sia.hackernoon.com

Back in January, I gave myself the challenge to make a fully AI-automated business by the end of 2025. What better way to get up to speed with AI, I thought. To be honest? It’s been painful.

By August 2025, I have 2 ‘agents’ which can prospect, build, & deploy web directories in 30 minutes with 3 human approval asks.

Lately, though, I’ve had this sinking feeling that has hit an existential crescendo.

How It Started

In January, I merrily started testing AI tooling out, doing tool reviews, and gearing up to tackle this challenge.

I knew my way around the big AI APIs already and had written plenty of automation code. I expected the ‘AI agent’ frameworks to be fairly mature - I believed the hype.

This was the first revelation:

‘AI’ is both amazing and hollow hype.

Reality Check: The AI Hype Train

There are so many voices shouting ‘AI, get your AI here’ right now, it’s hard to hear yourself think.

For every AI tool I test, which is really, really good, 4 or 5 or 10 fall short of the mark. Many don’t even work.

This should feel like entrepreneurial gold prospecting, but at times, I wonder if it’s more like Squid Games.

How it’s Going: Dreams and Duct Tape

Avg directory build time: 27-36 mins; human touches: 3; failure rate on initial listings: ~40%

I started out with a simple business model, which I figured shouldn’t be that hard to automate with AI: Web directories. I thought that if directories didn’t work, I could re-use a lot of what I learn and apply it to basically any other web-fronted company.

But we’re in August already. The SEO ground underneath me feels sketchy. I’ve only just started work on marketing automations.

I tell myself, ‘you took on this crazy challenge’ and remember that I did it as much to learn, and share, as to make blooming AI directories.

But sustaining that crazy challenge is hard sometimes.

My first two “agents” work. They can prospect a niche via keyword research, they can register a domain, change name servers, generate logos, a color palette, and actually build a passable directory site.

That all sounds good, and I guess in principle it is.

But it doesn’t feel like AI.

In fact, even though I’m using AI (LLMs) throughout, in truth, these are automated workflows and micro-agents. They still read like “a bunch of nested logic wired together, which regularly use LLMs to fill in blanks.”

At the start of this challenge, I dreamed of truly autonomous AI agents; of focusing on the deep philosophy of the business. Coupling together modern tools and slowly honed prompts of ultimate power.

That’s not where it’s at.

AI Entrepreneurship: Opportunity vs Uncertainty

Currently, bits and pieces of the puzzle work:

LLMs are getting better all the time. New models are undeniably more capable. This means we can vibecode internal tools and small marketing assets.
Automation platforms are getting more solid. We can automate many business processes and enhance ops with LLMs.

AI Agents are on the cusp of being ‘real’ - AI 2027 (scenario forecast by Kokotajlo et al.) continues to ring true in that agents are currently ‘unreliable’, but we can already use them to do simple tasks.

As software entrepreneurs, we can build tools faster than ever. This power-up is largely accessible to everybody with a net connection and enough money to pay for the (relatively cheap) access.

…but it’s not all positive. There are unresolvable issues with entrepreneurship in 2025:

If we don’t invest in AI, we risk falling behind, but equally, if we invest incorrectly, we risk wasting critical time and resources as new tools make our work redundant.

We wade through a monsoon of AI news each day; with so many voices, it’s getting 10x harder to know where/what to read. As LLMs get better at persuading, there will be more traps and pitfalls to avoid amongst the hype and progress posts.

AI Agents are on the cusp of being ‘real’ - but they are not yet - and because nobody really knows when they’ll evolve to the point of radically changing business, uncertainty prevails.

Is this repeating beat in my subconscious entrepreneurs' doubt or rational anxiety? It’s hard not to question what we’re building amongst all this. I find myself asking:

Where do software entrepreneurs add value in an AI-dev world?

Where should we spend delta v amongst such chaotic turbulence?

Right now, there’s no clear answer.

Turns Out We’re Still Early

I’ve not given up on my dream, but the ride is pretty lumpy.

As you can tell from the frothy AI news carnival, things are developing fast. Bit by bit, I think we’ll get there.

But I sustain that none of us can really picture what ‘there’ will look like. It’s a hard future to predict.

Where will the boundaries fall? Will super agents level the playing fields? Does marketing become a computing power-struggle as AIs battle to make better and better algorithms to secure attention?

What does a day in the life of an entrepreneur even look like in 2026?

Even though I’m fairly near the front of this wave, I struggle to see how this all plays out.

I’ve made some progress with this challenge, but I can’t help but feel I need to get more radical.

Follow the Build