4 ms·

Which is it is clear - the enthusiast have spent countless hours learning/configuring/adjusting, figuring out limitations, guarding against issue etc etc etc an

Which is it is clear - the enthusiast have spent countless hours learning/configuring/adjusting, figuring out limitations, guarding against issue etc etc etc and now do 50 to 100 PRs per week like Boris Others … need to roll up the sleeves and catch up

burnte 10mo ago
Or the tool makers could just make better tools. I'm in that camp, I say make the tool adapt to me. Computers are here to help humans, not the reverse.
- bdangubic 10mo ago
  so when you get a new computer you just use it, as-is, just like out of the box that’s your computer experience? you don’t install any programs, connect printer, nothing eh? too funny reading “tool should adapt to me” and there are roughly 8.3 billion “me” around - can’t even put together what that means honestly
paodealho 10mo ago
There isn't anything clear until someone manages to publish measurable and reproducible results for these tools while working on real world use cases. Until then it's just people pulling the lever on a black box.
- nfw2 10mo ago
  This is the measurable evidence you are talking about: https://a16z.com/revenue-benchmarks-ai-apps/ https://a16z.com/revenue-benchmarks-ai-apps/
  - overfeed 10mo ago
    Here's even earlier evidence: https://en.wikipedia.org/wiki/Tulip_mania https://en.wikipedia.org/wiki/Tulip_mania
    - nfw2 10mo ago
      this point is about revenue not valuations
  - FEELmyAGI 10mo ago
    All three companies profiled are selling products TO vibe-coders, not apps created BY AI-utilizers. The shovel seller in the gold rush analogy.
- conception 10mo ago
  Hundreds of millions of people use these every day on real world use cases. If they didn’t work, people wouldn’t use them.
zeroonetwothree 10mo ago
Merely counting PRs is not very impressive to me. My pre LLM average is around 50/week anyway. But I’m not going to claim that somehow makes me the best programmer ever. I’m sure someone with 1 super valuable PR can easily create more value than I do.
- Terr_ 10mo ago
  Maybe I'm just in a weird place, but I can't imagine 50 PRs a week. Maybe it's because I spend a lot of my time just turning problem reports reports on slack into tickets with tables of results and stack traces.
  - bdangubic 10mo ago
    automate that shit
    - Terr_ 10mo ago
      Unfortunately its mostly B2B integration stuff, where the other end is another company which can sometimes be just as a quirky as a user, except at scale. "I received your spreadsheet detailing 821 records that are in State A but still haven't been moved to State B by our system as it adds Datapoint X on a regular basis. From what I can tell, it seems your data is missing crucial pieces you assured us would always be there. What's that? You want us to somehow fix whatever is somehow making those records in your AcmeERP system? Don't you have a support contract with that giant vendor? We seem like an easier target to hit up for impromptu tech-support consulting work? Well, I'll escalate that to the product manager..."
- wakawaka28 10mo ago
  A bunch of tiny PRs is not hard to do manually. But LLMs can write boatloads of code to do kind of sophisticated things. You do have to figure out how to get to a point where you can trust the code. But the LLMs can help you write boatloads of tests too based on plain English descriptions.
  - 10mo ago

burnte 10mo ago

Or the tool makers could just make better tools. I'm in that camp, I say make the tool adapt to me. Computers are here to help humans, not the reverse.

bdangubic 10mo ago
so when you get a new computer you just use it, as-is, just like out of the box that’s your computer experience? you don’t install any programs, connect printer, nothing eh? too funny reading “tool should adapt to me” and there are roughly 8.3 billion “me” around - can’t even put together what that means honestly

paodealho 10mo ago

There isn't anything clear until someone manages to publish measurable and reproducible results for these tools while working on real world use cases. Until then it's just people pulling the lever on a black box.

nfw2 10mo ago
This is the measurable evidence you are talking about: https://a16z.com/revenue-benchmarks-ai-apps/ https://a16z.com/revenue-benchmarks-ai-apps/
- overfeed 10mo ago
  Here's even earlier evidence: https://en.wikipedia.org/wiki/Tulip_mania https://en.wikipedia.org/wiki/Tulip_mania
  - nfw2 10mo ago
    this point is about revenue not valuations
- FEELmyAGI 10mo ago
  All three companies profiled are selling products TO vibe-coders, not apps created BY AI-utilizers. The shovel seller in the gold rush analogy.
conception 10mo ago
Hundreds of millions of people use these every day on real world use cases. If they didn’t work, people wouldn’t use them.

zeroonetwothree 10mo ago

Merely counting PRs is not very impressive to me. My pre LLM average is around 50/week anyway. But I’m not going to claim that somehow makes me the best programmer ever. I’m sure someone with 1 super valuable PR can easily create more value than I do.

Terr_ 10mo ago
Maybe I'm just in a weird place, but I can't imagine 50 PRs a week. Maybe it's because I spend a lot of my time just turning problem reports reports on slack into tickets with tables of results and stack traces.
- bdangubic 10mo ago
  automate that shit
  - Terr_ 10mo ago
    Unfortunately its mostly B2B integration stuff, where the other end is another company which can sometimes be just as a quirky as a user, except at scale. "I received your spreadsheet detailing 821 records that are in State A but still haven't been moved to State B by our system as it adds Datapoint X on a regular basis. From what I can tell, it seems your data is missing crucial pieces you assured us would always be there. What's that? You want us to somehow fix whatever is somehow making those records in your AcmeERP system? Don't you have a support contract with that giant vendor? We seem like an easier target to hit up for impromptu tech-support consulting work? Well, I'll escalate that to the product manager..."
wakawaka28 10mo ago
A bunch of tiny PRs is not hard to do manually. But LLMs can write boatloads of code to do kind of sophisticated things. You do have to figure out how to get to a point where you can trust the code. But the LLMs can help you write boatloads of tests too based on plain English descriptions.
- 10mo ago