Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ask HN: If LLMs are so useful, why haven't we seen any spike in productivity?

news.ycombinator.com

Post reply on HN
Print view View original

51–58 of 58 posts

Re: Ask HN: If LLMs are so useful, why haven't we seen any spike in productivity?

#51

Post by gls2ro »

globular-toast wrote: post #21
proc0 wrote: post #6

Writing code is the easiest part of the process (relatively speaking). Figuring out the requirements, working with stakeholders to drive consensus, and understanding user needs is the bulk of the work. LLMs will certainly lower the entry barriers for new programmers, and might also create a new solopreneur economy because of it. Now non-technical people with ideas can start prototyping and raise money, but would soon…

That doesn't completely explain it. There are tons of open source projects with countless feature requests just begging for someone to implement them. There's not really any research to do. Often the feature being requested is ridiculously simple but difficult to implement.

I think we need to consider that sometimes a feature is not implemented or a PR is not merged also because any code added needs to be maintained.

So even if there are a lot of feature requests that does not mean that the maintainer wants to just implement them in any fast way because that is code/feature that needs maintenance further down.

Re: Ask HN: If LLMs are so useful, why haven't we seen any spike in productivity?

#52

Post by thiht »

As a developer, I do try to use LLMs in my daily work, and I get a slight productivity boost, but I wouldn’t say it’s a spike. It helps me write tests faster, it helps me write pure functions faster, and it helps me autocomplete some documentation faster. It also helps me go faster with 1-time fixes, for db data migration queries or stuff like that.

That said, LLMs in code editor come with some kind of "hyperactivity" which I find really unpleasant. They’re too "in your face", make the code move a lot and sometimes make it a bit harder to focus than without LLM. They can also be extremely frustrating and result in productivity loss, for example when they generate code that’s slightly wrong and you need to take some time to fix it. It’s harder than just writing the code.

Re: Ask HN: If LLMs are so useful, why haven't we seen any spike in productivity?

#53

Post by patagnome »

outside of the company walls as the context for empirical measure:

1. is the web becoming more [accessible](https://abilitynet.org.uk/news-blogs/inaccessible-websites-k... http://useragentman.com/wcag-wishlist/)?

2. are the web pages getting [faster](https://www.nngroup.com/articles/the-need-for-speed/) and lighter?

3. is it righting wrongs about existing non-performant [code](https://www.webperf.tips/tip/cached-js-misconceptions/)?

4. is it encouraging [smaller](https://dyf-tfh.github.io/)?

5. is it promoting historical [insights](https://qntm.org/clean)?

6. is it popping [bubbles](https://www.youtube.com/watch?v=Y7YAXUWG820)?

7. is it encouraging the correct interpretations of actual [innovators](https://mamund.site44.com/articles/objects-v-messages/index....)?

8. is it minimizing or eliminating [traps](https://www.gnu.org/philosophy/javascript-trap.html)? (also see the W3C's Web Sustainability Guideline's on javascript fallbacks)

9. is it avoiding the "[wars](https://tanzu.vmware.com/content/blog/framework-wars-now-all...)"?

10. is it shedding the object-form? https://dreamsongs.com/ObjectsHaveFailedNarrative.html

Re: Ask HN: If LLMs are so useful, why haven't we seen any spike in productivity?

#54

Post by sixhobbits »

eesmith wrote: post #50
sixhobbits wrote: post #47

Earlier quoted context omitted.

Sure if it's online and multiplayer it's similar. My comment is not google over Microsoft but shared/multiplayer docs for tracking changes and reaching agreements in minutes for something that used to take (and for many still does take) weeks or months of manual reconciliation of printed paper documents

Your comment was "instant 50x productivity increase for any legal process ".

Sue me ;)

Re: Ask HN: If LLMs are so useful, why haven't we seen any spike in productivity?

#57

Post by kledru »

My personal experience suggests that

* we are not developing features faster, but we have time for asking questions we had no time to ask before. More ambitious architectures and designs.

* we are fixing bugs faster and we are producing less bugs (because of better designs).

* not everybody is happy about less bugs.

* we discover more vulnerabilities. Again, not everybody is happy about that, they just want new features, not new knowledge of vulnerabilities and technical debt.

Re: Ask HN: If LLMs are so useful, why haven't we seen any spike in productivity?

#58

Post by grahamj »

lm28469 wrote: post #33
grahamj wrote: post #9

The technology still in its infancy and people's ability to harness it even moreso. Patience.

Two years ago people on this forum were trying to convince me the world would be a completely different place in no time This is one of these "in two years" technology, like self driving cars, asteroid mining, &c.

I think people underestimate the rate of progress of technology but overestimate the time it takes the world to adapt to it. AI is progressing rapidly but it won’t change the world overnight.
Post reply on HN