Now Anyone Can Code: How AI Agents Can Build Your Whole App

Thanks to rapid development in LLM’s, we are now at the point where AI is able to follow prompts and generate code to build functional custom software. So how does the tech landscape change when the ability to code is democratized? In this episode of the Lightcone, the hosts speak with Amjad Masad, the CEO of Replit, an AI-powered software development and deployment platform, to see how coding power can be given to everyday users.
0:00 Intro
1:15 Making an app with Replit
6:19 Feel the AGI, personal software era
8:07 Having AI code the way humans do
9:51 You should still learn to code!
11:42 The underlying tech
17:19 The path to AGI
19:41 What users made with Replit
25:56 Challenges in resetting the org
33:29 Future plans
36:12 Outro

[chapterme]

Chapters (Powered by ChapterMe) -
00:00 - Coming up
00:47 - Intro
01:15 - Making an app with Replit
06:19 - Feel the AGI, personal software era
08:07 - Having AI code the way humans do
09:51 - You should still learn to code!
11:42 - The underlying tech
17:19 - The path to AGI
19:41 - What users made with Replit
25:56 - Challenges in resetting the org
33:29 - Future plans
36:12 - Outro

[zgidwitz]

I’ll believe this when YC accepts a non-technical confounding team using these agents to build their products.

[jsalsman]

As someone whose business's flagship app was built in Replit with help from LLMs, Amjad is way too optimistic. For anything more sophisticated than toy apps, you've got to have software architecting, system and database administering, and lots of software engineering expertise just to pose the right questions to the AI. We're getting there, but it will take at least a couple more years to get where Amjad describes as today.

[adelhishem]

As a software developer i find the title of this video surreal. IT IS JUST NOT TRUE THAT YOU CAN DEVELOP A SOFTWARE PRODUCT (EVEN A VERY SIMPLE ONE) AND DEPLOY IT WITH JUST AI.

[nastied]

I used Replit for the first time Yesterday and it Started spinning and got stock in some loop so had to stop it and redo some times o but Eventually it did Create my first LLM API integration and deployed it which made me really happy

[rubncarmona]

who is this content for? is YC accepting non-technical founders that make their apps with these tools now?

[cryptonative]

If every can code, no one can build a business out of it.

ps: not everyone can code

[wookiee_]

News flash: Everyone has been able to code for a while now, even before LLMs. There's been a nearly limitless amount of resources to learn coding, many of them for free or at a negligible cost. Not many people do because it takes determination and work to become decent at it. Everyone is looking at LLMs as a free lunch to bypass that work, but code is still being produced by it, and it will still need to be understood if you ever want to create any serious apps.

[moderncontemplative]

This is an excellent roundtable discussion! Note that one of the speakers is talking about the difference between effective AGI and true AGI. It has become quite clear to many of us AI aficionados that effective AGI is just “a few thousand days away”

[morten-punnerud_engelstad]

This saves me like 5 second compared to just using Cursor AI that have been out for several months. The drawback is lock-in to replit, compared to Cursor AI that is just a good IDE over Claude Sonnet 3.5 and ChatGPT o1

[haroldpierre]

I haven't coded in almost 3 decades. Recently, I started using LLMs to write code. It's been fun, I mean REALLY fun because I remembered the days of coding freestyle. However, it has been frustrating too. There is no doubt that the more you know about the language, the faster it goes. I've spent 10+ hours trying to get something to work because of the limitations of LLMs and of course me. I have to prompt the LLMs correctly to reduce the development time and the coding errors. Also, I have to use multiple LLMs to accomplish the task. In my experience, the coding quality is in this order GPT o1 > Claude 3.5 > Gemini > Llama 3.1. If I get stuck in one, I switch to the other to fix the problem. Regardless, this beats learning a language when I just don't have time.

[SteveKowarsky]

I have aged out as a coder, but system architecture and understanding of data is an enduring skill. I would love to build a real application with AI, something with a database, user classes, interactive screens where users can add data, interact with data, and report data. One after another, the "application" examples in these videos are just wrappers around an AI conversation. That is not an application. It's a toy.

[EleshChannel]

Everyday people say everyone can now code but I am yet to see a single app that a non-developer has built with only LLM. Instead, my workload on upwork keeps climbing. Are these clients not aware they can build their apps with just LLM?😅

[RichardWatson]

Commenters, the direction is more important than the snapshot. I used it on day 1 and was impressed but it wasn’t fully baked. But it’s a work in progress and this is going to learn faster than most people will. Make your decisions with the assumption that these tools get better every year. They might not out-code a top coder, bit a top coder with it is better than one without. And many businesses don’t have top coders.

[adityakabra]

Replit Agent + Cursor + tldraw - Combined into one platform will be super useful

[balqaasem]

I’ll be appearing on this channel in a couple of weeks as a YC founder Insha’Allah

[Andy_B.]

Hi,

please do another podcast about Vertical Ai Agents -

Last one was great!

You could showcase some other SAAS start ups, that are using AI to leverage their business model.

Thank you!

[george_davituri]

AI driven agent-assistant will make engineers life less stressful and also help to spot a bug in a codebase in minutes instead of scrolling through hours

[solvriksh]

A 15 USD/month for Replit Core for using an AGI precursor, it doesn't seem a bad deal when you can play and build a ton. Team AI Aegnt Revolution 2025!

[saltcurrent]

Should have started earlier this year, exactly same thing in mind