Usual debate. In AI, better sell shovels or look for gold? The AI zeitgeist shifts every 6 months: • 𝐏𝐡𝐚𝐬𝐞 1 (Post-OpenAI launch): "AI apps are just wrappers; only labs (and NVIDIA) will profit!" • 𝐏𝐡𝐚𝐬𝐞 2 (Deepseek R1 drop): "Models are commodities; real value is

How can we prevent LLM outputs from being poisoned like Google results were/are after decades of SEO manipulation?

Luck is what happens when preparation meets opportunity

Funny fact of the day: The center of the AI revolution isn’t the Bay Area, it’s the tiny Caribbean island of Anguilla. They're making over $100 million a year just from selling their .ai domains. That’s 20–25% of their GDP. Talk about being in the right place at the right

The future of interfaces is voice in, text out, because we speak faster than we type, but read faster than we talk Next step: Neuralink

I’ve been regularly angel investing in the biggest 500 companies in the US. Haven’t missed a single one. Guaranteed by S&P.

x.com/i/article/1869…

What is a bot and what is an agent executing human will?

Bigger news than GPT5

Teenagers will be both brainrotted and brilliant. Imagine growing up with something that answers any question instantly. I’ve learned so much more in the (less than) 3 years since ChatGPT launched than I otherwise would have. Now imagine compounding that for 20 years.

William Goldman: "Nobody knows anything" Customers don't know what they want, and neither do visionary founders. Steve Jobs made a lot of products that flopped. People don't come with source code. That's why markets and business are so unpredictable. We're just making guesses

SaaS --> wrappers around databases AI apps --> wrappers around intelligence

No AI tool has figured out voice/tone for writing content. Why do you think that is? - Unrealistic expectations of the writer - Improper post-training/tuning of the models - Not enough context engineering by writers This feels like a solvable problem that no one has cracked.

Microchips made computation cheap. The internet made distribution free. AI is making creation nearly free. When costs collapse, revolutions follow. What comes next?

Availability and novelty bias in humans is fascinating. The AI apocalypse taking our jobs is kind of over. The real issue was slower hiring and "prove AI can't do it before adding headcount". It's happening now, likely to continue, and almost no one’s talking about it. A few

The number of times ChatGPT-5 uses “thinking” when it’s not necessary is annoying. It might be a clever trick. ChatGPT looks fast. Probably it is a smaller model. Fewer model layers = cheaper per step for OpenAI --> get quality by generating more output tokens. --> Costs shift

I see everybody is hiring Forward-Deployed Engineers. This is not new, Palantir is credited for inventing the role B2B AI rarely works out-of-the-box. Even after ingesting your data, most vendors need 30+ days of human-led onboarding to make it usable. Forward-Deployed