I love writing code.
I’ve always loved writing code.
That said, writing code is no longer part of my day job. Today my job is to build my teams and help push our engineering org forward, a very different role that requires a different toolset. In 2026 part of that, whether I like it or not, has involved AI.
To learn how to use AI I’ve also deployed it in some of my own work, but that use case has been far more limited and, frankly, not something I’ve found valuable or even particularly useful.
Where AI has helped me
To claim AI, as a tech, hasn’t been helpful at work would be a lie. The simple fact is that my workload is such that I simply couldn’t get it all done without AI (trust me, I’ve tried). For the most part where I use it is in a few areas:
First is meeting notes and transcriptions. My company pays for Fellow.ai but I’ve stuck with the standard Google Meet notes and transcriptions for nearly all my calls, often 10 or even more per day. Without it I couldn’t keep up. It records all the important parts and sends them to my email where I automatically filter them into a “Meetings” label for processing later. This has saved my butt so many times as I change context far too often during a given day to otherwise be able to keep up with everything.
Next is searching and reporting on internal information. Work data is in a few services including Notion, Google Workspace, Jira and GitHub. When I need to put these together I can spend hours I don’t have outside of meetings or I can ask our work AI, Claude, to find me what I need. I still proofread for accuracy, but it’s been very helpful in putting together what could otherwise require me spending hours going through dozens, or more, of pages and ideas that are rarely organized in any usable way.
Finally is drafting and editing. For work docs I’ll often use AI in one of two ways with documents. Either I ask it to draft me something or I ask it to stitch together something coherent from pages of notes and other info (it’s surprisingly hard to write a document of any kind in the fits and starts that come with my meeting schedule). For the most part the editing is helpful. The drafts, well, I will often keep the structure but, if I have time, I find myself rewriting most of the rest of it to make sure it is accurate.
In my personal life I avoid AI for writing but have also found it very helpful in one-off or other simple tasks to help me with this site and other projects. Over the spring I used it heavily to help me implement small scripts that I just didn’t have time for otherwise. They’ve been so helpful to my blogging workflow that I’ve even published many of those as either npm scripts or as VS Code extensions (where appropriate).
Taken together I’ve found AI helpful in tasks I can verify and that I wouldn’t have time to do myself otherwise.
Where I still won’t use AI
For as helpful as AI has been, there are areas I’ve found where it is better to avoid it.
The first of these is as a replacement of my search engine or really for searching for any facts at all (outside of my verifiable work data). AI is a statistical machine that predicts the next token in a pattern. It doesn’t, and can’t, understand if what it is generating is factually correct and it really should never be used as such.
Second is writing here. It’s helped me with some scripts but I absolutely will not use it to write. This blog is my voice and no one else’s. I’m not throwing that away to a machine.
Finally it is in drafting documents or code on anything I cannot verify. If I don’t know the material well I’ll do it myself. It’s far safer than suffering the embarrassment of submitting something that sounds good but is factually wrong. I see it all around me too, people taking shortcuts and trusting the AI to “do it for them.” It doesn’t take long to spot the errors in this approach and I won’t do that to others.
Where I still hate AI
I use AI because it is an expected, and indeed required, tool for my work in 2026. As ending my career isn’t an option I will continue to do so where it makes sense.
That said, I hate just how ridiculous the power behind AI is and what that power is doing to the world around us. Given that, I’ve experimented with local models and alternatives to the popular models and here I’m starting to find some helpful utility in the same use cases where I’ve found Claude helpful.
In the end, AI, like any tech, has its utility. Where it falls apart today is in the companies and politics behind it. Given that, I’ll use what I have to where I have to and continue to implement “better” solutions into my own workflow as much as I can.