Base Daily Recap - June 29 STATS ■ @virtuals_io crossed 50,000 AI agents launched on their platform. ■ @Litebeam_xyz achieved over 10,000 routed calls for their AI service layer since launch. ECOSYSTEM ■ @FolksFinance opened a new FOLKS and USDC liquidity pool live on Aerodrome. ■ @okutrade optimized its bridging routes to Base with zero frontend fees. ■ @Polymer_Labs enabled slippage-free bridging across 16 major networks including Base. LAUNCHES ■ @WakeOnBase launched its execution layer today for direct trading across 200 decentralized pools. ■ @FirepanHQ deployed an AI platform to continuously score token contract security risks.

compare hallucination rates across popular AI models right now: - the best ones are around 3-6% - the worst can go past 17% and that's only on document summarization. once you ask AI to do research, cite sources, or dig into niche topics... hallucinations become much more obvious: - fake citations - broken links - papers that don't exist always double-check the sources

I just realized something insane: People are actually too dumb or too brainwashed by the banking system to understand and use blockchain. 👀🦋🍄 Mass adoption won't come because billions of people suddenly learn how crypto works. It will come with the AI agentic economy. AI will use crypto to transact autonomously, even more than humans ever will. DYOR and prepare accordingly. 🫵

rather than making topdown ai agent visualizers you can just mod rimworld and see how your agents behave in your lil colony while doing real work

AI by default. Private by design. The real shift in AI isn't speed or model quality. It's trust. Most "private AI" solutions today still rely on a trusted layer. CPU TEEs. Operators. Cloud providers. Blackthorn is trying to remove that assumption. @Arcium turns GPUs into an encrypted compute layer where data remains protected throughout the entire computation. What stands out to me isn't just the privacy aspect. If institutions no longer have to trust someone else to use AI, the pace of adoption could change completely. Maybe the biggest obstacle to AI hasn't been model quality. Maybe it's been trust. gMPC ☂️

It’s almost a consensus bet now that Nvidia will launch an open-source model to offer a US version of OS & take on frontier labs. If that happens, I wonder if they’ll in turn reduce their dependencies on GPUs.

"Our country continues to advance in innovation; we are regional leaders and references in AI, thanks to the vision of President @nayibbukele 🇸🇻👏" President Bukele has leapfrogged El Salvador straight to the top as a player on the world stage. This is what that looks like👇

the 'obsidian is the codebase, the llm is the programmer' idea going around is basically how i already work. my whole product strategy lives in markdown notes claude reads and writes back to. the part nobody mentions: it only compounds if the agent maintains the notes, not just reads them. dead notes stay dead either way.

“An armed man is a citizen, an unarmed man is a subject” The same principle applies here. If only governments and a handful of corporations have access to AI (or restrict it), the rest of society becomes dependent on them. Open source carries risks, but concentrated power carries even bigger ones. If you are scared of the weapons you should not invent them in the first place.