Tomáš Repčík - 1. 2. 2026

Rise of Personal Software

How AI Changed How I Get Things Done

2026

First of all, a bit late, but Happy New Year 2026! 🎉

I have been in limbo since all AI tools started to emerge and have seen potential in them, but they were not quite there yet.

Hallucination, lack of context, and inability to perform complex tasks made them more of a better refactoring tool, than an actual developer.

Over the time, they still got better and better, and now in 2026, I can say that AI tools have become an integral part of my workflow.

Since I tried Claude Opus 4.5 in Antigravity IDE, I have been blown away by how much more productive I have become.

Suddenly, I could delegate complex tasks with the right harness to AI, and it would deliver results that were not only accurate but also contextually relevant.

It just did my job.

Since then, I have been experimenting and trying to find borders of the model. Of course there are still limitations and if you do not put correct guardrails, you will get bad results.

Give it the right prompt, context, and tools, and it will deliver. It is able to reason, plan, and execute complex tasks.

To give you a couple of examples:

This is just the beginning.

Age of Custom Software

We are entering the age of custom software, where AI tools will allow us to create software that is tailored to our specific needs and requirements.

Code has never been cheaper to produce, and with AI tools, it is becoming even more accessible. Of course, you still need programming experience, because I am able to guide the model and review the code it produces, but the barrier to entry is lower than ever before.

I have never been so willing to just throw away code or rebuild something from scratch, because I know that I can get it done quickly with AI assistance.

… the only limitation is the number of tokens …

Basically, the AI generated code is throw away code, because it is so cheap to produce. It is a new kind of legacy code. Even one off scripts are now worth writing, because you can get them done in minutes.

This is a paradigm shift in software development, and I believe that we are just scratching the surface of what is possible.

Software engineering is still relevant, because we need to guide the models, review the code, and ensure that it meets our requirements. However, it is less about writing code, but actually about engineering solutions with code.

That is why, I am able to write a command line tool in Golang, a web application in React, and infrastructure as code in Terraform, all in the same day, because I do not think in terms of syntax or language specifics, but in terms of problem solving.

The syntax might not be the most optimal, but it is good enough to get the job done, and I can always refactor it later if needed (It is throw away code after all, right?).

Just Try It

If you are still in the trenches from older incompetent AI tools, I urge you to try the latest AI tools.

You will be surprised, if you toy around with them and give them a chance.

Will they give you AI slop? Yes.

Will they do mistake? Yes.

Will they hallucinate? Yes.

But with right guardrails, context, and prompts, they will deliver results that are accurate, relevant, and useful. Mainly, it will save you tons of time.

This is how I spent my last days, experimenting with these tools, and I can say that I have never been more productive.

Now I pay for Claude Code, because it is worth every penny. Mainly with its speed and efficiency. However, Codex from OpenAI is also very capable in much more complex tasks, but it is a bit slower. Pick your poison.

Socials

Thanks for reading this article!

For more content like this, follow me here or on X or LinkedIn.

Subscribe for more
LinkedIn GitHub Medium X