The 'I' In AI

The 'I' In AI

guest@blog:~$ cat ./the-i-in-ai.md

Wed, Sep 9, 2026 · 3 min read · ai

2026 has been a wild year for us software engineers. On one side, everyone and their grandma is talking about how coding is a thing of the past. On the other, LinkedIn is filled with all the “solopreneurs” bragging about their vibe-coded empires and experienced SWEs desperately looking for a job after being laid off. Did we really solve coding? If so, what the hell are we doing here?

“Coding Is Solved”

As much as I find that statement annoying, I do partially agree with it. I have barely written any code manually throughout this year. LLMs are excellent at reading API documentation and churning through thousands of lines of code. For many straightforward problems, they can produce solutions that are as good as what a human would write. I’d be falling behind at my job if I were still writing code manually (and definitely won’t be hitting my required token limit 😂).

As awesome as it is to have coding solved, I do miss the old days when we used to code manually. I got into programming because I found coding exciting. I used to be obsessed with code quality and now no one seems to care about it.

I guess it’s the question of “Am I going to keep complaining about how much I miss driving a horse carriage, or am I going to shut up and adapt to driving a car?”.

If Coding Is Solved, What Value Do SWEs Bring?

Build the next billion-dollar app. Make no mistakes.

The advent of keyboards replaced the punch cards, not the engineers. While today’s LLMs are amazing at spitting out code, engineering is and always has been way more than just writing code. Here’s what I think I bring to the table as a SWE:

• Understanding Business Context: While LLMs work with what we provide, the SWE brings an understanding of the users, the problem space, and the business constraints.
• Making Tradeoffs: CPU Vs Memory, simplicity Vs modularity, consistency Vs availability, short-term Vs long-term, etc.
• System Design: Designing the system accounting for resources, tech debt, cost, etc.
• Debugging Production Systems: With LLMs writing most of the code, people tend to push code changes more often than before, increasing the number of SEVs. While LLMs help speed up the investigation, understanding what went wrong is very much still human.
• Knowing What Not To Build: With coding being cheap, the bottleneck now is to know what needs to be built, and in particular, what not to.

Am I Just A Naysayer?

Despite me yapping about why I can’t be replaced with a clanker above, I do find AI handy in the engineering process. In particular:

• Churning through documentation: LLMs solve the needle in the haystack problem. For a long time, being a good programmer has been knowing how to google and read through documentation. LLMs do it on steroids.
• Analyzing data: LLMs are good at walking through code, telemetry, logs, etc. While they do hallucinate at times and need to be manually audited for their proof-of-work, they speed up investigations significantly.
• Prototyping: LLMs make writing throwaway prototypes cheap.
• Generating Code: LLMs are great at generating code when provided a solid design plan.

Will I Stay Employed Forever?

LLMs are already cutting down team sizes. While I do believe most layoffs in the name of AI are bogus, I do see a world with smaller, more efficient teams with AI. Let’s see what the future holds for us. Until then, here’s a picture of a cat to keep you calm: