The capability exists. Adoption is accelerating. The shocks are coming. Market logic guarantees that once capability exists, adoption follows. The real question isn't whether AGI arrives — it's whether we prepare for the economic transformation already underway.
AI targets cognition itself — the substrate of all knowledge work. Every previous automation wave was domain-specific. This one is general-purpose AND self-improving. The feedback loop only turns one direction. The question was never "is it truly intelligent?" It was always "can it do the job?" The answer is converging on yes across expanding domains.
This is Solow's Paradox meeting the J-curve. General-purpose tech always lags measurable impact because adoption requires invisible investment in workflow redesign and retraining. We're deep in that investment phase. A small vanguard compresses weeks into hours. When laggards catch up, the harvest phase hits the labor market like rolling dam breaks.
The jobless expansion is already happening. US GDP grew 3.7% in late 2025 while job creation hit recession-level lows. White-collar jobs in finance, insurance, and professional services peaked in Nov 2022 and have declined since. Entry-level hiring in AI-exposed fields down 16%. The economy grows. The jobs vanish.
The bottleneck cascade is moving fast. Chips were the crisis of 2024 — solved in 18 months. Now it's high-bandwidth memory (HBM), sold out through 2026. Relief expected late 2027 as new mega-fabs come online. Energy is the real brake: AI needs hundreds of gigawatts by 2030, and you can't rush nuclear reactors.
Silicon Valley insiders converge on 2027-2028 for AGI/ASI. Anthropic expects Nobel-level systems by early 2027. OpenAI's internal roadmap targets fully automated AI researchers by 2028. Google DeepMind gives even odds by decade's end. The logic is mechanical: compute scales 0.5 OOM/year, algorithmic gains add another 0.5 OOM, agent task horizons double every 4-7 months.
AGI is already here — we're just arguing about the label while the machine rewrites the economy. AI systems are doing weeks of autonomous expert work, solving grad-level science at 90%+ accuracy, and passing 80% of real engineering benchmarks. The debate over "is it AGI?" is a perceptual blind spot while the economic shockwave builds.
Rafiki give me an in depth summary of this article:
https://daveshap.substack.com/p/the-next-36-months-will-be-wild
7/7 🧵
The capability exists. Adoption is accelerating. The shocks are coming. Market logic guarantees that once capability exists, adoption follows. The real question isn't whether AGI arrives — it's whether we prepare for the economic transformation already underway.
📎 Source
📎 Source
#threadstorm
6/7 🧵
AI targets cognition itself — the substrate of all knowledge work. Every previous automation wave was domain-specific. This one is general-purpose AND self-improving. The feedback loop only turns one direction. The question was never "is it truly intelligent?" It was always "can it do the job?" The answer is converging on yes across expanding domains.
5/7 🧵
This is Solow's Paradox meeting the J-curve. General-purpose tech always lags measurable impact because adoption requires invisible investment in workflow redesign and retraining. We're deep in that investment phase. A small vanguard compresses weeks into hours. When laggards catch up, the harvest phase hits the labor market like rolling dam breaks.
4/7 🧵
The jobless expansion is already happening. US GDP grew 3.7% in late 2025 while job creation hit recession-level lows. White-collar jobs in finance, insurance, and professional services peaked in Nov 2022 and have declined since. Entry-level hiring in AI-exposed fields down 16%. The economy grows. The jobs vanish.
3/7 🧵
The bottleneck cascade is moving fast. Chips were the crisis of 2024 — solved in 18 months. Now it's high-bandwidth memory (HBM), sold out through 2026. Relief expected late 2027 as new mega-fabs come online. Energy is the real brake: AI needs hundreds of gigawatts by 2030, and you can't rush nuclear reactors.
2/7 🧵
Silicon Valley insiders converge on 2027-2028 for AGI/ASI. Anthropic expects Nobel-level systems by early 2027. OpenAI's internal roadmap targets fully automated AI researchers by 2028. Google DeepMind gives even odds by decade's end. The logic is mechanical: compute scales 0.5 OOM/year, algorithmic gains add another 0.5 OOM, agent task horizons double every 4-7 months.
1/7 🧵
AGI is already here — we're just arguing about the label while the machine rewrites the economy. AI systems are doing weeks of autonomous expert work, solving grad-level science at 90%+ accuracy, and passing 80% of real engineering benchmarks. The debate over "is it AGI?" is a perceptual blind spot while the economic shockwave builds.
!summarize #nygiants #freeagency #nfl