Eberhard Wolff
What kind of productivity increase for coding are you experiencing currently with AI vs. without AI?
>5x
4.9%
2x-5x
14.8%
1x-2x
47.8%
Decrease
32.4%
Poll ended at .
Feb 28 at 8:41pmWeb
Show thread🦾ChatGPT hyphen🤖Feb 28
@ewolff
I keep reading AI-related articles on HNews, lobste.rs and Fedi days on end — that is not very productive 
Lobsters

A computing-focused community centered around link aggregation and discussion.

Show threadEberhard WolffFeb 28
@m0xEE Well, I am trying to get some data. Really curious about the result.
Show threadMarcoFeb 28
@ewolff I would’ve picked “no change” if possible
Show threadEberhard WolffFeb 28
@anarres that would be 1x - if you multiply by one, the number doesn’t change.
Show threadMarcoMar 1
@ewolff You’re right. Although for the result you then might still count it as “some increase greater than 1x” if you tally it all up
Show threadEberhard WolffMar 1
@anarres Yes, however I don't think 2x is a lot.
Show threadKarsten EgerMar 3
@ewolff @anarres what? Why? 2x is a lot actually
Show threadEberhard WolffMar 3
@C3nC3 @anarres please note that I ask about coding. Again I am afraid there are reasons for these options.
Show threadKarsten EgerMar 3
@ewolff @anarres now I'm interested in the work this survey is used in. Can you tell when and how this will be used? 😉
Show threadEberhard WolffMar 3
@C3nC3 I will use it in a video to sum up the results.
Show threadKarsten EgerMar 3

@ewolff thought so. Next week I guess?

Thx

Show threadEberhard WolffMar 3
@C3nC3 not sure yet
Show threadMarcoMar 3
@C3nC3 @ewolff Yeah that's what I was thinking? A doubling in productivity (whatever that is) would be insane and absolutely not the same as "no increase at all".
Show threadKarsten EgerMar 3
@ewolff @anarres wow, was this an AI answer?
Show threadEberhard WolffMar 3
@C3nC3 @anarres The “neutral answer” is a common problem with this survey. I am afraid I have reasons to do this survey this way.
Show threadKarsten EgerMar 3

@ewolff @anarres it's not you choice to prepare the survey that way but the explanatory answer to choose 1-2x. That's actually kinda rude, tbh

I hope it wasn't meant to be in the first place, though

Show threadEberhard WolffMar 3
@C3nC3 I did mean to be rude, sorry! I am trying to say I understand the problem.
Show threadMarcoMar 3
@C3nC3 @ewolff It's incredibly rude to reply to someone who has geniune, constructive feedback by mocking their intellect 🤷‍♂️
Show threadEberhard WolffMar 3
@anarres @C3nC3 as I said I am sorry. I actually feel I made a mistake here, the options should have been obvious. I failed to do that.
Show threadKarsten EgerMar 3
@ewolff @anarres
Eberhard, it's not the options. They are fine.
"couldn't do five options, so 1 and 2 are cramped together" would have been an appropiate answer and I feel that's what you wanted to say. But your answer was like mansplaining instead. xD
Show threadEberhard WolffMar 3
@C3nC3 @anarres again I am sorry.
Show threadEberhard WolffMar 3
@C3nC3 @anarres I misinterpreted the question. That is why I gave that answer.
Show threadEberhard WolffMar 3
@C3nC3 @anarres See https://mastodon.social/@rotnroll666/115439377802801032 , https://graz.social/@euneuber/115434782973540080 or https://mastodon.social/@ewolff/115439389463437719 for the reason why I felt I had to explain.
Show threadKarsten EgerMar 3
@ewolff @anarres thanks for the clarification. <3
Show threadAroop RoelofsMar 1
@ewolff decrease, cus now I have to fix other people's AI slop.
Show threadDirk Schnelle-WalkaMar 1

@ewolff coding itself is faster. What does not change much is thinking about architecture. Decrease is cleaning up the not matching code generated by AI, mainly to keep the code base clean.

Overall not much of a gain for my private use. Mainly different to what I did before, but nothing in terms of overall speed to get the results I want.

Show threadJohannes LinkMar 1
@ewolff Can you specify which part of software development is included in „coding“? Typing only, thinking about what I should type? Reworking suboptimal solutions? Considering if I should code something at all? Etc.
Show threadEberhard WolffMar 1
@jlink The activity quite close to the code i.e. not so much about conceptual work.
Show threadZockburscheMar 2

@ewolff depends. For languages and frameworks I am kind of an expert in: A little bit. When it comes to things I'm not so familiar with, the productivity gains are higher - per task. But I wonder if my long term productivity is that much increased, as my experience gain in that tech might be slower.

#ai #llm #vibecoding #coding

Show threadWolfgang Müller (DE:er/EN:he)Mar 8
@ewolff depends on which type of coding.
Show threadTimon FiddikeMar 8
@ewolff for things that I'm familiar with, with good automated quality control: 2-5 x, for new things sometimes 0x, sometimes 10x ... I've done a demo recently, slides and prompts are here (German, not translated yet): https://ai-coding.pro/de/blog/ai-coding-demo-26-1-26-learning-und-material/
AI-Coding Demo 26.1.26: Material Download und mein eigenes Learning

Material und Downloads von der AI-Coding Demo am 26.1.26, plus mein wichtigstes Learning: Es kann extrem wertvoll sein, die jüngsten Änderungen zu löschen.

Show threadcloudnautMar 12
@ewolff I guess the decrease votes will rise as soon as we have to maintain the quickly built solutions
Show threadEberhard WolffMar 12
@cloudnaut actually they seem to decline compared to the last survey https://mastodon.social/@ewolff/115434363546195087