Here's the neat thing: you don't.
I've tried, and I feel like I've got closer with faster models, but ultimately the agentic loop excludes you. Even if you're asking the agent to do simple short tasks, it's still: prompt, wait, wait, wait, check, and you never really feel like you're the one in control.
The problem with faster models is also that they're more stupid, so that additionally breaks your flow when you have to fix something dumb it's done.
LLM-powered autocomplete is a bit more like it, but that tends to be either so dumb as to be a net negative, or slow enough to be useless. And autocomplete is pretty distracting for me.
I feel like I'm missing a mode that works more like a pair programmer. Perhaps a multimodal model that can talk to you about what you're writing, as you write it, and offer suggestions rather than trying to take over and do everything for you.
Getting in the flow means continuous, deep concentration and attention, at least in my experience. Prompting and checking is more like managing an underling, I couldn't get in the flow that way. It would be like a driver trying to get in the flow with a vehicle that randomly does unexpected things.
I want something that works in the background, checking my work as I code, running tests and making suggestions... Without being obstrusive. Like a pair programmer.
I recently found llm-buddy (https://github.com/ahyatt/llm-buddy) which might be a good fit.
Llm-buddy is an “Emacs package that watches your recent buffer edits and asks an LLM to review them. When it finds something worth pointing out, it can add a short inline note in the relevant buffer or show a message in a popup buffer (it usually does the former).
The goal is lightweight feedback while you work: typos, logic mistakes, questionable edits, or prose issues that a normal compiler, linter, or spell checker may not catch.”
This is a UI/UX problem, no? While the current suppliers are mostly locked in the ‘chat for everything’ mode. Guess what, we didn’t go to the moon in chat mode, we don’t drive cars via chat and cyborgs don’t play chess that way. Domain specific interfaces are the way to go (opinion).
Edit with an example: Read some interesting science news yesterday regarding man made risk of high water (Nature). Mailed the author, found the article (popular news doesn’t do attribution) and data and code was open source. Claude Fable had it running very fast and explained the things I forgot from high school. Started on localization and adding some methods from my background (econometrics, extreme value theory). All nice in the /hobby/ way. I can overlap fields in hours now. A brilliant feeling (but probably not brilliant).
What I cannot do is assess the value and novelty of the created work on my own. So I still need to have a set of geologists and econometricians / actuaries work through ‘my work’. That’s what we need tools for! We need UI/UX in this case for novel fields interacting with quality controls made easy. I currently wouldn’t dare ask the author for her time based on my slop. And I cannot critically assess what I’ve made. I only learned today that Greenlands ice attracts water, that Manila and other cities are sinking due to exhaustion of their aquafiers and that the North Sea is surge heavy and unique that way.
I think "without being obtrusive" is a key point here, because any type of popups in the midst of working would break flow as well. Fixing compiler warnings is flow-inducing because it's very linear and has a fast feedback loop, so it would have to be something akin to that.
Side-note, I wonder if audio cues would work well here. When another person is commenting on something, we as humans can typically remember their point while still being focused on text, but if a popup with text comes up we usually get distracted by it. Just my two cents.
I've found AI code reviews as a first pass on PRs to be quite non-obstructive to my workflow and the AI code reviewer (while not instantaneous) is certainly faster on a response than the human pass will be.
> I feel like I'm missing a mode that works more like a pair programmer. Perhaps a multimodal model that can talk to you about what you're writing, as you write it, and offer suggestions rather than trying to take over and do everything for you.
This is exactly what I have also been thinking and wanting for a while now. A realtime agent that I can share a screen and mouse / keyboard with. and we can just work together at times. I think it will probably come at some point but we might be a few years away from it.
*You dont*
Skill issue, not a universal problem.