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

541–550 of 944 posts

Re: OpenAI O3-Mini

#541

Post by jumploops »

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.

LLMs are the new compilers.

As a student, you should continue to focus on fundamentals, but also adapt LLMs into your workflow where you can.

Skip writing the assembly (now curly braces and semicolons), and focus on what the software you’re building actually does, who it serves, and how it works.

Programming is both changing a lot, and not at all. The mechanics may look different, but the purpose is still the same: effectively telling computers what to do.

Re: OpenAI O3-Mini

#542

Post by abdullahkhalids »

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.

As a former prof. What you should be learning from any STEM degree (and many other degrees as well) is to think clearly, rigorously, creatively, and with discipline, etc. You also need to learn the skill of learning content and skills quickly.

The specific contents or skills of your degree don't matter that much. In pretty much any STEM field, over the last 100ish years, whatever you learned in your undergraduate was mostly irrelevant by the time you retired.

Everyone got by, by staying on top of the new developments in the field and doing them. With AI, the particular skills needed to use the power of computers to do things in the world have changed. Just learn those skills.

Re: OpenAI O3-Mini

#543

Post by BoorishBears »

chad1n wrote: post #244
othello wrote: post #201

Earlier quoted context omitted.

Would you have specific recommendations of such vendors?

For example, ` https://deepinfra.com/ ` which asks for $2.5 per million on output or https://nebius.com which asks for $2.4 per million output tokens.

As the sibling comment mentions, you're not getting anything production grade for less than $7 per million and that's on input and output.

Nebius is single digit TPS. 31 seconds to reply to "What's 1+1".

Hopefully Deepseek will make it out of their current situation because in a very ironic way, the thing the entire market lost its mind over is not actually usable at the pricing that drove the hype: https://openrouter.ai/deepseek/deepseek-r1

Re: OpenAI O3-Mini

#544

Post by BoorishBears »

druskacik wrote: post #248
othello wrote: post #201

Earlier quoted context omitted.

Would you have specific recommendations of such vendors?

Well, it's $2.19 per million output tokens even directly on deepseek platform. https://api-docs.deepseek.com/quick_start/pricing/

Their API platform has been down for 48 hours at this point

Re: OpenAI O3-Mini

#545

Post by MaxLeiter »

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,…

We've been working on solving a lot of these issues with v0.dev (disclaimer: shadcn and I work on it). We do a lot of pre and post-processing to ensure LLMs output valid shadcn code.

We're also talking to the cursor/windsurf/zed folks on how we can improve Next.js and shadcn in the editors (maybe something like llms.txt?)

Re: OpenAI O3-Mini

#546

Post by kristopolous »

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" and other function words; usually work fine today but if I'm having trouble, the best thing to do is probably be inclusive, not exclusive.

Re: OpenAI O3-Mini

#547

Post by mvkel »

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...

My point was that a prompt that simple could be held and executed very well by sonnet, but all other models (especially reasoning models) crash and burn.

It's a 15 line tsx file so context shouldn't be an issue.

Makes me wonder if reasoning models are really proper models for coding in existing codebases

Re: OpenAI O3-Mini

#548

Post by mvkel »

MaxLeiter wrote: post #545
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,…

We've been working on solving a lot of these issues with v0.dev (disclaimer: shadcn and I work on it). We do a lot of pre and post-processing to ensure LLMs output valid shadcn code. We're also talking to the cursor/windsurf/zed folks on how we can improve Next.js and shadcn in the editors (maybe something like llms.txt?)

Thanks for all the work you do! v0 is magical. I absolutely love the feature where I can add a chunky component that v0 made to my repo with npx

Re: OpenAI O3-Mini

#550

Post by bearjaws »

orbital-decay wrote: post #478
cheema33 wrote: post #462

Earlier quoted context omitted.

I like deepseek a lot. But they are currently very glitchy. The API service goes up and down a lot. Maybe they'll sort that out soon.

Apparently they're under a very targeted DDoS for almost a month, with technical details shared in Chinese but very little discussion in English. Which is surprising, it's not like major AI products are getting DDoSed out of existence every day.

Almost all of thm are protected by cloudflare if you look.

My guess is Deepseek didn't implement anti-DDOS until way too late.

Post reply on HN