551–560 of 1001 posts
Hi, I lead the teams responsible for our internal developer tools, including AI features. We work very closely with Google DeepMind to adapt Gemini models for Google-scale coding and other Software Engineering usecases. Google has a unique, massive monorepo which poses a lot of fun challenges when it comes to deploying AI capabilities at scale. 1. We take a lot of care to make sure the AI recommendations are safe and…
I'm continually surprised by the amount of negativity that accompanies these sort of statements. The direction of travel is very clear - LLM based systems will be writing more and more code at all companies. I don't think this is a bad thing - if this can be accompanied by an increase in software quality, which is possible. Right now its very hit and miss and everyone has examples of LLMs producing buggy or ridiculou…
In Google, there is a process called "Large Scale Change" which is primarily meant for trivial/safe but extremely tedious code changes that potentially span over the entire monorepo. Such as foundational API changes, trivial optimization, code style etc etc. This is a perfectly suitable for LLM driven code changes (in fact I'm seeing more and more of LLM generated LSC) and I guess a large fraction of mentioned "AI ge…
yeh, but the main problem is the quality. with algorithm bug can be fixed. with llm it's more complicated. in practice they do some mistakes consistently, and in some cases cannot recover even with assistance. (don't take me wrong, I'm very happy with the results most of the time)
Earlier quoted context omitted.
> the kind you might as well copy&paste from stackoverflow This bothers me. I completely understand the conversational aspect - "what approach might work for this?", "how could we reduce the crud in this function?" - it worked a lot for me last year when I tried learning C. But the vast majority of AI use that I see is...not that. It's just glorified, very expensive search. We are willing to burn far, far more fuel t…
What I can't wrap my head around is that making good, efficient software doesn't (by and large) take significantly longer than making bloated, inefficient enterprise spaghetti. The problem is finding people to do it with who care enough to think rigorously about what they're going to do before they start doing it. There's this bizarre misconception popular among bigtech managers that there's some tunable tradeoff bet…
There is no incentive to do it. I worked that way, focused on quality and testing and none of my changes blew up in production. My manager opined that this approach is too slow and that it was ok to have minor breakages as long as they are fixed soon. When things break though, it's blame game all around. Loads of hypocrisy.
I work for Google, and I just got done with my work day. I was just writing I guess what you'd call "AI generated code." But the code completion engine is basically just good at finishing the lines I'm writing. If I'm writing "function getAc..." it's smart enough to complete to "function getActionHandler()", and maybe suggest the correct arguments and a decent jsdoc comment. So basically, it's a helpful productivity…
Do people find these AI auto complete things helpful? I was trying the XCode one and it kept suggesting API calls that don't exist. I spent more time fixing its errors than I would have spent typing the correct API call.
Earlier quoted context omitted.
It's like blockchain right now. I'm sure there is some killer feature that can justify its problem space. But as of now the field is full of swamps. Of grifters, of people with a solution looking for a problem. Of outright scams of questionable legality being challenged as we speak. I'll wait until the swamps work itself out before evaluating an LLM workflow.
Blockchain was always a solution looking for a problem. LLMs are being used right now by a lot of people, myself included, to do tasks which we would have never bothered with before. Again, if you don't know how to use them you can learn.
It's a legal nightmare in my domain as of now, so I'll make sure the Sam Breaker-Friends are weeded out. If it's really all the hype it won't be going anywhere in 5 years.
Earlier quoted context omitted.
Between the fraction of my time I spend actually writing code, and how much of the typing time I’m using to think anyway, I dunno how much of an increase in my overall productivity could realistically be achieved by something that just helped me type the code in faster. Probably not 25% no matter how fast it made that part. 5% is maybe possible, for something that made that part like 2-3x faster, but much more than t…
I imagine that those who cherished the written word thought similar thoughts when the printing press was invented, when the typewriter was invented, and before excel took over bookkeeping. My productivity isn't so much enhanced. It's only 1%... 2%... 5%... globally, for each employee. Have you ever dabbled with, mucked around in, a command line? Autocomplete functions there save millions of man-hour-typing-units per…
Work gets scheduled on short time frames. 5% savings isn't enough to change the schedule for any one person. At most, it gives me time to grab an extra coffee. I can't string together "foregone extra coffees" into "more tasks/days in the schedule".
Earlier quoted context omitted.
I find this hard to believe. Can someone give me an example of something that takes months that AI can correctly do in hours?
Not hours; but days instead of months: porting around 30k lines of legacy livescript project to typescript. Most of the work is in tweaking a prompt for Claude (using Aider) so the porting process is done correctly.
Earlier quoted context omitted.
I'm continually surprised by the amount of negativity that accompanies these sort of statements. The direction of travel is very clear - LLM based systems will be writing more and more code at all companies. I don't think this is a bad thing - if this can be accompanied by an increase in software quality, which is possible. Right now its very hit and miss and everyone has examples of LLMs producing buggy or ridiculou…
My main gripe with this form of code generation is that is primarily used to generate “leaf” code. Code that will not be further adjusted or refactored into the right abstractions. It is now very easy to sprinkle in regexes to validate user input , like email addresses, on every controller instead of using a central lib/utility for that. In the hands of a skilled engineer it is a good tool. But for the rest it mainly…
Some people are touting this as a major feature. "I don't have to pull in some dependency for a minor function - I can just have AI write that simple function for me." I, personally, don't see this as a net positive.
Hi, I lead the teams responsible for our internal developer tools, including AI features. We work very closely with Google DeepMind to adapt Gemini models for Google-scale coding and other Software Engineering usecases. Google has a unique, massive monorepo which poses a lot of fun challenges when it comes to deploying AI capabilities at scale. 1. We take a lot of care to make sure the AI recommendations are safe and…