I have been utterly drained by the amount of AI content lately. Not only the volume but just the way people speak about AI, it’s (to be honest) dehumanising. Not just that it’s mostly written in that weird “AI-voice” but also because it’s all extremes. AI will save / destroy / eat humanity. This whiplash of opinions is really… uninteresting. These extremes don’t really speak to me. I’m not token-maxxing, sometimes AI does work that I find really annoying and pointless and it’s like “hey, that’s cool”. Would I rather live in a world without all this uncertainty about the future? Absolutely. Do I reckon I’ll need to depend on AI for my future career and to have the best chance of continuing to predictably provide for my family? Yes. So here I am, drowning in tools, models, harnesses, paradigms and extreme opinions all lacking in some relatable day-to-day humanity. So to me it comes across as noise screaming on by. Kinda like when I watch a Marvel movie.
So what does this mean for me?
AI has changed how I work, and for a while I was the person who typed things at an AI prompt all day and came away from it like a drained, emotionally exhausted, unfulfilled zombie feeling zero sense of accomplishment. Frankly it feels like mainlining something that’s going to take you directly to burnout city. And I kind of want to draw a parallel with another “paradigm shift” that I also kinda hated. Node.
Here’s me writing and thinking about PHP code:
- Code does A
- Code does B
- Code does C
Awesome, that makes sense, yes there are functions, classes and objects, etc, but the fundamentals of top to bottom sequential execution held most true for your work. Then I had to learn Node. Which is:
- Ask code to do A
- Ask code to do B
- B is done
- Ask code to do C
- A is done
- C is done
And I hated this. When it comes to code, I’m the most expensive thing in the chain and here I am dealing with async await crap all over the place because of some stupid runtime limitation. Eventually I broke my brain enough to code reasonable javascript in the node runtime and it was fine, I guess. (Oddly never had any issues with it on the browser, I guess it felt like a fair tradeoff there). And then there’s the constantly evolving frameworks, tools, etc that again, made me feel tired. Now let’s get back to AI.
How my day typically looks:
- Do thing A
- Do thing B
- Do thing C
Actually I use Cal Newport’s Time-Block Planner so it’s more like:
- 9:00AM Do thing A
- 9:30AM Do thing B
- 10:30AM Do thing C
Now with AI it’s more like:
- 9:00AM Kick off thing A with an agent
- 9:10AM Kick off thing B with an agent
- 9:15AM Thing A has questions
- 9:17AM Kick off thing C with an agent
- 9:22AM Thing C has questions
- 9:27AM Thing A is done, needs review
- 9:40AM Thing A reviewed, in the meantime thing B and C have follow up questions. Thing A continues
- ETC ETC.
By the end of the day it looks like (things spawn more things, hence D through H):
- Thing A, B, F, G are done.
- Thing C, D, E, H are still sort of done, sort of need reviewing but probably will carry on tomorrow. And will become A, B, C, D for tomorrow.
I am not a fan of this way of working. I never thought I’d feel empathy for the Node runtime but here I am. async / await is my job and I’m terrible at it.
I want to be clear here. I have a solution that makes this process less painful for me. But Deep Work is dead as far as I can tell when I’m on a computer. I will arrive, have 12 open browser tabs, 4 open agent windows, Slack and email and I will immediately get sucked in. This was mitigated by deep coding sessions where I had licence to think deeply about a particular programming problem in the code. That is dead. AI has taken it. And so I need to somehow stay sane in this career.
The Notebook
Yeah it’s another “isn’t pen and paper great” post. But to be honest, my notebook has been a life-raft to ground me. When I’m writing code or writing deeply on anything, I kinda get sucked into the machine mentally, I lose track of time, get into a flow state and I come away feeling good. Energised. Proud if I’ve had a great session and like I’ve accomplished something.
Now when I’m doing the same thing I still get sucked into the machine but I come away feeling kinda crappy. It’s not flow state, I don’t feel accomplished, I come away drained and numb. There are so many “tabs” of long, deep conversations, nothing feels closed off. Even after closing the “tab” of an AI chat, something lingers. This sounds extreme but in a small way it feels like rather than leveraging the machine as a tool to make something great, it’s taking something from me. Things feel a bit hollow afterwards.
So in attempting to reclaim more humanity in my work, it’s time to turn to pen and paper. If I need to think deeply on a topic, perhaps it’s architecture, perhaps it’s some process change, or breaking down a coding problem (we still need to do that), it’s done on pen and paper, or I’ll do it in Miro and stay there.
Each morning I open a 2 page A4 spread. On one page is all the random thoughts, tasks, observations that aren’t very interesting to share. Then this is how my “AI with work” page looks (not a real page, just an example).
I keep note of the conversations at key points throughout the day and then at the end of the day, review it. Was it helpful? Do I come away feeling better or worse using AI? Why?
Is this perfect? Definitely not. But I feel like I have a better handle on things and when someone in the pub asks me “How’s work with AI?” I can give something beyond a vague “meh”. On top of that, things feel a bit more finished. It’s also helped me be more intentional in my work. Rather than just rabbit-holing down a long AI thread, or running the risk of giving the AI too much responsibility and offloading all the thinking, I’m deliberate about how much I feel I should hand off vs do myself.
Are you a human who dislikes/hates my solution? Then as a human, share something that works for you. Not the token-maxxing, benchmarking noise, there’s plenty of that. How are you as a human working day to day in this? Share!