Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

OpenAI O3-Mini

openai.com

Post reply on HN
Print view View original

611–620 of 944 posts

Re: OpenAI O3-Mini

#611

Post by aprilthird2021 »

lukan wrote: post #332
lesuorac wrote: post #176

Earlier quoted context omitted.

I really don't think this is true. OpenAI has no moat because they have nothing unique; they're using mostly other people's (like Transformers) architectures and other companies hardware. Their value-prop (moat) is that they've burnt more money than everybody else. That moat is trivially circumvented by lighting a larger pile of money and less trivially by lighting the pile more efficently. OpenAI isn't the only comp…

"OpenAI has no moat because they have nothing unique" It seems they have high quality trainingsdata. And the knowledge to work with it.

They buy most of their data from Scale AI types. It's not any higher quality than is available to any other model farm

Re: OpenAI O3-Mini

#612

Post by aprilthird2021 »

Bjorkbat wrote: post #70

I have to admit I'm kind of surprised by the SWE-bench results. At the highest level of performance o3-mini's CodeForces score is, well, high. I've honestly never really sat down to understand how elo works, all I know is that it scored better than o1, which allegedly as better than ~90% of all competitors on CodeForces. So, you know, o3-mini is pretty good at CodeForces. But it's SWE-bench scores aren't meaningfully…

> My immediate thoughts, CodeForces (and competitive programming in general) is a poor proxy for performance on general software engineering task

Yep. A general software engineering task has a lot of information encoded in it that is either already known to a human or is contextually understood by a human.

A competitive programming task often has to provide all the context as it's not based off an existing product or codebase or technology or paradigm known to the user

Re: OpenAI O3-Mini

#613

Post by mhh__ »

throw83288 wrote: post #526
vok wrote: post #404

Well, o3-mini-high just successfully found the root cause of a seg fault that o1 missed: mistakenly using _mm512_store_si512 for an unaligned store that should have been _mm512_storeu_si512.

How do I avoid the angst about this stuff as a student in computer science? I love this field but frankly I've been at a loss since the rapid development of these models.

It's either over, or giving a lot of idiots false confidence — I meet people somewhat regularly who believe they don't really need to know what they're doing any more. This is probably an arbitrage.

Re: OpenAI O3-Mini

#615

Post by smallerize »

waynecochran wrote: post #576

I just had it convert Swift code to Kotlin and was surprised at how the comment was translated. It "knew" the author of the paper and what is was doing!? That is wild. Swift: // // Double Reflection Algorithm from Table I (page 7) // in Section 4 of https://tinyurl.com/yft2674p // for i in 1 .. Kotlin: // Use the Double Reflection Algorithm (from Wang et al.) to compute subsequent frames. for (i in 1 until N) { val X…

Wow, haven't seen a viglink in a while.

Re: OpenAI O3-Mini

#616

Post by tribeca18 »

cruffle_duffle wrote: post #568
macNchz wrote: post #533

Earlier quoted context omitted.

You can start it with something like: aider --architect --model o1 --editor-model sonnet Then you'll be in "architect" mode, which first prompts o1 to design the solution, then you can accept it and allow sonnet to actually create the diffs. Most of the time your way works well—I use sonnet alone 90% of the time, but the architect mode is really great at getting it unstuck when it can't seem to implement what I want…

I really want to see how apps created this way scale to large codebases. I’m very skeptical they don’t turn into spaghetti messes. Coding is basically just about the most precise way to encapsulate a problem as a solution possible. Taking a loose English description and expanding it into piles of code is always going to be pretty leaky no matter how much these models spit out working code. In my experience you have t…

I use Cursor and Composer in agent mode on a daily basis, and this is basically exactly what happened to me.

After about 3 weeks, things were looking great - but lots of spagetti code was put together, and it never told me what I didn't know. The data & state management architecture I had written was simply just not maintainable (tons of prop drilling, etc). Over time, I basically learned common practices/etc and I'm finding that I have to deal with these problems myself. (how it used to be!)

We're getting close - the best thing I've done is create documentation files with lots of descriptions about the architecture/file structure/state management/packages/etc, but it only goes so far.

We're getting closer, but for right now - we're not there and you have to be really careful with looking over all the changes.

Re: OpenAI O3-Mini

#617

Post by lukas099 »

Terretta wrote: post #528
dwaltrip wrote: post #525

Earlier quoted context omitted.

I haven't tried aider in quite a while, what does it mean to use one model as an architect and another as the implementer?

Aider now has experimental support for using two models to complete each coding task: - An Architect model is asked to describe how to solve the coding problem. - An Editor model is given the Architect’s solution and asked to produce specific code editing instructions to apply those changes to existing source files. Splitting up “code reasoning” and “code editing” in this manner has produced SOTA results on aider’s c…

Probably gonna show a lot of ignorance here, but isn’t that a big part of the difference between our brains and AI? That instead of one system, we are many systems that are kind of sewn together? I secretly think AGI will just be a bunch of different specialized AIs working together.

Re: OpenAI O3-Mini

#618

Post by esperent »

delichon wrote: post #577
torben-friis wrote: post #527

Earlier quoted context omitted.

My experience with cursor and sonnet is that it is relatively good at first tries, but completely misses the plot during corrections. "My attempt at solving the problem contains a test that fails? No problem, let me mock the function I'm testing, so that, rather than actually run, it returns the expected value!" It keeps doing that kind of shenanigans, applying modifications that solve the newly appearing problem whi…

Claude makes a lot of crappy change suggestions, but when you ask "is that a good suggestion?" it's pretty good at judging when it isn't. So that's become standard operating procedure for me. It's difficult to avoid Claude's strong bias for being agreeable. It needs more HAL 9000.

> when you ask "is that a good suggestion?" it's pretty good at judging when it isn't

Basically a poor man's COT.

Re: OpenAI O3-Mini

#619

Post by Terretta »

lukas099 wrote: post #617
Terretta wrote: post #528

Earlier quoted context omitted.

Aider now has experimental support for using two models to complete each coding task: - An Architect model is asked to describe how to solve the coding problem. - An Editor model is given the Architect’s solution and asked to produce specific code editing instructions to apply those changes to existing source files. Splitting up “code reasoning” and “code editing” in this manner has produced SOTA results on aider’s c…

Probably gonna show a lot of ignorance here, but isn’t that a big part of the difference between our brains and AI? That instead of one system, we are many systems that are kind of sewn together? I secretly think AGI will just be a bunch of different specialized AIs working together.

You're in good company in that secret thought.

Have a look at this: https://en.wikipedia.org/wiki/Society_of_Mind

Re: OpenAI O3-Mini

#620

Post by esperent »

zackproser wrote: post #518
mvkel wrote: post #510

I've been using cursor since it launched, sticking almost exclusively to claude-3.5-sonnet because it is incredibly consistent, and rarely loses the plot. As subsequent models have been released, most of which claim to be better at coding, I've switched cursor to it to give them a try. o1, o1-pro, deepseek-r1, and the now o3-mini. All of these models suffer from the exact same "adhd." As an example, in a NextJS app,…

Not trying to be snarky, but the example prompt you provided is about 1/15th the length and detail of prompts I usually send when working with Cursor. I tend to exhaustively detail what I want, including package names and versions because I've been to that movie before...

That sounds exhausting. Wouldn't it be faster to include you package.json in the context?

I sometimes do this (using Cline), plus create a .cline file at project root which I refine over time and which describes both the high level project overview, details of the stack I'm using, and technical details I want each prompt to follow.

Then each actual prompt can be quite short: read files x, y, and z, and make the following changes... where I keep the changes concise and logically connected - basically what I might do for a single pull request.

Post reply on HN