Building Your Career in AI: Real Talk from the Trenches

Source: Dev.to

Inspired by insights from Andrew Ng and Lawrence Moroney’s career‑advice talk

“Look, the AI space is absolutely wild right now. As someone who’s been deep in developer advocacy, cross‑platform development, and best practices in software engineering, I’m watching this transformation happen in real‑time, and honestly? It’s pretty exciting, but also chaotic as hell.”

What’s changed in the last year‑or‑two?

• Powerful tools – Large language models (LLMs) that truly understand context, smoother workflows, voice AI that sounds more human, and coding assistants that genuinely boost productivity.
• New AI building blocks – Large language models, RAG‑augmented workflows, voice AI, and deep‑learning frameworks that were either very difficult or didn’t exist a couple of years ago.

Andrew’s point: “We now have AI building blocks that were very difficult or did not exist a year or two ago: large language models, RAG‑augmented workflows, voice AI, and deep learning frameworks.”

I’ve been experimenting with everything from AI coding assistants to custom MCP servers, and the difference from six months ago is night‑and‑day.

🚀 Building stuff got stupid fast

• Recent prototypes:
  • A video‑watcher persona system that generates custom advertisement videos on the fly.
  • Multi‑agent AI content generators.
  • A “simple” vibe‑coded Conclave RPG game.

Andrew’s observation: “With AI coding, the speed with which you can get software written is much faster than ever before.”

The bottleneck isn’t typing code any more – it’s knowing what to build and how to architect it properly.

💡 Inspired by Andrew Ng’s “Product Management Bottleneck” concept

What I’ve seen on React‑Native TV projects & AI automation pipelines

• The hardest part isn’t implementing features; it’s deciding what to build, writing clear specs, and understanding user needs.

Andrew’s key insight: “When it is increasingly easy to go from a clearly written software spec to a piece of code, the bottleneck increasingly is deciding what to build or writing that clear spec for what you actually want to build.”

• Team dynamics shift – Traditional engineer‑to‑PM ratios (e.g., 8 engineers : 1 PM) are compressing toward 1 : 1 or even 2 : 1.

Andrew noted: “I’m seeing the engineer‑to‑PM ratio trending downward—maybe even two to one or one to one… some teams I work with have a head‑count of 1 PM to 1 engineer.”

Takeaway for developers: If you can think like a product person and ship code, you’re gold.

Andrew emphasized: “Engineers that can also shape product can move really fast… the subset of engineers that learn to talk to users, get feedback, develop deep empathy for users so that they can make decisions about what to build, those engineers are the fastest moving people.”

💡 Andrew’s advice on staying at the frontier of tools

• Tooling changes every 3–6 months in this space.
• Being one generation behind means working twice as hard for half the output.

Andrew’s experience: “If you ask me every three months what my personal favorite coding tool is, it actually probably changes every six months, maybe even every three months. Being half a generation behind in these tools means being frankly quite a bit less productive.”

Practical tip:

• Set aside weekly time to experiment with new tools. I literally block calendar time for this.

💡 Andrew & Lawrence’s emphasis on building and showing work

• Don’t just play around—build things that solve real problems.
• My recent work: automated video pipelines, MCP servers for app KPIs, and content‑automation systems. These aren’t toys; they solve actual workflow pain points.

Andrew’s encouragement: “Just go and build stuff right… your opportunity to build things and I think showing them to others is greater than ever before.”

Lawrence’s advice: “Don’t let your output be for the job you have; let your output be for the job you want.”

Portfolio tip: Show that you can ship production code or tools people actually use, not just prototypes. Conference talks are great, but shipped code gets attention.

🔥 Real‑world example

Andrej Karpathy recently shared how he used Claude Code to reverse‑engineer his entire Lutron home‑automation system in a single session. The AI:

1. Discovered controllers on his network.
2. Found open ports.
3. Pulled PDF documentation.
4. Connected to devices.
5. Tested by turning kitchen lights on/off.

He’s now “vibe‑coding” a master command center to replace the janky official app. This is exactly what building real things looks like—taking AI tools and solving actual problems you have.

💡 Andrew’s story about the student assigned to Java backend payment processing

• Lesson: The brand on your business card matters less than the people you learn from.

Andrew’s warning: “He told a story about a Stanford student who joined a company with a ‘hot AI brand’ that refused to tell them which team they’d join. After signing, they were assigned to backend Java payment processing instead of AI work and left after a year of frustration.”

What to look for:

• Engineers who are genuinely curious and stay current.
• Teams that ship regularly, not just talk about shipping.
• Environments where you can experiment without being shut down immediately.

Andrew’s emphasis: “One of the strongest predictors for your speed of learning and for your level of success is the people you surround yourself with… instead of working for the company with the hottest brand, sometimes if you find a really good team with hardworking, knowledgeable, smart people trying to do good with AI, even if the company logo isn’t as hot, you often learn faster.”

Bottom line: Access to emerging tech and working with smart people is invaluable. That’s what moves your career forward.

