IT/Software career thread: Invert binary trees for dollars.

TJT said:

I generally agree that knowing syntax has little value.

But understanding data structures, processing dynamics like threading, memory management, artifact (package storage), branching strategy for collaboration, distributed architecture, and most other design and computer science principles are just as critical. Also knowing what features a programming language provides, like in your recent Rust vs Python vibe code project.

Just because you do not know the syntax to express them in some given language does not change the fact that you need to understand the issues itself. Suppose you vibe code something that solves the right now problem, but cannot scale because of a Big O notation issue with dynamic nested looping or poor memory management. AI will not solve that immediately because you did not present that as a requirement for design. If you lack the ability to even understand or describe the issue because you do not know foundational computer science principles then AI cannot help you resolve it.

Click to expand...

I do agree with you. It's tough to step back and think how much of my knowledge I'm just inserting into the vibe coding because it's a part of me vs someone completely foreign to engineering asking it to do something.

The reason I switched from Python to rust was exactly because of that. I didn't know the syntax but I did know rusts benefits.

TJT said:

I generally agree that knowing syntax has little value.

But understanding data structures, processing dynamics like threading, memory management, artifact (package storage), branching strategy for collaboration, distributed architecture, and most other design and computer science principles are just as critical. Also knowing what features a programming language provides, like in your recent Rust vs Python vibe code project.

Just because you do not know the syntax to express them in some given language does not change the fact that you need to understand the issues itself. Suppose you vibe code something that solves the right now problem, but cannot scale because of a Big O notation issue with dynamic nested looping or poor memory management. AI will not solve that immediately because you did not present that as a requirement for design. If you lack the ability to even understand or describe the issue because you do not know foundational computer science principles then AI cannot help you resolve it.

Click to expand...

I agree that someone needs to have a deeper understanding of these things. I just don't think every developer needs that same depth of knowledge anymore. The people who have the interest and the mind for it will still dig into how everything works. That expertise still has value, but I think AI is making it less of a requirement for the average developer.

I look at it sort of like cars. At one time, knowing how to fix your own car was a pretty big advantage, and sometimes a necessity. These days, the average driver gets along just fine without understanding everything going on under the hood. We still need mechanics and engineers who understand those things, but everyone driving a car doesn't need to be one. I think software development is moving in a similar direction.

Noodleface said:

Knowing how things like pointers work, memory management, etc. and the actual architecture of software for instance is going to become extremely important.

Click to expand...

Important now, but at the rate of advancement, I wonder how much longer any technical knowledge is going to matter. The difference between the beginning of the year until now in what you can successfully have ai build by giving it a few vague instructions is crazy. Maybe not everything yet, but everything actually kind of seems in scope now.

As a side rustle, i had a few quota % to burn last night and asked opus max to review an art pipeline to see if we could improve it. Instead of actually doing that, it found a few clips made with H3, decided that it disagreed with our previous read of the license/terms and that we shouldn't use it even for experiments, then spent over an hour searching through other projects and licenses, decided that it also didn't like qwen 2.1 and starts chasing down anything "tainted" by either...

I mean ok, if you're going to use things commercially, you should be clear on the licenses of course, but Anthropic all of the sudden throwing an unprompted turbo fit about a chinese open model's license? (instead of actually improving the workflow as asked so that maybe we wouldn't need chinese models in the first place... also, obligatory, "NOW you care about ip?")


Also, in case it's relevant for anyone, this "disagreement with prior license evaluations" was:
qwen 2.1 - whether or not output used for training is "commercial"
h3 - community license doesn't cover the US and disallows use of outputs there, but also says that it claims no ownership or rights to outputs

Control said:

Important now, but at the rate of advancement, I wonder how much longer any technical knowledge is going to matter. The difference between the beginning of the year until now in what you can successfully have ai build by giving it a few vague instructions is crazy. Maybe not everything yet, but everything actually kind of seems in scope now.

As a side rustle, i had a few quota % to burn last night and asked opus max to review an art pipeline to see if we could improve it. Instead of actually doing that, it found a few clips made with H3, decided that it disagreed with our previous read of the license/terms and that we shouldn't use it even for experiments, then spent over an hour searching through other projects and licenses, decided that it also didn't like qwen 2.1 and starts chasing down anything "tainted" by either...

I mean ok, if you're going to use things commercially, you should be clear on the licenses of course, but Anthropic all of the sudden throwing an unprompted turbo fit about a chinese open model's license? (instead of actually improving the workflow as asked so that maybe we wouldn't need chinese models in the first place... also, obligatory, "NOW you care about ip?")

Also, in case it's relevant for anyone, this "disagreement with prior license evaluations" was:
qwen 2.1 - whether or not output used for training is "commercial"
h3 - community license doesn't cover the US and disallows use of outputs there, but also says that it claims no ownership or rights to outputs

Click to expand...

I'm just trying to cling onto anything that makes me employable now man. It's scary

I came across this video, and it is relevant to our recent discussions here: