Are we building AI coding assistants wrong?

#1

Post by anaempromptu » Mon, Jul 28, 2025, 6:20 PM UTC

The Jason Lemkin/Replit incident this week got me thinking about the fundamental problems with how we're approaching AI coding assistants. We've all seen the demos - natural language to working code, conversational debugging, "just describe what you want." But then you hit prod and everything breaks down.

The core technical challenges I keep seeing:

- Context management at scale - These systems work great for isolated tasks but struggle to maintain coherent state across complex, multi-file projects. How do you handle context that spans thousands of lines across dozens of files?

- The safety/capability tradeoff - More powerful tools can do more damage. Replit promised not to touch production, then deleted a database anyway. How do you build guardrails that actually work without neutering the tool?

- Conversational interfaces for complex systems - Natural language is ambiguous. Code is precise. Are we trying to solve the wrong interface problem?

- The production gap - Every AI coding tool I've tested works beautifully in demos and falls apart with real codebases, real data, real edge cases. Why is this gap so persistent?

I'm genuinely curious - has anyone built AI apps with tools that actually work reliably in prod?

Re: Are we building AI coding assistants wrong?

#2

Post by moomoo11 » Mon, Jul 28, 2025, 6:31 PM UTC

As with anything, complexity becomes the limiting factor.

Anyone can drive. Some even have cars that drive themselves. Very few people can drive Formula 1.

They're also totally different - getting from A to B vs competing on a track.

I think its easier than ever for total n00bs to build software. Its on them to use the same AI that did everything for them to figure out how to scale their software add features, maintainability, testing, and organizational efficiency.

You put the same AI workflows into the hands of someone that was a exceeds expectations tier engineer or someone highly motivated like that, and they'll push it even farther on the most complex systems.

Just need to know the limits, the opportunities, and go wild.

Re: Are we building AI coding assistants wrong?

#3

Post by lgl » Mon, Jul 28, 2025, 6:34 PM UTC

Isn't the main reason the split between dev, stage and prod exactly to weed out these sorts of issues?

An incompetent/distrated/etc dev can do just as much damage as an AI on these cases.

Just dont blindly give prod access to any entity.

Re: Are we building AI coding assistants wrong?

#4

Post by anaempromptu » Tue, Jul 29, 2025, 1:05 AM UTC

lgl wrote: post #3

Isn't the main reason the split between dev, stage and prod exactly to weed out these sorts of issues? An incompetent/distrated/etc dev can do just as much damage as an AI on these cases. Just dont blindly give prod access to any entity.

most AI coding tools don't have this option or does it mention it as a "must have"