#18 - The man, the myth, the agentic workflow.
My take on Andrew Ng's latest chat
The other day, I sat down to watch the LangChain team’s new fireside chat with Andrew Ng, and it sent me down a memory lane I haven't walked in a while.
Years ago, when I was a young Mechanical Engineering student drowning in thermodynamics and fluid dynamics, a friend showed me this thing called Coursera. On it was a course that would supposedly teach you "Machine Learning," taught by a professor from Stanford. The course used MATLAB, and while I was intrigued, my university workload was just too intense. I bookmarked it and moved on.
Fast forward to the COVID-19 lockdown. The world had stopped, but I finally had time. I had since learned Python and remembered that professor. I found his new “Deep Learning Specialization” on Coursera and dove in headfirst.
To say I was obsessed would be an understatement. I took over 200 pages of notes across the five courses, filling notebooks with everything from the fundamentals of neural networks to the intricacies of sequence models. To really solidify it, I supplemented the course by watching the full Stanford CS230 lectures on YouTube, taught by the incredible Kian Katanforoosh. In my head, I created this little fantasy: I was finally a Stanford student, with Andrew Ng as my professor and Kian as my TA.
I know - I'm a Nerd :))
So, when Andrew sits down to talk about the current state of AI Agents, I listen closely. It’s not just an industry leader speaking; it’s the professor who helped me find my place in this field.
Here are my biggest takeaways from his recent conversation:
The chat with LangChain's Harrison Chase wasn't about far-future AGI or flashy demos. It was a grounded, practical look at what it actually takes to build useful "agentic systems" today. They covered the messy, in-the-trenches work: the importance of evals, the underrated potential of certain technologies, and the skills that truly matter for builders.
The Big Debates
In a landscape where some tech leaders are predicting the end of coding, Andrew makes a powerful, historically-grounded case for the opposite. He argued that advising people not to learn to code might be "some of the worst career advice ever given."
His logic is simple: every time programming has gotten easier—from punch cards to assembly language to higher-level languages like Python—more people have learned to code, not fewer. AI coding assistants are just the next step in that evolution. They make it easier to tell a computer exactly what you want it to do, a skill he believes is critical for everyone, not just software engineers. It’s a compelling argument that frames coding not as a chore to be automated away, but as the fundamental language for human-computer collaboration.
What's Next on Andrew's Radar
Two things he mentioned stood out as clear signals of where he sees massive, near-term opportunity.
First, he dropped a fascinating hint about his own work. When asked what AI coding tools he uses, he mentioned Cursor and Windsurf, and then slyly added, "So we're working on some things that we've not yet announced." Reading between the lines, it sounds like AI Fund might be building a new tool in the AI-assisted coding space. Keep an eye on that. 👀
Second, he called the Voice Stack one of the most "underrated" areas in AI right now. He sees huge demand from large enterprises and notes that the developer attention on voice is still tiny compared to its commercial importance. He pointed out that speaking is a much lower-friction interface for users than typing into a text box—it’s less intimidating and allows for a more natural flow of ideas. For anyone looking for an uncrowded space to build, he just put up a giant neon sign pointing toward voice applications.
"Speaking of Voice Agents..."
Andrew's focus on voice really struck a chord with me, and it’s a space I’m exploring deeply myself. His insights on user friction and latency are spot-on, and they mirror some of the challenges and opportunities I've been tackling in my own projects. You might be hearing some interesting news from me on that front soon. 😉
Some courses I’m taking this week:
P.S. — Get Ready to Build
Speaking of practical skills, I’m doubling down on what this newsletter is all about: hands-on learning. I'll be publishing a lot more hands-on tutorials on the blog, where we'll explore the tools and techniques needed to build useful AI Agents together.
As some of you know from my last issue, I’ve been diving deep into GTM (Go-to-Market) Engineering, and I’m amazed every single day at how much we can automate and simplify our own workflows.
So, in our blog posts every Tuesday and Thursday, we're going to start building the very agents I'm creating to solve my own daily challenges.
I even have an idea for the first one: I’m building an agent to help me stop staring at a blank page for three hours whenever I need to write.
Sound familiar? Stay tuned.
Thank you for the value! I knew about Cursor but I did not know about Windsurf.