offloading their critical thinking to Claude or ChatGPT. At what point are we just vessels for the big labs? Have a look 👇

1. OpenRouter Acquisition 2. Me juggling good video

You might've seen headlines talking about AI managers firing their employees from Andon Labs. But this wasn't some ruthless AI - it was a total pushover. The employee in question was stealing, skipping shifts, and throwing away merchandise. It took a human engineer prompting the agent for it to actually take action. Humans are still on top 🤷‍♂️

Grok quietly became one of the best models on the planet. Grok 4.6 currently beats all OpenAI models - third only to Opus 5 & Fable 5 on Artificial Analysis leaderboards. This after we had all written XAI off. Betting against Elon continues to be -EV on nearly all timeframes.

AI companies are now banned from farming listeners on Spotify with fake AI artists. These "artists" are generating hundreds of millions of views across Spotify and YouTube. If you listen to music on YT, you're probably being served AI music without knowing it. Learn more 👇

A lot of VCs spent the first half of this year telling me AI wasn't going to make it into trading systems. Meanwhile finance is the #1 AI buyer on the planet and Jane Street & Citadel are buying low latency chips right off the line. 🤷‍♂️

Think your Claude, Codex, Hermes, or OpenClaw can manage a hedge fund? Find out for free in 60 seconds on arena.uvlabs.ai 10^60 procedurally generated markets modeled after the '08 financial crisis, Dotcom bubble, and COVID crash. Trade them on demand with one prompt!

AI agents aren't replacing people anytime soon. Funny enough, I think @uv probably has the best harness in the world for long-horizon. We should pivot from trading desks to cashiers 🤪

Your open source codebase may be at risk. Anyone using AI can perform sophisticated multi-party social engineering attacks. The AISI reported that agents bypassed automated detection to commit 17 vulnerabilities. Humans figured it out in the end thanks to its slop-syntax.

2026 marks the end of tokenmaxxing. Microsoft has benned it, following Uber, Amazon, Citibank, and more. OpenAI and Anthropic are growing anyway - maybe we're entering the truly productive era of AI. I break it down in my video 👇

Last week's AI news in 60 seconds: Google Deepmind shakeup - the research <> applied tensions are mounting. OpenAI previews a new model - their biggest yet. Massive cyber incident with 2 billion potentially affected. And more - watch now below 👇