Hacker News

Ask HN: Anyone else blurting?

6 points by l33tbro ago | 6 comments
Like most, I was stoked on LLMs when they became available. Particularly on the earlier free tiers where the slightly questionable answers were offset by generous tokens.

Now, I'm finding that having to use this tech is becoming much less fun. The constant corrections, the constant prompting for greater brevity, the grating certainty and sycophancy.

Others probably are forgiving considering the net benefit. But for me it honestly now feels like a chore, which is reflected in how my frustration is now appearing in how I communicate to it now.

Instead of articulating and making my prompts precise, I'm just straight-out blurting. Typos, missed words, poor semantic structure - I guess I maybe at some point reasoned that "it will get the picture, so why waste effort being articulate?" Or maybe I'm just anticipating that it is going to get it 15 percent wrong, so why even try and be articulate. Just blurt and steer. Blurt and steer.

Is anyone else finding this? I actually sometimes worry about it. I feel it's not just the fact that I'm offloading my cognition to it that is making me dumber, it is the way I'm communicating with it. They say the act of writing is tied to the act of thinking, and my prompts are basicslly the typographic equivalent of grunting right now.

muzani |next [-]

It's poor form. You're lifting from your back. You're throwing punches from your elbow.

You shouldn't anticipate that it will get 15% off and steer. You try to get it up to 100% before writing any code, and sometimes that involves getting it to scout what's already there.

Control the environment. The scope used to be code and architecture, now it's context. These models are designed to fill in gaps and propose solutions. If they don't know what you want, you have productivity in the the wrong direction.

If you want somewhere to start, look into Matt Pocock's public videos and https://www.aihero.dev/

A rule of thumb would be to speak to it like a teacher would. It has always worked on all models.

ozozozd |next |previous [-]

I do. Maybe not terribly but quite a bit. And for all the same reasons as yours :)

Although this happens less frequently with Chinese models, and recently Gemini Argon with medium level thinking.

Anthropic and OpenAI appear to have defined success as number of prompts sent or number of assistant responses, which means more work for me, so I am out. Plus, I wouldn’t talk to someone twice if they caveat every answer to a question and require me to follow up. I don’t ask questions expecting partial answers with exactly 1 caveat/nuance.

simonahm |next |previous [-]

I have found myself describing bugs and features quite casually, often sending a voice recording through Paseo from my phone, rather than sitting down to write it out. Which is quite a shift, since I've always prided myself in my ability to sit down and quickly produce an organized piece of writing - and yes, the process of writing definitely helped collect my thoughts and think through things. So I understand the feeling of something being lost.

But my semi-rambling voice notes are often enough, along with the context of the project, for the LLM to correctly explain back to me, exactly what I meant - And then typically do a good job at implementing it. So my input was good enough to get the thing done, and the discernment was moved from articulation to verification.

There's trade-offs to that, for sure. But it can also be surprisingly efficient, and allow me to apply that mental capacity elsewhere. I still enjoy stepping away from the computer and going for a walk or run, where I still find myself able to engage in big picture timing and getting everything lined up and connected in my mind. After that, I'm usually much better able to sit down and write down the more significant prompts or design decisions.

l33tbro |root |parent [-]

Maybe I'm way off, but this actually reads a bit like content marketing (Brand new account - product name in first 2 sentences - comment is 3 paragraphs about product).

Not an accusation. But is it content marketing?

ozozozd |root |parent [-]

I think you are spot on, elitebro.

nhecker |previous [-]

No, if anything I'm probably more formal in my communications toward AI with respect to grammar and punctuation and all that, than I am in my casual digital communication with friends and coworkers.