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

DeepSeek V4 Flash 0731

arcprize.org

Post reply on HN
Print view View original

421–430 of 474 posts

Re: DeepSeek V4 Flash 0731

#421

Post by solenoid0937 »

spaceman_2020 wrote: post #339
FernandoTN wrote: post #194

Earlier quoted context omitted.

I think you're overlooking the fact that for long-horizon tasks, even small errors compound over time and can lead to catastrophic outcomes. For simple queries, we have reached the threshold since the beginning of the year, and models are good enough from every provider to make a meaningful difference between one another. (ChatGPT, Claude, Gemini, Grok, MuseSpark, Kimi, DeepSeek, GLM...) The real unlock will be, and…

Majority of white collar work absolutely does not require sota models

White collar work will require SOTA models up until the point where said models can automate white collar work entirely. Then, we might finally see a meaningful commoditization crunch.

Re: DeepSeek V4 Flash 0731

#424

Post by stymaar »

solenoid0937 wrote: post #417
stymaar wrote: post #415

Earlier quoted context omitted.

> just a very last-gen way of using agents. Fable has only been out for a month but somehow everyone is supposed to have moved to a completely different way of working that supposedly only works for Fable and nothing else…

This stuff is just obvious to anyone working with the latest models. Fully autonomous agents are a game changer. Having to pair program with one is indeed "last gen", I haven't done that for a month and I won't ever be doing that again in my life, outside of personal projects.

> This stuff is just obvious to anyone working with the latest models. Fully autonomous agents are a game changer.

This kind of takes makes me cringe. Why don't you go back to LinkedIn?

I have no idea what you mean by “pair program with an agent”, but Opus have been able of autonomous coding since last November, and with any half-decent harness even local Qwen3.5 was able to do so 6 months ago.

Fable is a stronger model, which means it can solve harder tasks but it's also over-hyped, because only a small fraction of task is hard enough to be Fable-worthy.

Re: DeepSeek V4 Flash 0731

#426

Post by WarmWash »

DiogenesKynikos wrote: post #336
WarmWash wrote: post #216

Earlier quoted context omitted.

If it's hosted in China, they can tell you whatever you want to hear and do whatever they want to do. What are you going to do? Take a CCP company in front of a CCP judge?

You probably could sue them if they really did store your information without your consent. It would be a hassle, but China does have privacy laws. Companies do get sued for violating them.[0] 0. https://www.chinajusticeobserver.com/a/china%E2%80%99s-top-c...

China has laws that serve the state, not the individual. They are there for the party to use to prosecute. So you could file, but considering the state is the one responsible for holding the data, and the one that controls the courts, it's not going to go anywhere.

Remember China is still an authoritarian dictatorship. One leader with absolute power for life. Don't let the facade misguide you.

Re: DeepSeek V4 Flash 0731

#427

Post by solenoid0937 »

stymaar wrote: post #424
solenoid0937 wrote: post #417

Earlier quoted context omitted.

This stuff is just obvious to anyone working with the latest models. Fully autonomous agents are a game changer. Having to pair program with one is indeed "last gen", I haven't done that for a month and I won't ever be doing that again in my life, outside of personal projects.

> This stuff is just obvious to anyone working with the latest models. Fully autonomous agents are a game changer. This kind of takes makes me cringe. Why don't you go back to LinkedIn? I have no idea what you mean by “pair program with an agent”, but Opus have been able of autonomous coding since last November, and with any half-decent harness even local Qwen3.5 was able to do so 6 months ago. Fable is a stronger mo…

I use Kimi K3, Opus 5, and Fable on a daily basis.

Fable is the only one that reliably one shots complex changes and makes the right design choices. Everything else requires handholding.

I can let Fable loose on a 12+ hour (for AI) task and it will have performed it flawlessly when I come back the next day. K3 and Opus are not like this.

And no, our harness is not the limiting factor here.

Re: DeepSeek V4 Flash 0731

#428

Post by DiogenesKynikos »

WarmWash wrote: post #426
DiogenesKynikos wrote: post #336

Earlier quoted context omitted.

You probably could sue them if they really did store your information without your consent. It would be a hassle, but China does have privacy laws. Companies do get sued for violating them.[0] 0. https://www.chinajusticeobserver.com/a/china%E2%80%99s-top-c...

China has laws that serve the state, not the individual. They are there for the party to use to prosecute. So you could file, but considering the state is the one responsible for holding the data, and the one that controls the courts, it's not going to go anywhere. Remember China is still an authoritarian dictatorship. One leader with absolute power for life. Don't let the facade misguide you.

The case mentioned in the article I linked to involved a private citizen suing a tech company, based on China's GDPR-like laws. He won.

Re: DeepSeek V4 Flash 0731

#430

Post by thebytefairy »

LeBit wrote: post #291

I always find it confusing that a meaningful volume of the comments are saying "this reached parity with SOTA models. Best $/task." And a meaningful chunk of the comments are saying "this piece of garbage isn’t even at the level of gpt-oss 20B".

Definitely interesting. My guess it has to do with the level of hand-holding the users are doing. Ie. Full on vibing vs pair programming.
Post reply on HN