No PriorsNo Priors Ep. 128 | With Andrew Ng, Managing General Partner at AI Fund
Sarah Guo and Andrew Ng on andrew Ng on Agentic AI, Rapid Engineering, and Founder Mindsets.
In this episode of No Priors, featuring Sarah Guo and Andrew Ng, No Priors Ep. 128 | With Andrew Ng, Managing General Partner at AI Fund explores andrew Ng on Agentic AI, Rapid Engineering, and Founder Mindsets Andrew Ng discusses where future AI capabilities will come from, arguing that while scale still helps, real progress now depends on agentic workflows, better tooling, and disciplined engineering rather than sheer model size.
At a glance
WHAT IT’S REALLY ABOUT
Andrew Ng on Agentic AI, Rapid Engineering, and Founder Mindsets
- Andrew Ng discusses where future AI capabilities will come from, arguing that while scale still helps, real progress now depends on agentic workflows, better tooling, and disciplined engineering rather than sheer model size.
- He defines “agentic AI” as a spectrum of systems with varying degrees of autonomy, with coding agents currently being the most successful and economically valuable examples.
- Ng explains how AI-assisted coding is radically shrinking engineering headcount needs and shifting the bottleneck from software development to product management, thereby raising the bar for technical, high-empathy founders and product leaders.
- He also forecasts that individuals who aggressively adopt AI tools across professions will be dramatically more capable than their peers, and that small, highly skilled, AI-leveraged teams will increasingly outperform large, traditional organizations.
IDEAS WORTH REMEMBERING
7 ideasAgentic AI is a spectrum, not a binary category.
Ng coined “agentic AI” to capture varying degrees of autonomy—from simple LLM-driven steps to multi-step planning agents—so teams stop arguing about definitions and instead focus on building useful workflows.
The main bottleneck to agentic AI adoption is talent and disciplined engineering, not core model capability.
Teams that systematically use evals and error analysis to refine agents vastly outperform those tinkering randomly, and there aren’t yet enough people with these skills.
Coding agents are currently the most valuable and mature form of agentic AI.
Tools like Claude Code can autonomously plan and execute multi-step coding tasks, creating huge productivity gains, while more general “computer use” agents remain mostly demo-grade.
AI-assisted coding is shifting the startup bottleneck from engineering to product management.
Because one engineer can now do what used to take a small team, the real constraint becomes deciding what to build, how to prioritize, and how quickly to iterate based on customer insight.
Technical, AI-native founders have a decisive advantage in this era.
Ng believes founders who deeply understand fast-evolving AI capabilities will out-execute more traditional, business-only leaders, especially during periods of rapid technological disruption.
Small, highly skilled teams with heavy AI leverage can beat large organizations.
He sees small, co-located teams using AI tools outproducing much larger, lower-cost or bloated teams, suggesting future companies can stay lean yet highly impactful if they adopt AI aggressively.
Individuals who fully embrace AI tools will become dramatically more capable within a few years.
Just as AI transformed programmer productivity in two years, Ng expects similar capability jumps across other jobs, with early adopters gaining a large performance edge over peers.
WORDS WORTH SAVING
5 quotesThe single biggest barrier to getting more agentic AI workflows implemented is actually talent.
— Andrew Ng
Vibe coding makes people think it's easier than it is. After a day of AI-assisted coding, I'm exhausted mentally.
— Andrew Ng
Today, the bottleneck is deciding what we actually want to build.
— Andrew Ng
In moments of technological disruption, having a good feel for what this technology can and cannot do is the real knowledge.
— Andrew Ng
People who embrace this will be so much more powerful and so much more capable than they probably can imagine.
— Andrew Ng
QUESTIONS ANSWERED IN THIS EPISODE
5 questionsHow can non-technical professionals practically build the kind of AI instincts Ng says are now essential?
Andrew Ng discusses where future AI capabilities will come from, arguing that while scale still helps, real progress now depends on agentic workflows, better tooling, and disciplined engineering rather than sheer model size.
What concrete frameworks or playbooks exist for doing the disciplined eval- and error-analysis Ng sees as key to building effective agents?
He defines “agentic AI” as a spectrum of systems with varying degrees of autonomy, with coding agents currently being the most successful and economically valuable examples.
Where are the next “coding agent–style” breakthrough use cases likely to emerge outside of software engineering?
Ng explains how AI-assisted coding is radically shrinking engineering headcount needs and shifting the bottleneck from software development to product management, thereby raising the bar for technical, high-empathy founders and product leaders.
How can startups avoid the trap of staying too lean and underhiring while still maximizing AI-driven leverage?
He also forecasts that individuals who aggressively adopt AI tools across professions will be dramatically more capable than their peers, and that small, highly skilled, AI-leveraged teams will increasingly outperform large, traditional organizations.
What does Ng think are the most realistic near-term risks of agentic AI in production workflows, and how should teams mitigate them?
EVERY SPOKEN WORD
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome