Andrew Ng: Building Faster with AI
Welcome! This is a detailed, chronological summary and explanation of Andrew Ng's talk, "Building Faster with AI," including all the key lessons, insights, and memorable quotes. The talk is packed with practical advice for startups, founders, and anyone interested in leveraging AI to build products and businesses faster and more effectively. Let's dive in! 🚀
1. Introduction: Lessons from the Trenches
Andrew Ng opens by sharing his hands-on experience at AI Fund, a venture studio that builds about one startup per month. He emphasizes:
"We're in there writing code, talking about customers, designing features, determining pricing… not just watching others build startups, but actually being in the weeds, building startups with entrepreneurs."
He sets the stage for the talk's main theme: SPEED.
2. Why Speed Matters in Startups
Andrew highlights a core belief:
"A strong predictor for a startup's odds of success is execution speed."
He admires entrepreneurs who can "just do things really quickly" and notes that new AI technology is enabling startups to go much faster.
He promises to share best practices that are "frankly changing every two to three months" to help others achieve this speed.
3. Where Are the Opportunities? The AI Stack
Andrew explains the AI stack:
- Semiconductor companies (bottom)
- Cloud/hyperscalers
- AI foundation model companies
- Application layer (top)
He stresses:
"Almost by definition, the biggest opportunities have to be at the application layer… we actually need the applications to generate even more revenue so that they can afford to pay the foundation, cloud, and semiconductor technology layers."
He laments that media and social media rarely talk about the application layer, but "for those of you building startups, almost by definition, the biggest opportunities have to be there."
4. The Rise of Agentic AI
Andrew identifies the most important tech trend in AI:
"The rise of agentic AI."
He explains that a year and a half ago, he was advocating for AI agents, but the term got overused by marketers and lost some meaning. He clarifies:
"From a technical perspective, agentic AI is exciting and important and opens up a lot more startup opportunities."
What is Agentic AI?
Traditionally, we prompt LLMs to generate output in a linear, one-shot way:
"It's as if you're going to a human or an AI and asking it to please type out an essay for you by writing from the first word to the last word all in one go without ever using backspace."
But humans don't write this way, and neither does AI at its best.
With agentic workflows, you can:
- Ask AI to write an outline
- Do web research
- Draft, critique, and revise
- Iterate multiple times
"It is slower but it delivers a much better work product."
He shares that for complex tasks (compliance, medical diagnosis, legal reasoning), agentic workflows are often the difference between success and failure.
A New Layer: Agentic Orchestration
A new agentic orchestration layer has emerged, making it even easier to build applications by coordinating calls to underlying tech.
"The basic conclusion that the application layer has to be the most valuable layer of the stack still holds true."
5. Best Practices for Moving Faster
A. Work on Concrete Ideas
Andrew is adamant:
"At AI Fund, we only focus on working on concrete ideas."
A concrete product idea is:
"Specified in enough detail that an engineer can go and build it."
He contrasts:
- Vague: "Let's use AI to optimize healthcare assets." (Too many interpretations, not actionable)
- Concrete: "Let's write software to let hospitals, let patients book MR machine slots online to optimize usage."
"Concreteness buys you speed."
He warns:
"The deceptive thing for a lot of entrepreneurs is the vague ideas tend to get a lot of kudos… but when you're vague, you're almost always right. When you're concrete, you may be right or wrong. Either way is fine. We can discover that much more fast, which is what's important for startups."
B. The Power of the Expert Gut
After thinking deeply about a problem, subject matter experts can make rapid, high-quality decisions:
"The gut, which is an instantaneous decision, can be actually a surprisingly good proxy."
He notes:
"Getting data for a lot of startups is actually a slow mechanism for making decisions. A subject matter expert with a good gut is often a much better mechanism for making a speedy decision."
C. Focus on One Hypothesis at a Time
"At any moment in time, you're pursuing one very clear hypothesis… pick one, go for it, and if data tells you to lose faith in that idea, that's actually totally fine. Just pivot on a dime to pursue a totally different concrete idea."
He adds:
"If every piece of new data causes you to pivot, it probably means you're starting off from too weak a base of knowledge."
6. The Build-Feedback Loop: Engineering at Warp Speed
Andrew describes the build-feedback loop:
- Build software (engineering)
- Get user feedback (product management)
- Iterate (repeat)
He observes:
"With AI coding assistance… rapid engineering is becoming possible in a way that just was not possible before. The speed of engineering is going up rapidly and the cost of engineering is also going down rapidly."
Prototyping: Move Fast, Be Responsible
- Quick and dirty prototypes: "We're easily 10 times faster, maybe much more than 10 times faster."
- Production code: "Maybe we're 30 to 50% faster with AI systems."
He admits:
"I routinely go to my team and say, 'Go ahead, write insecure code.' Because if this software is only going to run on your laptop and you don't plan to maliciously hack your own laptop, it's fine to have insecure code… but of course, after it seems to be working, please do make it secure before you ship it to someone else."
He encourages:
"Move fast and be responsible."
AI Coding Tools: The New Generation
He traces the evolution:
- Code autocomplete (GitHub Copilot)
- AI-enabled IDEs (Cursor, Windserve)
- Highly agentic coding assistants (Claude 3, etc.)
"If you're even half a generation or one generation behind, actually makes a big difference compared to if you're on top of the latest tools."
He notes:
"We've completely rebuilt a codebase three times in the last month… because the cost of doing that has plummeted."
Two-Way Doors vs. One-Way Doors
"Choosing the software architecture of your tech stack used to be a one-way door… now, my team will more often build on a certain tech stack, a week later, change your mind, throw the code base away and redo it from scratch on a new tech stack."
7. Empowering Everyone to Build with AI
Andrew pushes back against the idea that AI will make coding obsolete:
"I think we'll look back on this as some of the worst career advice ever given… as better tools make software engineering easier, more people should do it, not fewer."
He shares:
"On my team, my CFO, my head of talent, my recruiters, my front desk person—all of them know how to code. And I actually see all of them performing better at all of their job functions because they can code."
He gives a vivid example:
"One of my team members knew art history and so he could prompt Midjourney with the genre, the palette, the artistic inspiration… had a very good control over the images he generated. Whereas in contrast, I don't know art history… I could never have the control that my collaborators could."
Key lesson:
"One of the most important skills of the future is the ability to tell a computer exactly what you want, so they'll do it for you."
8. The New Bottleneck: Product Management & Feedback
With engineering so fast, product management and user feedback are now the bottleneck.
He notes:
"I'm seeing this ratio shift… one of my teams came to me and for the first time… proposed having twice as many PMs as engineers."
He suggests:
"PMs that can code or engineers with some product instincts often end up doing better."
Tactics for Rapid Product Feedback
He shares a portfolio of tactics, from fastest to slowest:
- Look at the product yourself and go by your gut (if you're an expert)
- Ask three friends or teammates
- Ask three to ten strangers (e.g., in a coffee shop or hotel lobby)
- Send prototypes to 100 testers
- A/B testing (slowest)
He emphasizes:
"Contrary to what many people think, A/B testing is now one of the slowest tactics in my menu."
He also advises:
"When I A/B test something, I don't just use the result to pick product A or B. My team will often sit down and look carefully at the data to hone our instincts… to improve the quality of our guts on how to make product decisions faster."
9. Understanding AI = More Speed
Andrew argues that understanding AI gives you a real edge:
"Teams that actually get it, that understand AI, do have an advantage over teams that don't."
He explains:
"If you make the right technical decision, you can solve the problem in a couple days. Make the wrong technical decision, you could chase a blind alley for three months."
He illustrates:
"If you flip the wrong bit, you're not twice as slow. You spend like 10 times longer chasing a blind alley."
The Power of Building Blocks
He describes the explosion of GenAI building blocks (prompting, evals, guardrails, RAG, embeddings, fine-tuning, etc.):
"Knowing all these wonderful building blocks lets you combine them in much richer combination… the number of things you can combine grows kind of combinatorially or grows exponentially."
He likens it to Lego bricks:
"If you have a basic white building block, you can build some cool stuff. Add a black Lego brick, you can build something more interesting… get more building blocks, and very rapidly the number of things you can combine grows exponentially."
10. Summary: What Matters Most
Andrew wraps up:
"There are many things that matter for startups, not just speed. But… the management team's ability to execute at speed is highly correlated with its odds of success."
Key takeaways:
- Work on concrete ideas
- Speed and quality of decisions both matter, but speed absolutely matters
- Rapid engineering with AI coding assistance shifts the bottleneck to product feedback
- Have a portfolio of tactics for rapid feedback
- Stay on top of technology to buy speed
He closes with:
"If you haven't learned to go to a coffee shop and talk to strangers… just be respectful. That's actually a very valuable skill for entrepreneurs to have."
11. Q&A Highlights
Q: Should we focus on building tools or learning to use them?
"The people that are most powerful are the people that can make computers do exactly what you want it to do… people that know how to use AI to get computers to do what you want will be much more powerful than people that don't."
Q: The Future of Compute and AI Hype
Andrew debunks several overhyped narratives:
- "AI is so powerful, we might accidentally lead to human extinction. That's just ridiculous."
- "AI is so powerful, soon no one will even have a job anymore. Just not true."
- "AI needs so much electricity, only nuclear power is good enough for that. That's just not true."
He warns:
"Some of these hype narratives have been amplified… a distortion of what actually will be done."
Q: AI Safety vs. Responsible AI
"Safety is not a function of technology, it's a function of how we apply it… AI is neither safe nor unsafe. It is how you apply it that makes it safe or unsafe."
He prefers the term "responsible AI" over "AI safety."
Q: How to Compete When Everything Can Be Replicated?
"The number one thing I worry about is: are you building a product that users love? … Focus on building a product that people want, that people love. And then figure out the rest of it along the way."
He notes that moats are often overhyped and usually evolve over time.
Q: Agentic AI, Token Costs, and Building Block Flexibility
"My most common advice to developers is: to a first approximation, just don't worry about how much tokens cost. Only a small number of startups are lucky enough to have users use so much of your product that the cost of tokens becomes a problem."
He recommends architecting software to make switching between different building block providers easy.
Q: AI in Education
Andrew sees hyperpersonalization as the future, but the end state is not yet clear:
"Everyone feels like a change is coming in edtech but I don't think the disruption is here yet… I see a lot of experimentation, but the final end state is still not clear."
Q: Societal Downsides and Moving Fast Responsibly
"Look in your heart and if fundamentally what you're building, if you don't think it'll make people at large better off, don't do it."
He shares that AI Fund has killed projects on ethical grounds.
He also urges:
"Trying to bring everyone with us to make sure everyone is empowered to build with AI. That'll be an important part of what all of us do, I think."
Q: Educating the Public about AI
Andrew warns about the danger of gatekeepers:
"One of the dangers to inequality as well is if these regulatory proposals succeed and end up siphoning regulations, leaving us with a small number of gatekeepers where everyone needs the permission of a small number of companies to fine-tune the model or prompt in a certain way. That's what will stifle innovation and prevent the diffusion of this information."
He stresses the importance of open source and diffusing knowledge.
Final Thoughts
Andrew's talk is a masterclass in how to build faster with AI. The key lessons:
- Speed is a superpower—enabled by AI, but only if you work on concrete ideas and keep learning.
- Empower everyone to build—coding is for all, not just engineers.
- Stay on top of the latest tools and building blocks—the combinatorial possibilities are exploding.
- Move fast and be responsible—ethical considerations matter as much as technical ones.
- Don't get distracted by hype—focus on building things people love, and keep the door open for innovation.
"The people that know how to use AI to get computers to do what you want will be much more powerful than people that don't."
Thanks for reading! If you want to build with AI, now is the time—and Andrew Ng's advice is a fantastic place to start. 💡✨
Need a summary like this?
Get instant summaries with Harvest