281–290 of 323 posts
Earlier quoted context omitted.
True, this needs to be fixed. What I wanted to express was that I didn’t do any of the implementation , that is, any logic . I need to phrase this better.
Wait, I just saw what you meant: so no, for the Python tool my message stands. I did not write any code for it myself.
Earlier quoted context omitted.
I disagree, I think it's absolutely astounding that they've gotten this good in such a short time, and I think we'll get better models in the near future. By the way, prompting models properly helps a lot for generating good code. They get lazy if you don't explicitly ask for well-written code (or put that in the system prompt). It also helps immensely to have two contexts, one that generates the code and one that re…
> They get lazy if you don't explicitly ask for well-written code (or put that in the system prompt). This is insane on so many levels.
I agree with the author here, but my worry is that by leaning on the LLMs, the very experience that allows me to uniquely leverage the LLMs now will start to atrophy and in a few years time I'll be relying on them just to keep up.
Earlier quoted context omitted.
Like AI could write code perfectly as soon as I thought of it, and that would not improve my productivity 10x. Coding was never the slow part. Everything that goes around coding (like determining that the extra load here is not going to overload things, getting PMs to actually make their mind up what the feature is going to do, etc.), means that there's simply not that much time to be saved on coding activities.
Same argument can be said for not using any tooling really. "Tech is the easy part". No difference typing code on notepad and having zero process/engineering infrastructure I guess. Because stakeholder management is the main engineering skill apparently. Btw, AI doesn't just code, there are AIs for debugging, monitoring etc too.
1. Tooling obviously does improve performance, but not so huge a margin. Yes, if AI could automate more elements of tooling, that would very much help. If I could tell an AI "bisect this bug, across all projects in our system, starting with this known-bad point", that would be very helpful -- sometimes. And I'm sure we'll get there soon enough. But there is fractal complexity here: what if isolating the bug requires stepping into LLDB, or dumping some object code, or running with certain stressors on certain hardware? So it's not clear that "LLM can produce code from specs, given tight oversight" will map (soon) to "LLM can independently assemble tools together and agentically do what I need done".
2. Even if all tooling were automated, there's still going to be stuff left over. Can the LLM draft architectural specs, reach out to other teams (or their LLMs), sit in meetings and piece together the big picture, sus out what the execs really want us to be working on, etc.? I do spend a significant (double-digit) percentage of my time working on that, so if you eliminate everything else -- then you could get 10x improvement, but going beyond that would start to run up against Amdahl's Law.
Earlier quoted context omitted.
Photoshop is a tool designed for the 1% of people who want that level of control for their vision. Adobe has several other tools for other markets. Even the latest model from this week, which is undeniably impressive, can’t get close to the level of control that photoshop gives me. It often edits parts of the image I haven’t asked it to touch among other issues. I use photoshop as a former photojournalist, and AI man…
This rhymes with still developing your own film. I'm really eager to see how this pans out in a decade.
Well, guilty, I actually do occasionally develop my own film.
Film photography is actually expanding as an industry right now. We are well past the point where digital photography can do everything a film camera can do, and in most cases it can do it far better (very minor exceptions like large format photography still exist, where you can argue that film still has the edge).
I think that whether you embrace AI photo editing or not has more to do with the purpose of your photos. If you are trying to create marketing collateral for a valentines day ad campaign, AI is probably going to be the best tool. If you are trying to document reality, even for aesthetic purposes, AI isn't great. When I make a portrait of my wife, I don't need AI to reinterpret her face for me.
Earlier quoted context omitted.
Wait, I just saw what you meant: so no, for the Python tool my message stands. I did not write any code for it myself.
Not OP but did you edit the code via prompts, or was the whole thing a one-shot? That particular aspect is very confusing to me, I think you should clarify it.
Sometimes I liken the promise of AI to my experience with stereoscopic images (I have never been able to perceive them) -- I know there's something there but I frequently don't get it.
Earlier quoted context omitted.
Apologies for implying you are claiming to be an expert software engineer: I took the "senior" in the title and "25 years of experience" in the post to mean similar things as "expert". I don't doubt this is doing something useful for you. It might even be mostly correct. But it is not a positive advertisement for what AI can do: just like the code is objectively crap, you can't easily trust the output without a compr…
Because I wanted exactly this experience: can I get to the desired result — functionality-wise, if not code-wise! — even if I choose the stack that makes sense in terms of technology, not the one that I happen to be proficient in? And if I cannot bring language-proficiency to the table — which of my capabilities as a seasoned software&systems guy can I put to use? In the brown-field projects where my team and I have…
However, your writing style implied that the result was somehow better because you were otherwise an experienced engineer.
Even your clarification in the post sits right below your statement how your experience made this very smooth, with no explanation that you were going to be happy with bad code as long as it works.
Earlier quoted context omitted.
It is 5x if you are already a senior SE knowing your programming language really well, constantly suggesting good architecture yourself ("seed files" is a brilliant idea), and not accepting any slop / asking to rewrite things if something is not up to your standards (of course, every piece of code should be reviewed). Otherwise, it can be 0.2x in some cases. And you should not use LLMs for anything security-related u…
> It is 5x if you are already a senior SE knowing your programming language really well, constantly suggesting good architecture yourself ("seed files" is a brilliant idea), and not accepting any slop / asking to rewrite things if something is not up to your standards (of course, every piece of code should be reviewed). If you know the programming language really well, that usually means you know what libraries are u…
Everything boring can be automated and it takes five seconds compared to half an hour.
Rob Pike has the right idea but the wrong execution. As the amount of second and third party code we use increases, the search time goes up, and we need better facilities to reduce the amount of time you need to spend looking at the internals of one package because you need that time to look at three others. So clarity and discoverability both need to matter, and AI has no answers here, only more problems.
IMO, a lot of the success of Java comes from having provided 80% of the source code with the JDK. You could spend so much time single stepping into code that was not yours to figure out why your inputs didn’t cause the outputs you expected. But those are table stakes now.