4 ms·
5% speed improvement in my tooling would result in about 0.0000001% improvement in my productivity. first of all, so little of what we do is about the tooling.
by 8n4vidtmkvmk 5y ago
5% speed improvement in my tooling would result in about 0.0000001% improvement in my productivity. first of all, so little of what we do is about the tooling. yes, it's very helpful, but no tool, no matter how good, can tell you how to migrate a monolithic service into microseconds microservices with no downtime and no data loss. maybe if we didn't fuck up the original implementation such things wouldn't be needed but there's still trillions of lines of code out there and they aren't going to rewrite themselves. Google and Facebook aren't going to be overthrown by some hot new IDE with autocode built in.
secondly.. 5% is nothing. our tooling is so effing slow it needs a 1000x improvement. not in how fast we can write code, just to compile and test the damn stuff.
speaking of, amazons reasoning tech sounds interesting.
at best this tech will cut out some L3 maybe L4 jobs, but L5+ isn't writing code anymore. which in itself will be interesting because new grads are hired at L3.. how will they ever get hired if those slots are tightened up?
- 8note 5y agoI think of interns as L3 and new grads as L4. I love having L4s around. They don't have distractions so they can get stuff done
- 8n4vidtmkvmk 5y agoMaybe it's different at my company. I don't know what interns are classed as... not sure if they're L2 or L3-intern. But people that are fresh out of uni are always L3. A few years of industry experience is L4.
- jstanley 5y agoIf a 5% speed improvement represents 0.0000001% of your time, then isn't 1000x speed improvement still only 0.002%? I don't understand how you can simultaneously argue that almost 0 of your time is spent waiting on tooling, and also that your tooling is so slow that it needs to be 1000x faster.
- 8n4vidtmkvmk 5y agoAlright, let's do some legit math here. Compile times are often about 15 minutes, so a 5% improvement would shave off 45 seconds. That's not too bad. But because that's still a very long wait, I'm not gonna twiddle my thumbs for 14 minutes and 15 seconds, I'm going to go do something else. Probably some social site to kill a few minutes, maybe check some emails, who knows what. Then I'll come back 20-30 minutes later to check if my tests passed or failed. So that 45 seconds isn't really saving me anything because I've already moved onto other things. We have to get that down to a few seconds to avoid distractions. 15 minute / 1000 brings it down to less than a second. And then my other point was that depending on my day, I might not be writing any code at all. For the past few weeks I've just been reading and writing documents. So no amount of speed improvements will help me.
- gravypod 5y ago> our tooling is so effing slow it needs a 1000x improvement. not in how fast we can write code, just to compile and test the damn stuff Not just that. Why do we have 30 tools for 30 programming languages. JetBrains has one IDE per language (approximately). What happens when I have a project with two languages? It is a mess. Don't get me started about build systems. Something like Bazel is greatly needed which would help with the performance and metadata we can extract but we're decades away from demonstrating this to the average programmer.
- 8n4vidtmkvmk 5y agoYa.. I wasn't so sure about 'Bazel' at first but it really is great having one build system for every language that mostly just works once you get used to it.
- blindmute 5y agoJetbrains has one IDE (IDEA Ultimate) and redistributes it with various built in plugins for cheaper licensing and marketing. If you get the more expensive Ultimate, it has support for all languages via the same plugins.