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

GPT-5.6 Sol Ultra will be in Codex

twitter.com

Post reply on HN
Print view View original

281–290 of 433 posts

Re: GPT-5.6 Sol Ultra will be in Codex

#281

Post by sscaryterry »

tmountain wrote: post #178
whstl wrote: post #175

Earlier quoted context omitted.

The craziest to me was someone saying “we are using AI in daily processes, now we need to automate”. But of course to some asshole non-technical people it meant asking for their vibe coded bullshit to be merged into production without review and fighting about it.

The craziest for me is companies that sticking stochastic agents into automated business processes and expecting stable/reliable outcomes. Businesses want deterministic processes in the vast majority of cases.

I can argue that by applying multiple stochastic processes, with a human in the loop, that you will (may) converge on something that is deterministic. You use tests/test vectors to prove this.

We're no different to AI. The code we write to solve a particular problem can (and probably does) change from day to day, depending on your "mood", what you had for breakfast, if you've been fighting with your significant other, other problems/human emotions.

Re: GPT-5.6 Sol Ultra will be in Codex

#282

Post by davedx »

gbnwl wrote: post #85
andai wrote: post #4

For context: > Additionally, we’re introducing a new ultra mode that goes beyond the capabilities of a single agent by leveraging subagents to accelerate complex work. https://openai.com/index/previewing-gpt-5-6-sol/ Can someone explain how this compares with Pro? I thought Pro was already something similar.

How is this any different than what we have already? We've had this ability for ages (6+ months, decades in the AI world), you can literally today easily prompt CC or Codex to use subagents to accomplish tasks and they'll do it well. My entire workflow is one top level orchestrator chat creating tickets to dispatch to subagents to implement, and other subagents to verify. Why is this being sold as a new thing? Have H…

I'd love to read more about how to use this workflow. What kind of top level instructions does this actually work with? Is there an article out there with some concrete examples of how to do this effectively?

Re: GPT-5.6 Sol Ultra will be in Codex

#283

Post by HappMacDonald »

throwawayHBKFC wrote: post #219
throw394042 wrote: post #30

I'm working in large US corporation. And I see that I already have access to 5.6-Sol Ultra on my corporate account. I haven't really used it yet. 2 months ago management was showing us scoreboards, praising leaders who used most tokens. Last few weeks, we're getting weekly emails, telling us that whenever we can - we should use cheaper models, and that we should watch the page which shows our tokens usage.

Same here. It's insane that big, conservative tech companies would sign contracts saying "allow our engineers to use however much they want, we'll pay the bill later, no matter what it costs". In any other domain my company would insist on prior permission, and soft usage caps, and hard usage caps, and real-time tracking of actual dollar amounts (not just opaque tokens/credits, not just an after-the-fact view on a da…

This is the first time in history that I have seen the shovel-sellers convince a supermajority of the gold-miners that "amount of gold mined out of the ground" is less of an indicator of profit than "amount of shovels purchased".

Re: GPT-5.6 Sol Ultra will be in Codex

#284

Post by datakan »

adahn wrote: post #166
bloppe wrote: post #143

Earlier quoted context omitted.

"if I don't do it, someone worse will" Of course that's what Dario thinks because that's what every tech CEO thinks. Dario, Sam, Sundar, probably many Chinese CEOs as well. It's what everyone thinks. That's why they're competing so fiercely with one another. That's why they basically make all the same decisions. That's why we need properly open source AI.

It's hard to be certain what each individual thinks. We can do our best to judge based on what they each say and do. And there are significant differences in what each of these individuals have chosen to say and do over the years. The info available to the public makes it seem a lot like Dario's motivations & priorities differ from those of Sam and others. This doesn't seem like the right place to spend my time litig…

What they think is irrelevant. What they do is

Re: GPT-5.6 Sol Ultra will be in Codex

#285

Post by cyanydeez »

sscaryterry wrote: post #281
tmountain wrote: post #178

Earlier quoted context omitted.

The craziest for me is companies that sticking stochastic agents into automated business processes and expecting stable/reliable outcomes. Businesses want deterministic processes in the vast majority of cases.

I can argue that by applying multiple stochastic processes, with a human in the loop , that you will (may) converge on something that is deterministic. You use tests/test vectors to prove this. We're no different to AI. The code we write to solve a particular problem can (and probably does) change from day to day, depending on your "mood", what you had for breakfast, if you've been fighting with your significant othe…

the humans only there to take the blame. you arn't not goingto be a cyborg, nor is your idea of determinism ever going to last the cycle of the agents, no matter how many memorrry layers, skills and other context guardrails you place.

You're there for blame, not much else. The systems are still going to churn garbage, but because it's a business, that business will rather pay less for garbage they can sell than pay you a living wage, eventuially.

Re: GPT-5.6 Sol Ultra will be in Codex

#286

Post by eiifr1 »

wjnc wrote: post #155
noduerme wrote: post #134

Earlier quoted context omitted.

That's kinda hilarious. Pretty soon they might just ask people to write code themselves.

I’m in Finance and learned pretty quickly that to point out the implicit future cost raises based on the cost the LLM-providers need to recoup was unpopular at best (STFU better describes the situation). Running full force into a bear trap.

[dead]

Re: GPT-5.6 Sol Ultra will be in Codex

#287

Post by cyanydeez »

davedx wrote: post #282
gbnwl wrote: post #85

Earlier quoted context omitted.

How is this any different than what we have already? We've had this ability for ages (6+ months, decades in the AI world), you can literally today easily prompt CC or Codex to use subagents to accomplish tasks and they'll do it well. My entire workflow is one top level orchestrator chat creating tickets to dispatch to subagents to implement, and other subagents to verify. Why is this being sold as a new thing? Have H…

I'd love to read more about how to use this workflow. What kind of top level instructions does this actually work with? Is there an article out there with some concrete examples of how to do this effectively?

in opencode, if you directly referencee a subagent keyword or the name of a defined agent, it'll often spawn the agent.

If you don't mention it directly, it's 50/50 whether any given request will invoke a subagent.

The same with tools, skills, etc. No matter how smart these LLMs appear, they rarely do thinking as you expect.

So basically: learn how the harnesses operate, and know the names of the tools they have.

Re: GPT-5.6 Sol Ultra will be in Codex

#288

Post by cyanydeez »

layla5alive wrote: post #26
postalcoder wrote: post #2

I wonder if it's related that that OpenAI has found a way to cut inference costs by half, according to The Information. https://www.theinformation.com/newsletters/ai-agenda/openai-...

https://archive.ph/NEwVz "However, these inference optimizations, which rival Anthropic refers to as “compute multipliers,” are a big focus for all the labs. Anthropic CEO Dario Amodei has been publicly talking about the concept since at least mid-2023, when he said on a podcast that the company limits “the number of people who are aware of a given compute multiplier” because it could give other AI labs a leg up if t…

Keep in mind, these tech leaders have deluded themselves with the infinite jest of "effective altrusim" which can effective ignore any problem today for some imagined future problem that they're solving. So if they have to enslave the human race because there's a super killer asteroid 10 million lightyears away heading towards us, they'll do it, because obviously, saving future humanity is the only thing that makes sense (because only their genius can save the day)!

Re: GPT-5.6 Sol Ultra will be in Codex

#289

Post by cyanydeez »

alightsoul wrote: post #67
layla5alive wrote: post #26

Earlier quoted context omitted.

https://archive.ph/NEwVz "However, these inference optimizations, which rival Anthropic refers to as “compute multipliers,” are a big focus for all the labs. Anthropic CEO Dario Amodei has been publicly talking about the concept since at least mid-2023, when he said on a podcast that the company limits “the number of people who are aware of a given compute multiplier” because it could give other AI labs a leg up if t…

i really hope it's just what Deepseek V4 does. Deepseek V4 is very cheap and highly performant OpenAI tried to pull off the same trade secret thing with RL when they announced o1 and o3, aka "Compute time scaling". Then Deepseek revealed it with Deepseek R1. Could also be something like Deepseek DSpark. Or using diffusion like DiffusionGemma as a draft model. The timing between the release of those, and this article,…

deep down, i suspect they're all just drafting on implementations to llamacpp.

Re: GPT-5.6 Sol Ultra will be in Codex

#290

Post by bananaflag »

Iolaum wrote: post #261
bananaflag wrote: post #256

Earlier quoted context omitted.

Indeed. Any meaningful AGI/ASI will have to have a form of memory / continual learning. Sam Altman said last year that this will be the focus for GPT-6. The whole "soul.md" stuff today is a poor approximation to that. But I wonder whether it will grow into it, like chain of thought prompting grew into reasoning models.

LLM's as a technology - currently - are stateless. The memory layer is controlled by the agent. I m surprised with the gpt-6 reference unless it has to do with vertical integration between the agent and the layer.

Yeah, I am aware of this statelessness.

This is what I was referring to:

https://www.cnbc.com/2025/08/19/sam-altman-on-gpt-6-people-w...

It's marketing speak, but the goal is clearly there, no idea how achievable.

Post reply on HN