Shaun Prince

Trolling AI never gets old.

@mattshumer_ This is why you don't use YOLO-mode. ALL agent harnesses have the ability to prevent this. OpenAI is NOT responsible for idiots with skill issues.

I'm so angry... the OpenAI team is looking into it, but this feels like something that should happen with GPT-3.5. Not a mid-2026 frontier model on the highest reasoning level.

@AmbRealismX @kr0der Terra is for well structured delegated tasks from sol. Luna is probably great for human interfaces, summarizing things, compaction/compression and any other task that does not require much thinking/reasoning. There's no game here. It's already figured out.

Too many people are hyper focused on Sol. The real game is figuring out how and when to use Terra and Luna. Look at the chart attached and look at every spot on the Sol line below High. Then draw a horizontal line to the left. Challenge yourself to use 2 models at a minimum, then do the same thing with all 3 being required. Figuring out which, where and when is the whole game.

it's been 1 day now - what reasoning level is everyone using for GPT 5.6 Sol?

@kr0der high

@Emmyzaza77 @DavidOndrej1 Yeah, and what I noticed about its behavior in Hermes Agent, it just keeps going. If you just tell it to go through all the tasks and just finish the list, that's exactly what it's going to do.

@Suparious @DavidOndrej1 I use it with with Hermes as my orchestrator and yo! The difference is very visible. Run for way longer I had to raise my tool limit ceiling. Not crazy expensive too imo

first impressions of GPT 5.6 are not good...

@remoteoperating To add context, I was using Hermes Agent from my MacBook to ssh to my new ASUS laptop, and setup NVIDIA PRIME in KDE to get the machine to use both the Intel and NVIDIA GPU together, for power efficiency on battery.

@Suparious Is that cmux

Trolling AI never gets old.

@remoteoperating Yes. I love cmux so much. On windows, I tried several implementations of wmux, but it was not the same. In linux, I can just use raw Ghostty and make it behave like cmux.

@DavidOndrej1 yes. the xhigh (max) and ultra are not going to give better results and will be slow as a snail. I use high, David suggests medium. David is probably right.

avoid using 5.6 Sol Ultra at all cost just stick to Medium

GPT-5.6 Sol on 750 Token/s in Blender. Not sped up. Holy moly. That’s so much more impressive than looking at benchmarks

@DavidOndrej1 You need to use them both the same way that you used to use your Opus and Then GPT-5.5 with your cmux setup. Only thing that changes for you is the model versions. The two different models are specifically good at different things, so that's how to use them.

Fable over 5.6 change my mind.

And now I have to backpedal using Fable to audit what Haiku did to make sure it didn't mess anything up.

I just had an entire session by accident using Claude Haiku, and I was believing that I was using Fable, and I was amazed about how it was getting through all my tasks so quickly.

What do you mean? It was released at 10 a.m. Pacific this morning. If you don't see it in your software, then you need to update it or use a proper agent harness like Hermes agent. Also, if you use homebrew or Linux brew to manage your agent harnesses, you're gonna have to wait a day or 2 because that's how long that shit takes. It's better to just install the thing directly with the instructions that OpenAI provides, and not use homebrew. Or you can just wait a day or 2 for homebrew to catch up. Also, if you live in the UK, you might have problems with your stupid regulations.

@cocktailpeanut us poor plebs who use codex don't get it 5.6, sad.

wait, so is it chatgpt-cli now?

@IamARoosterMAN @JThePhoto @jasperdevs I mostly see people with the desktop app complaining. Also, if you use linuxbrew or homebrew to install codex or claude-code, you'll suffer a 1-2 day delay on new releases.

@JThePhoto @jasperdevs I think the confussion for many is that they have it in Codex & Work already, but not in their regular chatgpt chat area.

so GPT sol is not coming to ChatGPT...

@alive_prolly @jasperdevs It is available. It was on the API since 10AM PST - If you live in Canada or the USA. Software take time to update. If you want real-time access to new models, use a proper agent harness, like Hermes Agent, ect..

