Rapid prototyping, internal tool development, spreadsheet replacement, and translating business needs into buildable software

Spreadsheets are great, until they are not. With spreadsheets scattered everywhere, they are where data goes to die.

More recently I've been raising a 'red flag' mote to myself whenever I see a spreadsheet being used, or being tempted to use. And when I see that happening, I ask whether an internal app can be made instead.

If there is a potential great use of AI, it is for wherever you use a spreadsheet. This is the mindset we currently have. Wherever we have recently used a spreadsheet to help us get something done, perhaps we should vibe code it instead. And it feels good.

I feel it is too overwhelming to go in and redesign everything with an AI lens, maybe one day we will, but what is more manageable and sensible is to seek efficiencies. And that is what we've been doing - seeking efficiencies. This also feels like a less risky approach. We need to have confidence in the transistions we make.

One example we currently, is the creation of our certifications. With multiple authors and content flying about everywhere, we had a spreadsheet that was religiously updated. I was always too scared to go near it! But this time Simon has used Claude to build an internal app to help keep track of all the requirements of our next cert. No, it won't be an automated process, but it we believe it will be more efficient and joyful.

I've been asked a few times "why can't my prompts work like I see others do?".

The easy answer to give is that they have better prompts, skills, instructions etc, but I think this is just a small part of it. Yes they do have impact, but they are not the main cause.

I saw a great You tube video from ThePrimeagen and he nailed it, the issue is:
"Its easy to get AI to make impressive things, but difficult to make it make exactly what you want"

You can get 80/90% of what you want with one prompt in minutes, but the last 20/10% of exactly what you want can takes hours of frustration.

Pro-code developers will often do that bit manually, but if you are #vibecoding that is not an option.

For me that leaves to options:

Change your approach, be comfortable with a solution that delivers everything you need but maybe not exactly how you in-visioned it.

Or when it really matters use other tools, don't believe the hype, Canvas #powerapps are not dead and still a great solution for the right problem.

Using #AI is often not a silver bullet to get a quick fix, it requires you to:
🔷Adapt your workflows
🔷Change your expectations
🔷Learn new techniques

and even then it is still not always the best solution

I’m a Principal Developer and I haven’t written a line of code in a year.

That’s a strange sentence to write.

A year ago, I was still deep in C#, TypeScript, APIs, infrastructure, architecture reviews, debugging production systems, Terraform, and CI/CD pipelines.

Today?

I mostly describe systems.
I talk to AI.
I architect with AI.
I review with AI.
I direct, refine, test, challenge, and iterate with AI.

But physically typing code?

Almost never.

The last thing I manually “coded” was tweaking a bit of Terraform. Even that now feels one voice command away from disappearing entirely.

And honestly, it’s unsettling.

I genuinely feel like an accountant in 1863 who’s just been handed a MacBook Pro and a subscription to Xero.

Not because it’s impossible to comprehend.

Because within minutes you realise entire industries are about to change around it.

And then the terrifying thought arrives:

What could somebody from that era have built if they’d truly understood the tool they were holding?

That’s the uncomfortable part about the current AI wave.

Not the hype.

Not the demos.

The speed.

Because we’re rapidly moving toward a world where a non-technical person says:

“I want a CRM system that connects warehouse operations, customer service, complaints, sales, marketing, IT, security testing, and technical teams” and I want it to solve operational problems.

And increasingly, the answer is no longer:

“That will take a team of developers 18 months.”

The answer is:

“Okay.”

That’s the shift.

Not years away.

Months away if not days.

Software development itself is becoming abstracted.

The value is moving higher up the stack:

Understanding systems
Understanding businesses
Understanding people

I’m obsessed with AI because I understand what it can deliver.

The closer you are to the technology, the less theoretical it feels.

I sit there sometimes thinking:

What do you even tell your children to learn now?

What skills still compound?

What does society look like in 18 months if this pace continues?

For decades we built society around knowledge accumulation.

Go to university.
Build expertise.
Become specialised.

But what happens when intelligence itself becomes massively accessible?

What happens when execution collapses from years into days?

It’s beginning to feel like the bottleneck is no longer software development.

Delivery is rapidly becoming commoditised.

