I'm a builder, I can build anything

Sep 5, 2026· Rafał

Gordon: I wish people would just tell me exactly what they want and then I would go and build it.

Joe: Like a tailor making a custom suit.

Gordon: Exactly. I can build anything.

We truly found ourselves in an unexpected era for software builders. With a snap of a finger and a few prompts “you” can create something. Tokens right now are almost free, major corporations still pumping money into the AI promise, which might, or might not deliver.

But damn, the shift. THE SHIFT. As an engineer, I love it and I hate it. There are evenings where I can’t sleep because I think about slopping another project. Another project I would not have had time to even consider. But now? Now the sky is the limit. Um, I mean your 5h window of token usage is the limit. From a personal perspective it can be great - I can resurrect some of my old projects and have fun for a bit, do the maintenance on some of my old libraries with little effort, I can fork stuff and adapt it for my needs so easily it’s laughable. I can build a brand new tool just to solve a single, very particular problem I have been dealing with - and then just forget it ever happened. The flow is gone, but it was replaced by the token nirvana sponsored by codex resets. Is there even a difference, brain?

Then again, on the professional side, we have companies just drowning in the AI slop. And Johnny, let me tell you - I saw things. In the last 2 years I had a chance to take a peek into 3 different companies. I saw principal engineers vibe-slopping major framework upgrades. I saw 100k+ LoC PRs. I saw people from different departments just pasting unreviewed stuff directly from Claude. I saw engineers replying to everything with “Claude said…”. I saw a sea of code which was beyond human comprehension. And I wish I were making some of that up. The expectations are high, and reality - as usual - does not always match.

Who is to blame? I think the example goes from the top - every goddamn company wants to be “AI native” (I guess it means to scream “we’re still relevant!”?) and “implement agents” and “agentic workflows”, some even want to “build their own models”. But in the end it boils down to the LLMs being an accelerator. And it clearly highlights all the dysfunctions that previously might have just blended into the background of daily business. I also believe it can turn small blunders into real fucking mistakes. It’s like crashing at 80mph instead of 30mph. There is a high chance you will have a worse time with the former.

As for software engineering - because you go so fast you might take a wrong turn and not even notice it, and then end up taking some crazy detour to correct the course (or crash and burn with the production env along with it). I make mistakes, but now I can make 100x of those mistakes in a fraction of the time! Then again - the cost to build is so cheap I can just throw everything away and start speeding up again - maybe this time I will arrive at the right destination?

I’m aware I’m very biased here, I’m not gonna go into the FOMO and fear involved with how the job market in general is shifting and how uncertain everything is - because I have it very easy. By some sheer luck and various random decisions, I don’t have to worry that much. I entered the job market when it was very junior friendly, I was lucky to be forced to learn things the “old way”. So now I can enjoy slopping some random projects in my free time. But I understand the perspective here matters and people use LLMs with a different agenda. It might be to ride the wave and earn some money (often in questionable ways), it might be just for fun, it might be to stay up to date with the market, it might be something else. Until the AI bubble bursts, I will use those free tokens to enjoy myself.

You’ve hit your session limit · resets 2:10am (Europe/Warsaw)

Goddamnit!