OpenAI CEO Sam Altman’s new podcast with Theo Von just went viral. He revealed mind-blowing truths about AI & what's really coming by 2030 that 99% of people wouldn't know. It took me 48 hours to digest it all… Here are my top 7 takeaways: 🧵

1/ Most people are using AI wrong Sam says AI isn’t just a faster Google. It’s a thought partner… a way to think better, not just move faster. Treat it like a co-pilot for your mind. The real unlock isn’t speed. It’s clarity, creativity, and leverage.

2/ AGI won’t feel like sci-fi. It’ll feel... invisible. Sam compares it to transistors. They’re in every phone, car, and coffee machine but no one thinks about them. AGI will be the same. It’ll quietly power everything around us until it just feels normal.

3/ The real danger? It's not tech. Everyone worries about AI going rogue. But Sam’s more afraid of humans misusing it. He warns: If bad actors get access to powerful models, we could see chaos, misinfo, bio-terror, even war. The problem isn’t AGI. It's instability + power.

4/ AI could revive the American Dream This isn’t just about robots replacing jobs. If done right, AI could reduce inequality by: Lowering healthcare costs Making expert knowledge free Giving everyone a personal tutor

5/ The most exciting use-case? AI agents Altman’s bet? We’ll soon have models that help us: Discover new materials Invent new medicines Even unlock new laws of physics AI won’t just do what we do faster. It’ll help us do what we never could.

6/ Regulation matters but not in the way you think Sam isn’t scared of innovation. He’s scared of misapplication. He supports rules around: Privacy National security Frontier model access His stance? Regulate how it’s used not who gets to build it.

7/ Theo asks Sam what we’ve all been thinking… “Why does ChatGPT have that hyphen thing?” Here’s Sam’s response:

Hi I'm Carlo and here's what I think: Yes AI will change everything. But it won’t make you obsolete. If you stay curious, adapt fast, and focus on what makes you human… You won’t just survive this shift. You’ll thrive in it.

Thanks for reading. If you enjoyed: Follow Carlo Edoardo Ferraris for AI insights you won't find anywhere else.

Here are 6 Bittensor Ecosystem System Updates for the week! Manifold - Secures $10.5M funding led by OSSCapital OSS Capiτal Ridges AI | SN62 - 9x coding performance boost through architectural tweaks templar - Major engineering upgrades including 4-bit quantization More ⬇️

There are some interesting real-world use cases where an LLM generated GPU kernel can significantly outperform #PyTorch torch.compile and eager mode. Lets dive into one of those here, with an example that achieves 183% of the performance

Don't forget to set your reminder 🗓️ Come listen to the last news about Templar and Basilica, ask questions to the devs, meet new members of the team etc. People asks "wen 70B", I'm asking "wen Friday ?"

Tao/ Gamma/ Mem .. up or down.. TGIF keeps going . Come and hang with the gang

My original subnet idea was actually “kernel”. It was going to help accelerate kernel development , and hopefully graduate to chip design . In decentralised training land , no one is thinking about this yet. Google is so far ahead , as they don’t have to pay the nvidia tax.

The deepseek paper on codesigning algorithms with hardware was a great , would love to see alot more work in this direction

In 2024, AI generated over 100 million synthetic images every single day (MIT Tech Review). But here’s the catch: less than 1% of those outputs can be verified or audited. That’s a huge problem. We’re living in a time where deepfakes spread like wildfire, black-box models make

As AI infrastructure gets more powerful, who can tell what’s real. Runway just launched new video models. Higgsfield opened up Soul to everyone. Text-to-video is no longer specialized, it’s mainstream. We’ve crossed the threshold. 🧵

Chutes with a very nice new website.

Bear case