TL;DR:

• A non-coder can now build complex AI tools in hours.

• Traditional SaaS companies are fucked if they only treat AI as a feature.

• Software is becoming disposable and hyper-personalized.

• Agent chains will be the new normal for building tools.

• The "app gap" is dead – if it doesn't exist, you can build it instantly.

We're about to witness the biggest displacement in software history. For some, this means massive opportunity. For others – particularly horizontal SaaS businesses that fail to adapt – it's going to be devastating.

Most won't adapt. They can't. They're internal workflows, processes and human incentives will prevent them from adapting.

This isn't theoretical bullshit. I just experienced it firsthand by building something that completely changed my perspective on AI and software.

In the context of SaaS, I have mostly considered AI as value-add. Nice-to-have augmentation. Then I built my own AI agent, and now I'm not so sure anymore.

It started with a simple need. I wanted better ways to surface my past notes and article highlights. Something that could generate creative prompts for my writing and help condense my content for X and Farcaster.

Couldn't find a tool that did exactly that. So I said fuck it, I'll build one.

The initial version took less than an hour. Here's the BRG Agent with a simple UI:

Some context: My coding skills are basic. Made some apps and websites 10 years ago, but what I built here is way beyond what I could normally do. And it took less than three hours...

From my previous research into AI agents, I knew I wanted one with personality, trained on my content. Using Replit (an AI-augmented code editor), I set up the initial structure using ElizaOS agent framework and Claude as the LLM.

The first version was simple:

- Give it a prompt, get a thought back

- Autonomous thoughts every ten minutes

That last part is wild. Keep the browser tab open, and it just... creates. Like magic. Some of it is crap, some it sparks inspiration.

Playing around with it gave me more ideas. Two hours later, I had built:

- Readwise integration pulling my reading highlights

- Thought generation based on highlights

- RSS feed subscription to my newsletter

- Favorite marking system

- Topics, tagging, and search

- Preview styling

- LinkedIn post generator (yeah, I post there sometimes)

- Feedback mechanism to improve future generations

Is it perfect? No. It produces a mix of interesting stuff and garbage. The interface is a bit janky.

But here's the thing: It's a fully functioning AI agent, with an interface, specifically trained on my content, integrated with multiple external APIs. Built by a non-coder in 3 hours.

In my past life I ran a company with multiple software engineering teams. I know what kind of time would have gone into creating something like this following 'best practices', putting an engineer, a pm and a designer together to work on it, etc.

(Spoiler: it would take more than three hours.)

Some more screenshots:

What This Means for Software

Building software has been a gold mine, because of skill and capex hurdles. I see a path where software in the future becomes ephemeral and hyper-personalized. Think of agents as the new interface – you communicate what you need, and they generate the exact tool for that moment.

That's exactly what happened here. I chatted with an AI agent that created and customized an app based on my specs and feedback. Yeah, it's meta that it's an agent programming an agent, but this chain-of-agents pattern will be normal soon.

Right now it's raw – I see the code, use programming tools, handle deployment specs. But all that could be (will be) abstracted away.

The implications are massive: The barrier to "there's an app for that" is basically gone.

Some legacy SaaS products think slapping a chatbot in their sidebar and calling it "AI-enabled" is enough. It's not.

They'll be outcompeted by two forces:

1. People rolling their own custom software

2. Living, evolving applications that constantly iterate based on user needs

The future of software isn't just AI-enhanced. It's AI-native. And it's coming faster than most people realize.

There's a substrate of software, the vertical SaaS, niche tools built for specific industries. In some industries, these products are still onprem, haven't yet made it to the cloud. In others, it's barely entered the cloud. In most of these cases, there's an important moat: The complexity of building tool depth and integrating with other legacy systems adds overhead that makes it difficult and expensive for challengers to enter the market. I think this challenges are about to collapse.

And it's going to displace many incumbents that managed to hold on through the "cloud-phase", but they will be absolutely washed out by cloud+ai enabled competition.

All of this almost makes me wanna find a good industry and build a new venture.

Hmm...

As always, reach out to me on X or Farcaster if you have feedback or are working on something cool you want to discuss.