The people who win over the next few years probably won’t be the people who produce the most output manually.

They’ll be the people who can identify valuable problems and direct intelligence effectively.

That’s partly why I’m so focused on AI now.

Because it feels inevitable.

And honestly, the biggest challenge no longer feels technical.

The challenge is figuring out where to apply all of this capability before the rest of the world catches up.

Because for the first time in my career, I’m not sure where the ceiling is anymore.

And I’m not sure anybody else does either.

I'll be honest — I've always been a bit cynical about the glorified learning posts on LinkedIn. But here we are.

6 weeks ago I had never written a single line of code. I always wanted to learn - but never took the time! I had a rough concept of vibe coding - but no idea what to build. Until I found my "have to fix it" problem.

Territory planning reviews were taking hours of pivot tables, multiple dashboards, and customer PowerPoints - and still leaving gaps in the story. That felt like something worth breaking.

So I did. With GitHub Copilot powered by Claude as my co-pilot, I vibe coded my way through React, TypeScript, and data joins I didn't fully understand until they fell apart. Then I rebuilt them. Then broke them again. Then GitHub Copilot fixed them!

The result? A self-contained territory planning web app for our Security business - multi geo versions, one file, no install. Uniformed approach across our entire team.

What I took from it:
→ You don't need to understand everything to build something useful
→ Breaking things is faster than reading about them
→ AI doesn't write your solution — it helps you think through the problem

If you're sitting on a problem you think needs a developer to solve, it might be worth trying it yourself first.

Worst case you learn something. Absolute best case you ship something.
Middle ground - you have a great demo and someone far more qualified takes it from there. (we still need developers!!)

Spent an hour debugging a dropdown that wouldn't work.

Checked the logic. Checked the state. Checked it again.

Then I asked the AI one question:

"Are these two different components?"
"Yes."

The AI built TWO separate dropdown handlers instead of one. Two. For the same job.

I wasn't debugging bad code.

I was debugging bad architecture I never asked for.

That's the thing about building with AI. It doesn't argue. It doesn't push back. It just... builds. Even when what it's building is dumb.

Your job isn't to write the code anymore.

Your job is to catch the AI when it gets "creative" ⚡️

Side note: we build internal tools this way for a fraction of what most teams pay for SaaS subscriptions.

What's the dumbest thing AI has confidently built for you? 👇

Interesting take on the state of application development, at least for web apps.
I wanted to get an app to help me read through articles and books faster. Normally I'm not a fast reader, but using the RSVP (Rapid Serial Visual Presentation) technique really helps me go faster. It shows one word (or more) at a time in sequential flashes from a source document. So I went and downloaded some apps from the app store.

All were ok but had ads, were not exactly what I needed, and didn't fit my needs well -- so I made one. Well, I defined it and Claude made it.

I haven't looked closely at the code, but after one bug that was quickly corrected, I had an app working that perfectly suited to my needs. As a user I didn't care that it was a completely front-end app. I didn't care about the technical tradeoffs (that I has Claude explain to me). It just worked.

What does this say about the market and the value of paid applications if you know how to prompt an AI and have some basic application hosting skills? Will this undermine the market for simple utility application on the web, desktop, or mobile? Will it open the door for more apps, which will make standing out in a crowd even harder for developers? Will the market be willing and able to pay the price tag on your hand-coded application that perfectly fits the product user persona you defined? Or, will they just make their own?

Let me know your thoughts as this directly impacts everyone who makes paid software.

Interesting time to be in software engineering.

Not trying to start a war but… it kinda feels like something shifted this year.

I’m seeing solo founders shipping like crazy. Full apps. Landing pages. Internal tools. Stuff that used to need a small dev team + designer + PM.

Now it’s just one person + AI + caffeine.

I’m not saying AI replaces skill. If you don’t understand what you’re building, it shows fast. But if you do know your domain? It’s almost unfair how fast you can move.

I’m building a niche product right now and honestly some days it feels like I have 3–4 invisible teammates. And other days it feels like I’m duct-taping chaos together 😅

Are we actually entering the era of “1-person serious companies” or is this just early hype and we’ll hit a wall soon?

Curious what you’re seeing in real life, not Twitter threads.