4 ms·
> You get the LLM to draft some code for you that’s 80% complete/correct. You tweak the last 20% by hand. How much of a productivity increase is that? Well jeep
by blackbear_ 4y ago
> You get the LLM to draft some code for you that’s 80% complete/correct. You tweak the last 20% by hand. How much of a productivity increase is that? Well jeepers, if you’re only doing 1/5th the work, then you are… punches buttons on calculator watch… five times as productive.
Really makes me think that the author has no idea how professional software developers work.
At least 30% of the time is spent in meetings, add another 20% spent on thinking and gathering information about the feature/bug, and maybe 20% validating that the thing implemented works as intended. That leaves only 30% of the time spent on actual coding, give or take. And sure, say that a LLM saves you 80% of the time for coding, then the real productivity increase is.... punches button on calculator... 28%. Even assuming, generously, that 50% of the time is spent on coding, the productivity increase is 67%. Considerable, but not nearly a doomsday change.
- gk1 4y agoI suspect the author knows a lot more about how SWEs work than even many SWEs... https://en.wikipedia.org/wiki/Steve_Yegge https://en.wikipedia.org/wiki/Steve_Yegge
- namaria 4y agoAdd to that calculation that often asynchronous communication blocks work - finishing up code more quickly doesn't necessarily mean shipping feature faster.
- esquire_900 4y agoI'd even argue that if you are given a bunch of code of which 20% is incorrect, it's going to take more than 20% time to fix that. If you've written 100%, you know what the code is doing, what kind of data the variables are supposed to hold etc. In the 80-20% case you've got to build the internal model, test the code, reason and think about it, all of which you would have done during writing. There are so many nuances to these kinds of statistics (same as the github copilot claims) that they give me a little marketing nausea every time they are claimed as truth.
- teaearlgraycold 4y agoHe knows. He’s just selling a product.