💡 Lawrence’s “Business Focus” Pillar

The best opportunities come when you understand the why behind what you’re building. Whether it’s optimizing a TV interface for 10‑foot viewing or designing an AI workflow for content creation, knowing the business context makes you infinitely more valuable.

Lawrence’s framework:

“Business focus is non‑negotiable… Everything is geared towards production. Everything is biased towards production… the bottom line is that the bottom line is the bottom line.”

💡 Lawrence Moroney’s “Bifurcation of AI” Prediction

You’ve got two paths forming:

Lawrence’s insight: “Over the next five years there’s going to be a bifurcation… Big AI will be what we see today with the large language models getting bigger in the desire to drive towards AGI… The other side is self‑hostable models that are exploding onto the landscape.”

Both are valid. I work across both—using cloud LLMs for content generation while also thinking about edge processing for TV platforms. Pick what aligns with your interests and the problems you want to solve.

💡 Lawrence’s framework on “Vibe Coding” & technical debt

AI systems can accumulate technical debt fast. I learned this the hard way building automation pipelines. Think about technical debt like financial debt.

“Think about debt the way you normally would, right? Buying a house… you end up paying back the bank about a million dollars on half a million owned. That is probably a good debt to take on… A bad debt would be an impulse purchase on a high‑interest credit card.”

Every time you generate code with AI, ask yourself: Is this worth the technical debt I’m taking on?

Framework for avoiding bad technical debt

1. Clear objectives met – You knew what you needed to build; you didn’t just spin out code randomly.
2. Business value delivered – How’s this helping the business? “It’s really cool” isn’t enough.
3. Human understanding – Can your team understand and maintain this code?

Balance quick prototyping (which AI enables) with sustainable architecture. Document your work—your future self (and your team) will thank you.

“Code is cheap now in the age of generated code. Finished code, engineered code is not cheap.” – Lawrence

💡 Lawrence’s discussion on the “Anatomy of Hype”

The AI space is drowning in hype, and you need to learn to filter signal from noise.

Key insight:

“The currency of social media is engagement. Accuracy is not the currency of social media… If you are the kind of person who can filter the signal from the noise and then encourage others around the signal and not the noise, that puts you at a huge advantage.”

When someone says, “We need to implement agents” or “We need AI,” your first question should be “Why?”

Story: A European company CEO asked Lawrence to implement agents because “everybody’s telling me I’ll save business costs.” Lawrence kept asking why until they uncovered the real need: making salespeople more efficient by reducing research time from 80 % to 20 % of their work.

Learn to make things as mundane as possible to truly understand them. Strip away the magic and hype to see what’s actually happening under the hood.

💡 Lawrence’s advice on avoiding being a “One‑Trick Pony”

I’m not just an AI person; I work on TV platforms, React Native, developer advocacy, content creation, friction logging of new SDKs and tools. This breadth makes you resilient and more interesting to work with.

“Don’t be that one‑trick pony who only knows how to do one thing. I’ve worked with brilliant people who are fantastic at coding a particular API or framework and then the industry moved on and they got left behind.” – Lawrence

💡 Lawrence’s emphasis: Everything Is About Production Now

Move past proofs of concept. Learn deployment, monitoring, scaling. The bar has moved from “Can you build cool demos?” to “Can you deliver business value?”

“What’s it actually like working in AI right now? As recently as two or three years ago, working in AI was ‘if you can do a thing, you’re great.’ Unfortunately, that’s not the case anymore. Today you’ll see the P‑word: production.” – Lawrence

💡 Andrew & Lawrence on showing your work

Share what you’re learning. Write posts on dev.to, speak at conferences, create GitHub repos. The community feedback loop is invaluable, and it opens doors you didn’t know existed.

Example: When Lawrence interviewed at Google, instead of answering random questions he showed code he’d built—a Java application running in Google Cloud for predicting stock prices.

“My entire interview loop was them asking me about my code… It gave me the power to communicate about things that I knew.” – Lawrence

💡 Lawrence’s breakdown of what makes something “agentic”

If you’re going to work with AI agents, understand the four‑step pattern:

1. Understand intent – Use LLMs to grasp what needs to be done.
2. Planning – Declare available tools and create a plan.
3. Execution – Use the tools to get results.
4. Reflection – Review results against intent, iterate if needed.

This isn’t just buzzword compliance—it’s engineering discipline applied to AI systems.

💡 Lawrence’s discussion on the AI bubble & industry maturation

There’s probably a bubble coming. Lawrence laid out the anatomy:

1. Hype at the top.
2. Massive VC investment drying up.
3. Unrealistic valuations.
4. Me‑too products everywhere.
5. A small kernel of real value at the bottom.

Lesson from the dot‑com bubble:

“Amazon, Google… they did it right. They understood the fundamentals of what it was to build a .com. They understood the fundamentals of what it was to build a business on .com. And when the bubble of hype burst, they didn’t…” – Lawrence

Survivors will be those who…

• Focus on fundamentals
• Build real solutions