Impact of AI on Tech 2026

Intro

Since the release of ChatGPT 3.5 in November 2022, we have been under a constant bombardment of claims that the end of everything as we know it is upon us. We have been just 6-12 months away from the end of X, Y, Z job across what seems like every sector multiple times. No doubt, there has been change, but is it as dramatic as we are lead to believe?

In this post, I will explore the recent history of AI, and how it has already impacted my profession in the technology sector, in this, the year of our AI-Overlord 2026.

AI Tools

To set the scene, let's go over the progression of tools over the last few years.

Web Chat

This is where it all stared. Web interfaces that you interact with via conversational text chat. First there was ChatGPT by OpenAI, then others quickly followed; Claude by Anthropic, Perplexity by Perplexity AI, and the big players like Google have quickly followed with their own offerings.

These types of interfaces are more of a replacement for Websearch now. They have added features that allow you to upload files, provide your own context, and now have the ability to acceess live data, not just rely on their training data.

Editor Integration and Editors

Next up, came the wave of coding assistants via Editor interegrations such as Github CoPilot and full blown editors such as Cursor and Windsurf dedicated to using AI to assist with generating code.

This was a big step forward in AI assisted code generation. When running in your editor, the AI has access to the code base. This allows it to make more informed decisions about code style, layout, library choices etc..

Tools and MCP Severs

Then came tools and soon after Model Context Protocol (MCP) servers. Tools are programming functions that models can use to get something done. MCP servers host tools and can also 'advertise' the tools they have available to use.

I have currently not made use of tools or MCP servers. I am sure they have uses, I just have not found the need to use them yet. They are also looking less useful with what came next.

Agents and Skills

Here we arrive to the current state of affairs (Feb 2026) with Agents and Skills.

Agents are autonomous workers that when given a task, work towards completing it with minimal human intervention. They can think and reason independently, plan their work and go about completing it like good little worker bees.

Skills are sets of domain knowledge that is packaged up in a way that agents can easily understand and make use of. Skill are only used when needed to achieve an objective. This helps to keep the context window clean.

So, Agents and Skill are the current wave off tooling and they have resulted in a staggering increase in usefulness and results. There is Claude Code from Anthropic, Github Coplit CLI from Github and Opencode an opensource player that lets you use many models including local ones.

Winds of Change

When the winds of change blow, some people build walls and others build windmills.

Chinese proverb

My Teeth Hurt

In January 2026, I was off work for a week with an infected wisdom tooth. I was pretty wrecked and did not feel like doing anything. However, I thought I would give this opencode tool a try to see what the buzz was about.

Last year on a whim, I decided rebuild this blog yet again. The details can be found here. Well, that turned out to be the easy part. Migrating the 330 odd posts from HTML/Jinja style syntax, to Rust data types, going back almost 10 years was taking me AAAGGGEESSS. I was only about 50 pages down after 6 months of effort here and there (It was major boring shit).

So anyway, I figured what the hell, this is a low stakes project and i'll experiment while i'm on drugs and feeling edgy. Well long story short, It took opencode using the claude sonnet 4.5 model less than a day to migrate the remaining 280 odd posts. This was with around 4-5 hours of input/planning from me. Remember, this was my very first time using this tooling and I was going pretty slowly at the start as I had no idea what I was doing. And the most amazing part; The tool, did almost ALL of the work by itself, AND IT WORKED! I wrote less than 1% of the resulting code. MAGNIFIQUE!

This actually blew me away. The migration was taking me so long I was considering giving up and shutting my blog down. It is hard to imagine if you have not used these tools yourself. The productivity increase is genuinly astounding.

But, Why?

Why is it so good though? My observation is this: Agents run in your terminal via a TUI application rather than in your code editor. It's a small shift, with a big impact. Agents have access to your code base AND also tools installed on your operating system. This allows them to be much more affective at completing work. Additionally, because the project was written in Rust, the agent could also compile the site and if it wasn't compiling, it new something was broken, then was able to find what it was and fix it on it's own.

I didn't even have to tell it, it was Linus Torvalds encarnate 100x engineer or, don't make any mistakes!

This is what changed my mind about the usefulness of AI in tech. In the last few months, this agentic workflow (I hate the term 'agentic' BTW) has appeared and it is an actual real game changer. This is what the grifters have been promising us for the last 4 years. Well, it's finally here! The stuff that came before this, was an ankle deep wave, this is a tsumani with the whole ocean behind it.

YeAhOkBuDdY

I get it, if I had not experienced it myself, I would not believe me either.

So don't take my word for it. Here are some posts/videos from heavy hitters in the industry who are at the forefront in embracing this new technology.

Mitchel Hashimoto (Founder Hashicorp, Ghostty)

My AI Adoption Journey.

