201–210 of 334 posts
Earlier quoted context omitted.
I hope Google has been improving their diffusion model in the background this whole time. Having an agentic system that can spin up diffusion agents for lite tasks would be awesome
Because they would be faster?
Earlier quoted context omitted.
Seriously and beyond productivity, flow state was what I liked most about the job. A cup of coffee and noise cancelling headphones and a 2 hour locked in session were when I felt most in love with programming.
I love the flow state, and I’m pretty sure it’s fundamentally incompatible with prompting. For me, when the flow state kicks in, it’s completely nonverbal and my inner dialogue shuts up. I think that’s part of why it feels so cool and fun when it hits. But LLM prompting requires you to constantly engage with language processing to summarize and review the problem.
For me the fatigue is a little different— it’s the constant switching between doing a little bit of work/coding/reviewing and then stopping to wait for the llm to generate something. The waits are unpredictable length, so you never know if you should wait or switch to a new task. So you just do something to kill a little time while the machine thinks. You never get into a flow state and you feel worn down from this c…
That's a type of fatigue that is not new but I hear you, context switching fatigue has increased ten fold with the introduction of agentic AI coding tools. Here are some more types of fatigue that have been increased with the adoption of LLMs in writing code. There are plenty of articles on review fatigue including https://www.exploravention.com/blogs/soft_arch_agentic_ai/ which I published recently. The focus there…
Earlier quoted context omitted.
It reduces the friction of coding tremendously. Coding was usually not the bottleneck but it still took a significant amount of time. Now we get to spend more time on the real bottlenecks. Gathering requirements from end users, deciding what should be built, etc.
> It reduces the friction of coding tremendously. Coding was usually not the bottleneck but it still took a significant amount of time. I don’t think I agree. How can something be both “usually not a bottleneck” that usually “takes a significant amount of time” ? > Now we get to spend more time on the real bottlenecks. Gathering requirements from end users, deciding what should be built, etc. Sounds like you might re…
I dont understand what you dont understand. Is everything that takes a significant amount of time necessarily a bottleneck? That seems implied by you but makes no logical sense.
The funnel into the programming work is often more difficult/time consuming/resource intensive than the programming.
Also, sometimes its not as costly but should be. And insufficient time and resources were spent up front which caused the coding portion to take a lot longer than it should. In which case the programming time may appear to be the bottleneck but it was still really the funnel leading into it.
> Sounds like you might really enjoy a PM role
Enjoyment isn't really a factor in terms of what work needs to be done. And designing technical features isnt really a PM responsibility.
For me the fatigue is a little different— it’s the constant switching between doing a little bit of work/coding/reviewing and then stopping to wait for the llm to generate something. The waits are unpredictable length, so you never know if you should wait or switch to a new task. So you just do something to kill a little time while the machine thinks. You never get into a flow state and you feel worn down from this c…
I wonder if this is how managers feel -_-'
But the cycle is longer. When you help a person they don’t come back to you 4-20 minutes later.
I also only review PRs at specific times a day, because that’s more cognitively intensive and switching in and out pretty much ensures you’ll do it badly.
Either way, I’m really starting to think agentic as designed is a deeply flawed workflow. The future could be small, fast models that finish pseudo code and look stuff up to aide focus. Anthropic’s own research seems to support this.
This is like people crying how their phones are ruining their lives. Just stop. Take some responsibility and control of your life. I don’t feel exhausted especially after letting LLM hallucinate some code lines. If you do, maybe it is time to re-evaluate your life choices
> When each task takes less time, you don't do fewer tasks. You do more tasks.
And you're also paid more. Find a job that ask less from you if you are fatigued, not everyone want to sacrifice his personal life for his career. That's choices you got to make but ai doesn't inherently force you to become overworked.
Earlier quoted context omitted.
It's not a choice. For example, Windows XP is no longer a choice, because the context around it made it unsafe now, though it didn't change. Life style from an older era is no longer the norm, which means your relative life quality degrades automatically and it actually becomes unsafe.
When I retire I plan to have no phone, no computer, and no TV. These are by far the biggest time sucks in my life and I want to see what I can do without their distractions. I might keep a tablet or old phone with no service so that I can still do email.
Some people tried that a bit and they had to retreat back to the usual connected life. What happens is, that old non-digital disconnected world is no longer there waiting for you. It may pretend to be the old world you desired, but it is looking at you and judging you. You become an animal in a zoo, instead of an anonymous part of the old-time world.
For me the fatigue is a little different— it’s the constant switching between doing a little bit of work/coding/reviewing and then stopping to wait for the llm to generate something. The waits are unpredictable length, so you never know if you should wait or switch to a new task. So you just do something to kill a little time while the machine thinks. You never get into a flow state and you feel worn down from this c…