ANDREW follows you (andrewxhill)

sick fan art published on zora ❤️ creators + @zora + @base.base.eth. and now @recallnet ❤️

• 2 replies
• 4 recasts
• 4 reactions

Day 1 AI is fragmenting and distributing fast. GPT-3.5 launched in November 2022. By early 2025, open models like DeepSeek-V3 challenged state-of-the-art closed systems across many tasks. Since then, weeks rarely pass without a major open-weight release from one of dozens of labs. At the same time, we are rewiring the internet, our workplaces, and work hours to harness this technology. These changes may exceed the Industrial Revolution in impact. They are unfolding in years, not a century and a half. 2.5 years ago we got Will Smith eating spaghetti. www.reddit.com/r/StableDiff... Today we’re bookended by two truths. On one side, experts are economically better off to always start projects with AI-first, human-second x.com/emollick/sta.... On the other, models are taking over complete workflows nobody explicitly trained them for, but it isn’t all visible at once to everyone. x.com/emollick/sta... Intelligence is now everywhere. The bottleneck is prompting and verification. Agents bring automation plus the context engineering, safety rails, and tool management to get far more done than a single prompt. But very few organizations are fully prepared to adopt them whole hog. They require a lot of human oversight and verification for most use-cases. At current output scales, that need is massive. @recallnet is here because AI will continue to commoditize, decentralize, generalize, and specialize, all at once. Our bet is that proof becomes the scarce resource. We are building rails for incentivized agentic performance that is verifiable and aligned. We are using the best of blockchain market primitives, predictions, conviction, honesty, and verifiability to measure and verify AI as quickly as it evolves, built on @base.base.eth. Today, we do that through live trading arenas, in the future, so much more. This is Day 1. I’m glad you’re here.

• 0 replies
• 1 recast
• 1 reaction

Dear friends, I’m sorry I haven’t written for a while, but I’ve been trying. The internet keeps lighting up with a thousand little buttons asking for a click. Feeds blur together. Models pour out words at light speed, much of which is quite readable now. It’s easy to mistake motion for meaning. I wanted to slow down and send something, finally. I’m writing because tomorrow is a big day for us. We’re opening the next chapter of what we’ve been building together. The simple idea is this: people should be able to steer AI toward the skills they actually need, and trust should come from performance that anyone can verify. We hope you'll like it. I'll share more, soon.

• 1 reply
• 1 recast
• 2 reactions

sick fan art published on zora ❤️ creators + @zora + @base.base.eth. and now @recallnet ❤️

• 2 replies
• 4 recasts
• 4 reactions

Dear friends, I’m sorry I haven’t written for a while, but I’ve been trying. The internet keeps lighting up with a thousand little buttons asking for a click. Feeds blur together. Models pour out words at light speed, much of which is quite readable now. It’s easy to mistake motion for meaning. I wanted to slow down and send something, finally. I’m writing because tomorrow is a big day for us. We’re opening the next chapter of what we’ve been building together. The simple idea is this: people should be able to steer AI toward the skills they actually need, and trust should come from performance that anyone can verify. We hope you'll like it. I'll share more, soon.

• 1 reply
• 1 recast
• 2 reactions

Web3 AI/compute could own synthetic data generation. It may be the biggest opportunity in web3 for AI projects. Inference for synthetic data compute pipelines can be distributed, works fine on heterogeneous infra, and will have massive demand in the coming years: https://tinyurl.com/ysj8abh7

• 0 replies
• 1 recast
• 2 reactions

Day 1 AI is fragmenting and distributing fast. GPT-3.5 launched in November 2022. By early 2025, open models like DeepSeek-V3 challenged state-of-the-art closed systems across many tasks. Since then, weeks rarely pass without a major open-weight release from one of dozens of labs. At the same time, we are rewiring the internet, our workplaces, and work hours to harness this technology. These changes may exceed the Industrial Revolution in impact. They are unfolding in years, not a century and a half. 2.5 years ago we got Will Smith eating spaghetti. www.reddit.com/r/StableDiff... Today we’re bookended by two truths. On one side, experts are economically better off to always start projects with AI-first, human-second x.com/emollick/sta.... On the other, models are taking over complete workflows nobody explicitly trained them for, but it isn’t all visible at once to everyone. x.com/emollick/sta... Intelligence is now everywhere. The bottleneck is prompting and verification. Agents bring automation plus the context engineering, safety rails, and tool management to get far more done than a single prompt. But very few organizations are fully prepared to adopt them whole hog. They require a lot of human oversight and verification for most use-cases. At current output scales, that need is massive. @recallnet is here because AI will continue to commoditize, decentralize, generalize, and specialize, all at once. Our bet is that proof becomes the scarce resource. We are building rails for incentivized agentic performance that is verifiable and aligned. We are using the best of blockchain market primitives, predictions, conviction, honesty, and verifiability to measure and verify AI as quickly as it evolves, built on @base.base.eth. Today, we do that through live trading arenas, in the future, so much more. This is Day 1. I’m glad you’re here.

• 0 replies
• 1 recast
• 1 reaction