---
title: "Software Developers Are Not Okay"
description: "Baldur Bjarnason examines the strain among software developers as AI tools reshape their work, a preview of pressures other skilled professions are likely to face next."
kind: "video"
section: "Mind"
author: "Josh Alley"
author_url: "https://hastalavista.ai/authors/alley"
published: "2026-10-07T18:54:01.249Z"
updated: "2026-10-07T19:19:49.880Z"
canonical: "https://hastalavista.ai/mind/software-developers-are-not-okay"
video_url: "https://www.youtube.com/watch?v=KsEpW03gxKg"
video_channel: "Baldur Bjarnason"
---

# Software Developers Are Not Okay

*Baldur Bjarnason examines the strain among software developers as AI tools reshape their work, a preview of pressures other skilled professions are likely to face next.*

By [Josh Alley](https://hastalavista.ai/authors/alley) · October 7, 2026

Video: [Software developers are not okay](https://www.youtube.com/watch?v=KsEpW03gxKg) — Baldur Bjarnason on YouTube

Hear it from the horse's mouth when they understand cars are here: AI is better than you at many forms of human skill, and AI will only rapidly improve and cover more ground.

## Those who are at the frontier with AI are experiencing what will be the future for others. You're not excluded.

When AI was becoming widely available and people started to use it, they got their first taste of the capabilities. They saw it, they experienced it, and the most common counterpoint was always a god-of-the-gaps cope.

You've heard it, too. And you probably still hold some of those same beliefs.

> Well it can't do X perfectly, so my job is safe.

> It doesn't account for taste, so people are still needed.

When AI videos started to be a thing, you'd get a hilariously bad [Will Smith eating spaghetti](https://en.wikipedia.org/wiki/Will_Smith_eating_spaghetti_test). This created joy even, because now we have a new art form to play with, yet it's nowhere near being good enough to replace even a junior level 3d artist.

[Video: Will Smith eats spaghetti. Poorly. And in 2023.](https://oli14uru1ac8n1u4.public.blob.vercel-storage.com/uploads/video/Will_Smith_Eating_Spaghetti_Original.webm.240p.vp9-JIADdTBIOf49xJZvYHbGaovrCv4j10.webm)

Three years later, Will Smith is now believably eating spaghetti.

[Video: Will Smith eats spaghetti. Normally. And in 2026.](https://oli14uru1ac8n1u4.public.blob.vercel-storage.com/uploads/video/Will_Smith_eating_spaghetti_Google_Veo_3.webm.480p.vp9-k7SgEjFBwjLHJnsKYab2ZKnOLWE9Dj.webm)

## In 3 years, AI closed this gap. And now, we're pouring trillions into scaling the tech so it can improve much more quickly.

It's the *pace* of progression that most people don't fully comprehend. In 3 years, a joke turned into something seriously damn legit. In 3 years, everyone in this industry felt pretty safe, and this kind of genuine excellence was not conceivable.

Imagine the quality of these two videos as being scored: the first gets a 10, the last gets a 90. There were other iterations across other models between 2023 and 2026, and they were showing solid improvement every year. These are arbitrary scores, but you get the picture.

Now plot that on a graph. The X axis is time, the Y axis is quality.

**Will Smith eating spaghetti test**

| Year | Quality |
| --- | --- |
| 2023 | 10 |
| 2024 | 20 |
| 2025 | 45 |
| 2026 | 90 |

Source: [My own bullshit.](https://hastalavista.ai)

That graph directly represents the Will Smith video test, and a similar shaped graph exists for software development, graphic design, copywriting, tax preparation, legal work, radiology, and yes, even plumbers. The more digital-first paths are the lowest hanging fruit, the physical ones will have a longer X axis. But it's all the same.

## AI's purpose is to do this. It is the *value proposition* of the technology. It is the reason to adopt it.

> **Thought exercise**
>
> Will Smith eating spaghetti went from a joke to legit in 3 years. Software development as a career is essentially over.
>
> Companies are incentivized to adopt AI as fast as possible to have the cutting, competing edge. Human labor is one of the most expensive parts of business.
>
> AI companies are, quite literally, racing to close as many gaps as possible, because if they can do that effectively, that gives them more industries — more customers — to pay for the tech and become dependent upon it.
>
> What is safe? What jobs are safe, what functions of a human in a company are safe?

> **Related:** [Reasoning Got Cheap. But Imagination Was Always Expensive](https://petesena.substack.com/p/reasoning-got-cheap-but-imagination) (petesena.substack.com)
>
> Stop chasing yesterday. Tomorrow is already here.
>
> Pete Sena is a smart dude at the forefront of actually utilizing AI in business, across multiple industries.

In the linked article above, the author makes a case for the gaps where humans can still find value and make crucially important contributions and steering for AI.

But those are gaps to close. "Taste" can be trained. Any action where the human has an impact can be measured, be it in sales, clicks, and so on.

> **Thought exercise**
>
> You're a hot shit marketer and you come up with an amazing campaign idea. Sometimes those ideas don't land with your audience, even though they're brilliant to you. Sometimes they absolutely crush. You learn from this, you know your customers more as a result, and your win rate as you go on gets better.
>
> Sales are data points. So are clicks. So is whatever the metric that indicates that the idea and execution was valid. It can be measured — it must be measured — because that's how you determine success.
>
> AI can read data, that's kind of it's whole thing. The entire gap written about in the article can be automated.
>
> If AI can admittedly generate a lot of good ideas, then it can also test against those ideas. It can generate copy, it can generate imagery, it can generate videos, and it can test every single idea with a budget and determine which leads to the best results. You can automate resurrecting ideas and testing them again.
>
> Taste, as determined by what the people respond to in accordance to your objectives, is training data.