Building exactly this right now. We're moving our entire marketing agency operation and clients into an agent-based system. Structured tasks, contextual input bundles, continuous learning loop.
Right now we're closing the end-to-end Meta cycle. Research to creatives to live campaigns to reporting, one loop. Every cycle compounds. The system learns from its own output.
Point 3 you mentioned is crucial. Once you have the system, point 4 becomes obvious. You can productize because delivery is repeatable. Right now most companies are stitching together 10+ fragmented tools, each still requiring someone in-house to operate. The tools aren't the bottleneck anymore. The coordination layer between them is.
Currently our Imps (agents) keep spawning across departments. One runs competitive intelligence, another writes Meta ad copy, another handles reporting.
Meta i Andromeda pchają Cię w kierunku wrzucania jak największej liczby reklam (patrz ads library dużych ecom brands), a targetowanie dzieje się głównie przez samą kreację. Dlatego największym wyzwaniem nie jest już ustawienie kampanii, tylko ogarnięcie i analiza setek wariantów reklam..
Manus będzie robił to za Ciebie, będzie monitorował dane, wyłapywał schematy i pokazywał, co naprawdę działa, co skalować, a co wycinać. Dostajesz konkretne wnioski, a decyzje podejmujesz na podstawie twardych, skondensowanych danych.
@ttunguz a fantastic list, @ttunguz. One thing I’d add, by 2026, AI won’t be constrained by chips anymore, but by power.
Chip supply scaled fast. Energy, grids, and permitting didn’t keep up. Electricity is becoming the real moat. The winners will be those who find a way to solve energy
What would the world look like if students were taught to defy skeptics instead of become them?
The future isn’t built by doubt - it’s built by belief.
Jony Ive x OpenAI
Rick Rubin x Anthropic
Next: Pharrell Williams x Google?
The next wave of GTM isn’t just technical, it’s deeply creative.
Designers, producers, and culture-makers are now shaping the future of AI.