Read through a great thread by Coursera founder Andrew Ng, where he broke down why AI-native teams (those building processes around AI agents from day one) operate nothing like classic big tech:
🔹 Great engineers now aren't just writing code. They're simultaneously product managers, designers, and sometimes even marketers.
🔹 Small teams (2–10 people) working in the same office, able to talk face to face, move at insane speed.
🔹 The main bottleneck now isn't code — it's deciding what to build. So the best teams are drastically shrinking the engineer-to-PM ratio (from 8:1 down to 1:1 or even lower).
🔹 The fastest teams are those where the engineer personally understands the users, makes the product decisions, and immediately ships them. No extra approval loops.
When coding gets 10–100x faster, everything else starts dragging: design, marketing, legal, compliance. New bottlenecks appear that nobody even thought about before.
In small AI teams, generalists win — people who are deep in their own specialty but can quickly pick up adjacent areas. AI tools help a lot here.
And the most important thing: even in a two-person team, minimizing communication overhead is critical for maximum speed. That's why the best results come from teams working together in the same room.
Worth noting that this thread is specifically about small AI-native teams (2–10 people). Andrew promises to write about coordinating larger teams later.
> "I recognize that these role changes are hard for many people. But I'm also excited: individuals and small teams willing to learn new skills can now accomplish so much more than before. This is a golden age of learning and building!" ✨
> — Andrew Ng
(Just remember to step away from the screen occasionally and actually talk to your teammate about the feature.)