Armin Ronacher (Creator of Flash/Jinja2)

AI Changes Everything.

Thorsten Ball (Software Developer, Author - Writing An Interpreter In Go, Writing A Compiler In Go)

How AI Might Change Programming?

There are many more examples. If you ignore the dramatic headlines, you will find the minds of the industry are turning.

Perceived Value

This raises an interesting question. In this new world, how is the value of your work measured?

In tech, you generally need to be good a learning new things, know how things work, and how to quickly find out how they work to fix something that is broken.

Since the release of ChatGPT and the subsequent deep integration with editors/IDE's and now Agents, the perceived value of writing code and engineering systems has dropped down to close to zero.

Why is that? Did building reliable systems become so easy a mere mortal with 10 minutes to spare could whip out a SalesForce competitor while taking a dump? Of course not! But the perception that it's possible has permiated the atmosphere and, this perception has pushed your market value to hystorically low levels.

If you don't believe this to be true, look at the state of the job market. We have seen record layoffs over the last couple of years, with less jobs in the market and more competition, salaries on offer have gone back in time 5 years.

But, don't despair friends! People who actually build and manage system know the truth. How? Because we use the AI in anger. AI is just a tool, tools can help you go faster and improve quality but the laws of building software remain true: "The the first 90% takes 90% of the time, the last 10% takes the other 90% of the time."

The window might be smaller, but the law still holds.

Future Planning

If you want to survive in Tech, It helps to look at what is on the horizon, try to filter out the noise, and keep the promising/interesting things in sight. It's better to work towards your goals over months and years rather than needing to make a sudden pivot with little experience, likely, thanks to our friend Murphy, when you will probably need it the most.

From what I have seen so far, if there are no more improvements, we already have a 10-100x improvement over doing everything the old way. If progress continues even further, it's a bit scary to think where things could be in another 3 years from now.

Jevons Paradox is observed to occur when technological improvements that increase the efficency of a resource, lead to an INCREASE in it's use, rather than a decrease. This is due to the reduction in cost which increases accessability and in-turn, demand.

We see this all the time in tech. Recent examples are; data transmission and cloud computing. And now, we are seeing this with AI models. A model you can run locally on a Mac Mini or a consumer grade GPU gives you comparable results to frontier models from a year ago that run on a datacenter compute cluster (at everyday tasks).

Therefore (I am wearing my optimism hat), if Jevon is right, we will see an explosion of productivity due to AI, not a reduction. AI will help us go faster and futher than ever before. But it will require you to make a shift.

Powertools

Was the invention of the nailgun the end the carpenter? No, it was not even the end of the hammer. We have circular saws that make cutting faster and better quality, then why do we still have hand saws? These tools still exist because sometimes, they are the right tool for the job.

AI tools are like powertools for people with soft hands. There will still be a place for doing things the way we used to, it will just be reserved for instances when it is the right tool for the job.

These are powerful tools, but don't let the tools think for you or, own your output.

Help Me Obi-Wan

These tools are not magical, they are not the Force. Sorry to tell you, it does take effort and there is a skill to using them effectively. Unfortunately, there are not many guides around at the moment and the best way to learn is to just try and start using them.

At the moment, the best advice I can give you, is get started NOW, like TODAY! Chip away at it a little at a time, be consistent. Look for tutorials on how people are using these tools and experiment yourself. Youtube is full of pretty useful content with many people sharing their experience and workflows. Depending on what you are doing, different tools have different benefits.

You don't learn how to play guitar by watching someone on youtube play guitar. You have to actually pick the thing up and play it yourself.

Albert Einstein - Probably

Here are the AI tools I use the most as a guide to get started.

Perplexity - Research topics and sounding board.
VSCode - With Github Copilot plugin, to help understand code and fix bugs.
Opencode - Terminal based code assistant to plan and write code.

I plan in writing some posts, or posting Youtube videos in the future once I have a better handle on my workflow. Things are moving fast at the moment, you have to keep your ear to the ground on new features and tools.

Outro

This has been a bit of a long one. This shift has been weighing on my mind and I needed to get my thoughts out. I have nothing to sell, the only skin I have in this game is needing to pay my mortgage and keep my family clothed and fed.

I have never been this unsure about what the future holds for my career and I changed careers from IT to construction and back to IT (where I have spent the last almost 20 years). At the same time I feel very optimistic that new and exciting opportunities will open up. Opportunities that at the current time, we can't even dream of.

It really feels like a very big shift is coming, and it's coming quick. In the mean time, with some effort, you can achieve the mythical: "Good + Fast + Cheap - Pick Three". At least until the VC money runs out anyway 😚

✌️ Peace out nerds. Stay weird! ✌️