SYSTEM NOTICE

Auto translation by AI. Be sure, accuracy, nuances and authorial intent may not be fully reflected.
見出し画像

How I Solved the ADHD Struggles I Was Having with AI Development

I work as an engineer at a certain cloud vendor in the United States. I want to state clearly from the beginning that this blog post is about my own opinions and experiences, and does not represent my company. Please read this as just one instance.

Further trials for ADHD...

This is a super trivial story, but it has been a while since I started doing most of my system development via chat or automatically. Thanks to agents, I feel like my productivity has gone up like crazy. However, as someone with ADHD, there was one thing I was really struggling with. That is, "context switching." ADHD makes it easy to get distracted as it is, but when running multiple agents in parallel, I often find that when Agent A finishes a task, while I'm reviewing it, Agent B sends a notification, I get distracted by that, and I end up thinking, "Wait, what was I doing again?" While I'm doing that, it seems like Agent C has finished some task.

Please become Shohei Ohtani

For someone with ADHD, multitasking is the thing I am worst at, but in an "AMA (Ask Me Anything)" at my company, I was told, "We expect you all to become Shohei Ohtani" (laughs). Baseball isn't really a major sport in the US, so it was surprising that such an analogy came up, but it seems they want us to do the work of a PM (Product Manager) in addition to the traditional scope of an engineer. By the way, Shohei Ohtani is a player who comes along maybe once in a hundred years (laughs).

Shifting to the talent that is needed

It probably depends on the department, but it seems my boss was told that "People Managers" are no longer needed, and that managers are now expected to be more like technical leaders. And they said that ICs (Individual Contributors) are important. Ultimately, we are building things, and if AI comes along and makes engineers ten times more productive (and productivity is actually going up significantly), what is needed in that world is output that blends several roles, and I feel like the value of people who can actually build "things" is rising.

Where did the theory that engineers are unnecessary go?

By the way, if you ask if it feels like engineers are no longer needed because AI has arrived, that is not the case at all. To do ten-person-worth of engineering with decent quality using AI, you honestly need a certain level of skill and understanding. Moreover, since that is approaching at breakneck speed, humans are definitely the bottleneck. That is why my manager is telling me not to focus on one thing, but since I am at the top level of seniority, I have to do many things at the same time.

Even though I have ADHD and am bad at multitasking. Moreover, since it is the age of agents, I have become a manager of agents, and I can just have multiple agents do work in parallel. But, if I could just give them a prompt and have the agents take care of the rest, that would be great, but life isn't that sweet, so I need to correct their direction or review their work. In the end, that requires solid experience and knowledge of computer science. Seriously, who was it that said engineers would become unnecessary?

Stop multitasking (my own)

Well, in such an environment, as a serious but incompetent person, I have no choice but to put in my own time and do my best, but since I have ADHD, I am already not good at it, and I also have to do context switching, which I am even worse at. I was seriously wondering what I should do. My brain gets super tired, and when I go back and forth, I get totally confused, and I don't make as much progress as I thought I would, which was a very pathetic state, but then one day I suddenly thought that maybe that strategy would solve it. That is the strategy of stopping multitasking. You might think, "Wait, notifications come from various agents, what do you do about that?" but since the bottleneck is me, I should just stop "my own multitasking," regardless of the agents.

If I don't write it down, I won't start

Specifically, as an ADHD countermeasure I have used for a long time, I always write things down before I start. I use OneNote or something, but anything would probably work. If I don't write it down, I won't do it. It's just that, but it works well for me with ADHD. Even if I get distracted, if I have it written down, it's easier to go back to it and focus. It's like an agent. And what do I do? For example, I give instructions to Agent A. The agent does some work for a while, so I become unnecessary for a while. I could kill time by looking at X, but that would be a waste.

Don't switch tasks until you become unnecessary

I write in OneNote that I have sent such-and-such prompt to Agent A and am waiting for the task. Then I write that I will first understand what to do with Agent B, have an interactive conversation with Agent B, summarize the policy, and input it into Agent B. Then, I note in OneNote that I have instructed Agent B to do YY and am waiting for execution. At that point, if Agent A has finished its work, I can review it. Even then, I write in OneNote that I am going to review Agent A's work before I do it. The point is, for example, when I have given instructions to Agent A and am having an interactive conversation with Agent B, if Agent A finishes its work and sends a notification, I ignore it. Because I don't multitask. Even if the agent is flashing, I ignore it. After I finish the conversation with B, I write in OneNote that I will review Agent A, and then I do the review. I always finish one task before moving on to the next.


In this way, I simply ignore notifications from agents while I am in the middle of work, and I just work at my own pace while taking notes. That's all.

Parallel, but multitasking-free is complete

When I did that, my productivity went up like crazy. The bottleneck was indeed me. I got an insane amount of work done today. I was able to do many things in parallel. The principle was the same. Multitasking is inefficient. At least for me. This might seem less efficient, but in the end, if I focus on one thing at a time, my brain doesn't have to go here and there, so it's much easier for me. Moreover, even while working in parallel, by using the method of switching contexts while writing in OneNote, it's easy to resume my train of thought even if I do another task and forget. Because it's written down. And while I'm doing one thing, I can forget about other things. At least for me, it worked incredibly well. When I think about it, it's like removing interrupts from a CPU mechanism.

This time, I shared a method I found by chance that worked very well for me.

Bungeishunju is going to publish my book soon. Please give it a read if you'd like.

 
 

いいなと思ったら応援しよう!