strangescript

So many people on their high horse when Nexus attempted to do collections, as if Nexus Mods was their god given place to post their mods for free and control everything about how they are used.

Congratulations! You are going to be so happy moving forward!

What exactly is the alternative? Wish it away? Developers have been automating away jobs for decades, its seems hypocritical to complain about it now.

Everyone is still thinking about this problem the wrong way. If you are still running one agent, on one project at a time, yes, its not going to be all that helpful if you are already a fast, solid coder.

Run three, run five. Prompt with voice annotation. Run them when normally you need a cognitive break. Run them while you watch netflix on another screen. Have them do TDD. Use an orchestrator. So many more options.

I feel like another problem is deep down most developers hate debugging other people's code and thats effectively what this is at times. It doesn't matter if your Associate ran off and saved you 50k lines of typing, you would still rather do it yourself than debug the code.

I would give you grave warnings, telling you the time is nigh, adapt or die, etc, but it doesn't matter. Eventually these agents will be good enough that the results will surpass you even in simple one task at a time mode.

there are plenty of open source LLMs trained on harry potter, is that fine?

I read harry potter, and you ask me about a page, and I can recite it verbatim, did I just commit copyright infringement?

If you look at their finances you will realize that there is nothing left in that company.

Yeah, but why aren't they attacking that problem? Is it just impossible, because it would be a really simple win with regards to coding. I am huge enthusiast, but I am starting to feel a peak.

I have used claude code a ton and I agree, I haven't noticed a single difference since updating. Its summaries I guess a little cleaner, but its has not surprised me at all in ability. I find I am correcting it and re-prompting it as much as I didn't with 3.7 on a typescript codebase. In fact I was kind of shocked how badly it did in a situation where it was editing the wrong file and it never thought to check that more specifically until I forced it to delete all the code and show that nothing changed with regards to what we were looking at.

it feels like openai are at a ceiling with their models, codex1 seems to be another RLHF derivative from the same base model. You can see this in their own self reported o3-high comparison where at 8 tries they converge at the same accuracy.

It also seems very telling they have not mentioned o4-high benchmarks at all. o4-mini exists, so logically there is an o4 full model right?

In open source its super useful to be able to immediately have an idea of how big the model is and what kind of hardware it could potentially run on.

Speed is great, but you have to set the bar a little higher than last year's tiny models

The smaller models have been creeping upward. They don't make headlines because they aren't leapfrogging the mainline models from the big companies, but they are all very capable.

I loaded up a random 12B model on ollama the other day and couldn't believe how good it competent it seemed and how fast it was given the machine I was on. A year or so ago, that would have not been the case.

The 0.6B model is wild. I like to experiment with tiny models, and this thing is the new baseline.

There is more complicated math systems that computers have solved, just like Chess, and Go. Systems that seemed impossible for a machine to beat and eventually they do.

Coding will be exactly the same soon.

Yeah, anytime I am about to do some long multiplication, I start reaching for my calculator and stop, "no, you will go to the multiplication gym, and do this by hand, need to stay sharp"

I would use Aider if it had an agent mode. It needs to catch up with UX, frankly just have a mode that copies what claude code does.

curious why you went with Phi as the default models, that seems a bit unusual compared to current trends

Claude Code still feels superior. o4-mini has all sorts of issues. o3 is better but at that point, you aren't saving money so who cares.

I feel like people are sleeping on Claude Code for one reason or another. Its not cheap, but its by far the best, most consistent experience I have had.

"There's only one component per file"

If you don't do this in React, its your own fault.

I think they have realized that even if OpenAI is first, it won't last long so really its just compute at scale, which is something they already do themselves.

I have been using Claude Code, and its a bit pricey, but has exceeded my expectations so far.

I think its interesting they left out Gemini 2.0 Pro in the benchmarks which I find to be markedly better than flash if you don't mind the spend.

This is just a bad model. I can't believe they released it. Yes it does have few interesting properties, but nothing that justifies the speed or cost when people are running R1 distillations on toasters for nothing.

This is sooooo dependent on where you work and your current outside of work situation (ie, you need the money bad)

That is what 90% of science is really. There aren't a lot of truly "aha" moments where someone discovers something fundamental with little outside influence.

I think its about scope and expectations. I have had some form of AI code completer in my neovim config for 3 years. It works flawlessly and saves me tons of keystrokes. Sure sometimes it suggests the incorrect completion but I just ignore it and keep coding as if it didn't exist. I am talking about line by line, not entire code blocks, but even that it does well at times.

From what I have seen the people that have the most success have AI building something from scratch using well known tooling (read: old tooling).

The problem is that doesn't immediately help most people. We are all stuck in crap jobs with massive, crusty code bases. Its hard for AI because its hard for everyone.

OpenAI is pivoting away from MS. MS also has their own internal AI interests. Need to frame this for investors that doesn't look like we are losing out. "Nadella doesn't believe in AI anymore". Done and done.

AI will get faster and more energy efficient over time. Deploying physical hardware will never improve in any meaningful way that fixes the biggest problem, deploying X amount of things everywhere you need it. Its a non-starter.

Exactly, so many first of its kind moments in HL1. The first time you saw a scripted sequence, like wtf am I playing, this is amazing.

HL2 was more refined, and the art style did help elevate the story telling.

As context sizes get larger (and remain accurate within the size) and speeds increase, especially inference, it will start solving these large complex code bases.

I think people lose sight of how much better it has gotten in just a few years.