GPT-5.6 Sol Ultra will be in Codex • Hacker News Board

Re: GPT-5.6 Sol Ultra will be in Codex

#361

Post by satvikpendem » Mon, Jul 06, 2026, 4:36 PM UTC

reinitctxoffset wrote: post #273

solenoid0937 wrote: post #231

Earlier quoted context omitted.

It is in OpenAI, Anthropic, and the US Gov's best interest to slow China down and ban Chinese models. Literally none of what you wrote prevents them from doing so. Once China starts to get scary, Commerce will export control GPUs and declare Chinese models "foreign munitions." Any nation doing business with the US will not be allowed to use these models either, and that will be the end of that. It is just not in the…

The assertion that it's in the United States Government's best interests ban Chinese open weight models is a very strong opinion that is not a consensus even at the fringe of Thiel-adjacent psycho thought: Alex Karp is on record about open weight models being necessary, the fucking "we bombed a bunch of kids with Claude doing rubber stamp target selection" guy. He thinks "trust OpenAI and Anthropic" is a radical posi…

Again, none of what you're saying will make the US government, especially the current administration, care, as they have not cared about many things with consequences either; you're applying logic where there may not be any heed to it.

As well, it is a false equivalence to say that local models are only Chinese and otherwise we would use cloud models, but there are American or European ones, so a ban would simply force companies to use these, even if they are inferior to Chinese ones. It's simply a matter of national security to the US government, and they will not care what random people in media say.

Re: GPT-5.6 Sol Ultra will be in Codex

#362

Post by solenoid0937 » Mon, Jul 06, 2026, 4:45 PM UTC

petesergeant wrote: post #278

solenoid0937 wrote: post #141

Earlier quoted context omitted.

HN is just a massive Anthropic hate fest now, probably funded/manipulated by OAI's $8B PR budget. OP phrases it as a bad thing that Dario is keeping compute multipliers to Anthropic. How naive can one be? Compute multipliers are the whole business. Those are the trade secrets every lab is built on. It is the alpha of the business. How does protecting this make Dario evil? This website is getting out of hand with the…

I'm not as conspiratorial as you, but it does seem the tide of opinion here is turning against Amodei, for no particularly obvious external reason. At the same time, there does seem to be at least some evidence of adversarial attempts to oppose data-centers by America's competitors.

At my company we are pretty certain that a lot of datacenter opposition is being instrumented and funded by the PRC.

AI is a bid for control of all humanity. The first to superintelligence owns our future. I'd say it's okay to be a bit suspicious or conspiratorial of anti-AI/anti-lab narratives.

Re: GPT-5.6 Sol Ultra will be in Codex

#363

Post by szundi » Mon, Jul 06, 2026, 4:45 PM UTC

alightsoul wrote: post #294

schnitzelstoat wrote: post #202

Earlier quoted context omitted.

It gives them a massive advantage because they can cut the cost per token by a lot and eat Anthropic's market share. In what universe is any company going to give that advantage away? In any case if they take away a lot of market share it's basically the same in the end - most people will be using these optimisations.

Deepseek gave it away with R1. They did it again with V4 and DSpark. You don't use it because it's Chinese

[dead]

Re: GPT-5.6 Sol Ultra will be in Codex

#364

Post by deercs » Mon, Jul 06, 2026, 4:58 PM UTC

fny wrote: post #348

d4rkp4ttern wrote: post #259

Earlier quoted context omitted.

> single line addition to prompt to use subagents proactively. This misses an important detail. In Claude Code [1], ultracode suggests the agent create a JavaScript code to deterministically orchestrate sub agents. This is different from just having the main agent launch sub agents and (non-deterministically) manage them. The resulting workflow is called “dynamic” because CC creates this orchestration script dynamica…

Can you explain what you find useful about ultracode? I've become wary of agent swarms since the early days and now just prefer to have a single agent spin for hours at time. Parallelism never got me anywhere worthwhile.

I got a security scan by route for one of our services that was pretty good and raised some useful results. Uses are few and far between for me tho tbh, also really expensive. I usually run a single agent with a subagent or two MAX

Re: GPT-5.6 Sol Ultra will be in Codex

#365

Post by solenoid0937 » Mon, Jul 06, 2026, 5:03 PM UTC

bigyabai wrote: post #355

solenoid0937 wrote: post #227

Earlier quoted context omitted.

The only reason Chinese labs release their research is because they are behind. They lose nothing by doing so because the alpha has already been discovered. This is not hard to understand. Do you really think DeepSeek would publish their algorithms if they led the American companies? Lmao.

Except this is clearly not true. Opus-tier models like GLM 5.2 have their weights freely available. Even when they're reaching parity with American models, Z.ai, Qwen and Deepseek are upholding their end of the bargain. I'd criticize them too, if they were due any scrutiny.

I use GLM 5.2 daily. Calling it Opus tier is a bit of a misnomer. It is maybe comparable to Sonnet 5, definitely not Opus 4.8.

It is a great model for the price, but it has much worse autonomy and long context performance, and I don't trust it for anything beyond personal projects, whereas I use Opus/Fable and GPT for work.

Re: GPT-5.6 Sol Ultra will be in Codex

#366

Post by solenoid0937 » Mon, Jul 06, 2026, 5:04 PM UTC

stingraycharles wrote: post #330

Der_Einzige wrote: post #300

Earlier quoted context omitted.

The fact that they don’t release model weights for free to download on huggingface means that Sam, Mira, Dario, etc are ontologically evil and may they all reincarnate as either durian fruits or cockroaches, ideally as durian fruit infested with cockroaches…

Is this a serious take or sarcasm?

Sometimes you just gonna channel your inner Karp

Re: GPT-5.6 Sol Ultra will be in Codex

#367

Post by nullbio » Mon, Jul 06, 2026, 5:07 PM UTC

internet2000 wrote: post #79

Hope this forces Anthropic to be less stingy with Fable.

Hopefully they aren't, and their business dies. They're not a good company.

Re: GPT-5.6 Sol Ultra will be in Codex

#368

Post by HDThoreaun » Mon, Jul 06, 2026, 5:08 PM UTC

jvuygbbkuurx wrote: post #113

make3 wrote: post #110

Earlier quoted context omitted.

I don't think that people relying on the tools too much is the first sign I would identify to mean the bubble is popping

The market is priced at expecting AGI levels of breakthroughs. Just a very useful tool for programming is definitely not enough to keep the music playing.

If the market thought AGI was imminent the labs would be worth hundreds of trillions of dollars. AGI is one step away from fully automated economy.

Re: GPT-5.6 Sol Ultra will be in Codex

#369

Post by rowanG077 » Mon, Jul 06, 2026, 5:09 PM UTC

embedding-shape wrote: post #253

rowanG077 wrote: post #241

Earlier quoted context omitted.

The big problem is that a person making a mistake can be taught to not make that mistake again. That's also not foolproof but at least it works a lot of the times. AI are unteachable, if you have given them a good prompt and they do something wrong 90% of the time you are shit out of luck. That is to say I do agree that building reliable processes out of unreliable parts with feedback is the modus operandi. However A…

> The big problem is that a person making a mistake can be taught to not make that mistake again. That's also not foolproof but at least it works a lot of the times. AI are unteachable, if you have given them a good prompt and they do something wrong 90% of the time you are shit out of luck. I feel like this line of thinking is kind of an unfair comparison. I'm not saying LLMs are magical beings that can suddenly lea…

I'm not against the prompt being changed, the point I was making is that an LLM is prone to the exact same mistakes even if you change the prompt. A trivial example is the very basic character counting mistake, I just asked chatgpt:

> How many p are in strawperry?

> There are 0 “p”s in strawperry.

And I can trigger the same mistake with various words even when adjusting the prompt many times. So I cannot teach chatgpt to correctly count characters.

Re: GPT-5.6 Sol Ultra will be in Codex

#370

Post by alightsoul » Mon, Jul 06, 2026, 5:09 PM UTC

estearum wrote: post #351

alightsoul wrote: post #276

Earlier quoted context omitted.

So how can Deepseek publish them without killing themselves?

Because they are a state-owned enterprise tasked with destroying the aforementioned alpha of private American labs...?

oh here we go again saying everything in china is done by the state. it's not a democracy but they're not the soviet union either. if that were true, why don't they have just one model initiative instead of several?