🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

It turns out delegating/automating things in your life and operating as a platoon in group flow has alpha. Return of the hacker houses.

someone will build the Netscape of AI first $10B valuation → worth nothing in 36 months killed not by competition but by the platform evolving past needing them my guess: it's already in your portfolio

For decades, my grandfather created something that drew 50 souls to gather. Every birthday, holiday, Sunday dinner - his house overflowing. Perfect gardens, cousins swimming, aunts laughing in the kitchen. The kind of gatherings where you'd drive 3 hours just to be held by that

almost every AI company has an expiration date RAG startups: dead with Claude 4 (now) code assistants: dead by Claude 5 (18 months) biz agents of any kind: dead by Claude 6 (30 months) but platforms for human creativity? eternal.

not all hype is backed by substance ours is.

If you've used Claude Code then you know it's obviously the last way that humans code when the goal is economic productivity

If you like your own tweets there's an almost 100% chance you're afraid of failing too I don't make the rules I've been both. I recently quit. You should too

average AI denier "i have faith my stone age skill will be relevant next year" faith is cope

the model that wins will be the first that can actually think from first principles

every dollar you spend on fine-tuning is a prayer to a dying god GPT-4's failures won't exist in 18 months. neither will your startup. unless you learn the difference between building for today's weaknesses and tomorrow's strengths most founders are optimizing themselves into

my crazy takes are backed by shipping 100K lines of code every 3 months for the last 10 years

your competitor raised $50M to build AI infrastructure in 18 months: their RAG pipeline: worthless (models have infinite context) their vector database: worthless (models have perfect memory) their prompt engineering team: worthless (models understand intent) but the founder

in 2022 I built tool use before it existed in any API 500 lines of code that barely worked with GPT-3 when GPT-4 shipped, same code was suddenly worth millions the best AI products do almost nothing. they build for the wind.

every AI coding tool is building for a world that's already dead Cursor, Copilot, Replit - they're perfecting the horse carriage in 1899 the end state of cooking: humans don't code at all pre-sigma launching today. join me in 2030.

Jacob currently one of the few who is reliably understanding the trajectory.

the consumer social platforms of this era will make Facebook like a look like a peanut

4o-mini