If you are a dev and can't get value out of AI, you are not a senior dev. You are just a dev who doesn't know what the tool is for yet. In the real world, no one is going to drop a senior dev onto a project where every line of code is "muscle memory." They drop you on stalled projects that combine a dozen pieces of proprietary tech that you've never seen before. If you can't make use of an AI that can analyze and cross reference all the manuals in a minute, you're just a bad dev.
Too bad--you've already been fired and replaced with a subscription-based LLM service. Our new codebase is composed entirely of detailed descriptions of anime girls taking their socks off. This is also how the new automated help line answers all queries.
While I do worry the exec level push to use LLMs will negatively impact the job market, I also really don't want to work anywhere that puts that much reliance on an objectively bad pattern.
It's usually useful as a search engine. Quicker to ask it something about config options than looking through the docs. Except when it begins hallucinating or outright lying, then you have to fall back on the actual docs.
It's a useful tool a lot of the time, but sometimes it's more of a waste of time trying to get it to do what you want and explaining to it over and over that what it's trying to do is now allowable in the specifications of the language or tool. Then having it say you're right, and don't another completely wrong thing, and then being cocky enough to write "why this fix works" after every wrong fucking thing or writes.
I let it stub out my unit tests and enhance intellisense for one line completions, that's it.
If you're a Jr dev or just getting into coding right now learning to actually code rather than relying heavily on llms is going to be money in the bank 5-10 years down the road when there is a massive demand for competent devs as the last of the old guard seniors are promoted/retired out and there is a tonne of AI tech debt to dig out of.
Exactly. In the long run companies are going to find that LLMs MIGHT be able to replace low level stuff, but they are going to need real human experience when the house of cards collapses.
It won't necessarily be our coding skills, but our ability to debug, investigate, and understand that will be our key skill.
Because it slows things down and breaks concentration. Because it can't retain enough context to be helpful. Because it will give you patterns that have better options these days. Because you still have to go through all the code to understand it. Because there are often syntax or logic issues you have to find and fix.
Outside of simple and small things it's just not trustworthy and for those things, any competent coder can do those things faster and with less effort (e.g. muscle memory).
Unfortunate500
If you are a dev and can't get value out of AI, you are not a senior dev. You are just a dev who doesn't know what the tool is for yet.
In the real world, no one is going to drop a senior dev onto a project where every line of code is "muscle memory." They drop you on stalled projects that combine a dozen pieces of proprietary tech that you've never seen before.
If you can't make use of an AI that can analyze and cross reference all the manuals in a minute, you're just a bad dev.
PaleChapter
Too bad--you've already been fired and replaced with a subscription-based LLM service. Our new codebase is composed entirely of detailed descriptions of anime girls taking their socks off. This is also how the new automated help line answers all queries.
iamgnat
While I do worry the exec level push to use LLMs will negatively impact the job market, I also really don't want to work anywhere that puts that much reliance on an objectively bad pattern.
grokitman
It's usually useful as a search engine. Quicker to ask it something about config options than looking through the docs. Except when it begins hallucinating or outright lying, then you have to fall back on the actual docs.
Kyrorayne
Because AI is shit and breaks code
UWAGAGABLAGABLAGABA
It's a useful tool a lot of the time, but sometimes it's more of a waste of time trying to get it to do what you want and explaining to it over and over that what it's trying to do is now allowable in the specifications of the language or tool. Then having it say you're right, and don't another completely wrong thing, and then being cocky enough to write "why this fix works" after every wrong fucking thing or writes.
OriginalSyn
I let it stub out my unit tests and enhance intellisense for one line completions, that's it.
If you're a Jr dev or just getting into coding right now learning to actually code rather than relying heavily on llms is going to be money in the bank 5-10 years down the road when there is a massive demand for competent devs as the last of the old guard seniors are promoted/retired out and there is a tonne of AI tech debt to dig out of.
iamgnat
Exactly. In the long run companies are going to find that LLMs MIGHT be able to replace low level stuff, but they are going to need real human experience when the house of cards collapses.
It won't necessarily be our coding skills, but our ability to debug, investigate, and understand that will be our key skill.
anarchoFeline
intellisense? found the c#/.net guy
OriginalSyn
Not for decades but I haven't heard a better name for it
LitterBoxKing
Have you bothered to ask why that is?
iamgnat
Because it slows things down and breaks concentration. Because it can't retain enough context to be helpful. Because it will give you patterns that have better options these days. Because you still have to go through all the code to understand it. Because there are often syntax or logic issues you have to find and fix.
Outside of simple and small things it's just not trustworthy and for those things, any competent coder can do those things faster and with less effort (e.g. muscle memory).