​Founder Mode Episode 49 - The Future of AI-Built Software with Nima Keivan


The Future of AI-Built Software with Nima Keivan

Welcome back to Founder Mode!

In this episode, we sat down with Nima Keivan. He has built autonomous systems in both robotics and software, and now he is focused on one of the hardest problems in AI-built software: reliability.

We talked about autonomous coding, why code generation is no longer the main bottleneck, and what it takes to move from a demo to real production software.

This one was a strong reminder that writing code is only a small part of building something that lasts.

Let’s get into it.

1. Code Is Not the Hard Part Anymore

AI can now generate code very fast.

That part is no longer the real breakthrough.

As I said in the episode:

“We're seeing tools that generate code instantly, but code isn't the problem. It's context and sort of requirements.”

That is the real shift.

The hard part is knowing what to build, how it should behave, and how it fits into the messy reality of an actual business.

Code is getting cheaper.

Clarity is not.

2. The Autonomy Gap Is What Breaks Everything

Nima introduced a simple but important idea: the autonomy gap.

That is the gap between what the system can handle on its own and what reality still throws at it.

If the system cannot carry that gap, a human has to.

That means the human is still part of the workflow. They just got pushed into the failure path.

That is why so many AI demos look impressive but still fall apart in real use.

A workflow is not truly automated just because AI touched it.

It is automated when it works reliably enough that a person does not need to babysit it.

3. Requirements Matter More Than Ever

This episode made one thing very clear.

If you want durable AI software, you have to start with requirements.

Not guesses.

Not vibes.

Not just prompts.

You need to know:

  • What the workflow is
  • Who is involved
  • What systems it touches
  • What success looks like
  • What should happen when something goes wrong

Nima’s point was that software becomes much more durable when it is anchored to real requirements first.

That sounds obvious, but most teams still skip this step or rush through it.

AI will happily build the wrong thing if you are not clear.

4. Scenario Testing Beats Simple Testing

Another big takeaway was that normal testing is not enough.

For AI-built software, you need to test full scenarios.

That means asking:

  • What happens when this touches a live CRM
  • What happens if the outside system is slow
  • What happens if the data is incomplete
  • What happens if the workflow hits a weird edge case

This is where Nima’s robotics background really stood out.

In robotics, you do not get points for a nice demo.

It either works in the real environment or it does not.

Software is now entering that same phase.

The lesson is simple.

You do not just test the code.

You test the real world around the code.

5. The Best Builders Will Understand Systems, Not Just Syntax

This conversation was not really about replacing engineers.

It was about changing jobs.

A lot of repetitive implementation work will shrink.

But the need for people who can think clearly about systems will only grow.

That means the next wave of strong builders will be the people who can:

  • define requirements
  • understand workflows
  • design for edge cases
  • test real outcomes
  • connect technical work to business reality

That is why this episode felt bigger than just code generation.

It was really about how we build software that survives contact with real life.

Final Thoughts

This episode sharpened something I have been feeling for a while.

AI can help write software faster.

But faster does not mean better.

And generating code does not mean you solved the real problem.

The hard part is still the same in many ways.

You need to understand the system.

You need to define the requirements.

You need to test the messy parts.

And you need to close the gap between the demo and production.

That is where the real work is.

The teams that understand that are going to build things that last.

If this changed how you think about AI, software, or what production really means, share it with someone who is building.

🎧 Listen to Episode 49 here:

show
The Future of AI-Built Softw...
Mar 26 · Founder Mode
31:32
Spotify Logo
 
video preview

This podcast builds on the Founder Mode newsletter.

Let’s build.

-kevin

2810 N Church St #87205, Wilmington, DE 19802
Unsubscribe · Preferences

Founder Mode

Founder Mode is a weekly newsletter for builders—whether it’s startups, systems, or personal growth. It’s about finding your flow, balancing health, wealth, and productivity, and tackling challenges with focus and curiosity. Each week, you’ll gain actionable insights and fresh perspectives to help you think like a founder and build what matters most.

Read more from Founder Mode
Kevin shares how Pretty Good AI reveals hidden workflow assumptions, exposes broken rules, and forces teams to align processes for real AI success.

Rules Nobody Questions Welcome back to Founder Mode. There is something I have started to notice more and more as we build Pretty Good AI. Most businesses run on rules that nobody questions. Not because they are right. But because they have been there for a long time. They live in conversations, habits, and assumptions. They get passed down from one person to another. Over time, they harden into policy. Then AI shows up. And suddenly, those rules have to be written down. That is when things...

Kevin shares lessons from Pretty Good AI on why teams stick with broken systems, how sunk cost bias slows growth, and when to walk away to win faster.

The Sunk Cost Trap Welcome back to Founder Mode! There is a pattern I keep seeing that has nothing to do with AI models or technology. It has everything to do with people. Teams hold onto broken systems long after they stop working. Not because they believe in them. But because they already paid for them. At Pretty Good AI, we run into this all the time. The hardest part of implementation is not building something new. It is helping people let go of what is not working. And that is a much...

AI agents are changing how companies operate. Learn why founders must shift from building to system design and orchestration.

The End of Prompt Engineering with Dennis Pilarinos Welcome back to Founder Mode! We’re in the middle of a shift. Not a small one. Not another tool. Not another framework. A real shift in how companies get built. As I said in this episode: “It used to be build the product, hire the team, scale the ops. Now it's design the system, orchestrate the agents, and make the calls that can't be automated.” That’s the job now. And if you don’t see it yet, you will soon. AI Isn’t a Tool Anymore. It’s a...