@jasperdevs You just outed yourself, it is not yet available in homeless tiers of subscription, yes

@The_Coaxys @jasperdevs Yes. $20/month sub give me sol. If you are in the UK, you are S.O.L.

@jasperdevs Can a plus sub confirms access to sol ?

@digi_cristina Well, it's akin to a 30min-1hr meeting - A huge opportunity to learn, potential opportunity to propose something or even a chance to be offered a role of a lifetime.

Why is lunch with billionaires a big deal?

@ThePrimeagen I remember you complaining that AI coding now is just sitting around watching spinners. Definitely worth a video.

Results are in and I have one strong complaint Grok4.5-Fast is fast enough that it feels like I'm always on, typically I will ask for work and then go program part of a problem that I want to solve. It's just so fast it's hard to have that loop.

@Star_Knight12 Statistically, it is normal. If you chart all the model releases together on a graph, you can see that it's a mathematically predictable trajectory. It's almost like Moore's Law. I can't understand how this progressively keeps getting faster. I think that's what's happening.

a new AI model is dropping every 12 hours now. this is not normal

@MatthewBerman Thank you, Matthew.

I’ve been using GPT-5.6 Sol internally for the past two months, I've spent probably 25+ billion tokens. Here’s my review and comparison to Fable 5: > Let's start with the analogy because everyone seems to be giving theirs - GPT-5.6 is likely the last version of the GPT-5 training run series. It's kind of like an athlete at their peak. Through years of experience in the game, they've become the most reliable player and has the highest game IQ. But, there's no more room to grow. Fable on the other hand, being essentially the first version of a new training run, is the first round draft pick rookie. Raw talent mixed with the energy only a young person would have results in some incredible plays we didn't think possible, but also mistakes due to lack of experience. But that rookie will only improve and likely will be better than the veteran ever was because it's a new game and a new era. > GPT-5.6 is genuinely better at long, sustained work. With /goal, I've had it running complex projects for days with almost no intervention. It built a Minecraft-style game, kept adding features and mobs after the core game worked, and only stopped because I stopped the run. I never felt as though I had to jump in and guide it back to the right path. > It keeps finding useful work when you give it a concrete finish line. I had it recreate Excel with a loop. It inspected the real desktop excel app with Computer Use, comparing that against its own build, and closing the gaps. I stopped it after six days after it had built an incredible amount of functionality. > It's faster than other models in two different ways. The raw generation speed is higher, something OpenAI has been putting effort into. But it also takes a shorter path to solutions. It wanders less, changes less code, and generally knows how to get things done directly. In daily use, it feels about 2-3x times faster than Fable. That's my impression, not a controlled benchmark. The difference is large enough that I notice it constantly. > It works well across a wide range of tasks. I use it for one-line edits, quick questions, browser chores, and multi-day builds without changing my prompting style. Speaking of browser control, its the best ever I've used. To the point where I actually use it often. If a task lives on a website, GPT-5.6 usually opens the browser and does it there instead of asking for an API key or forcing everything through the terminal. When I switched back to GPT-5.5, it went straight to the command line even when the browser was clearly the better tool. > And it can handle real browser work, not just toy demos. During a data import, I had it monitor Supabase and resize instances as the load changed. It stayed on the dashboard, adjusted capacity, and checked the result without an API or a custom script. > I also gave it a full Google Workspace migration. It moved Forward Future from forwardfuture.ai to forwardfuture.com, preserved the old aliases, and configured MX, SPF, and DKIM. Before a consequential save, it stopped, explained exactly what would change, and waited for confirmation. > The reasoning setting matters a lot. Light is good for questions and small edits. High and Extra High are the sweet spots for serious work. Ultra usually takes longer than the extra thinking is worth and burns tokens. > I love that 5.6 is split into 3 sizes. Not only can you control speed and cost that way, but you still also have the thinking effort setting for each of them. Very precise controls. I just wish Codex automatically routed my prompts for me. > Its personality is blunt and a little bland. Claude feels warmer and more natural to talk to. GPT-5.6 is more clinical, but I like that for work. It gives me enough explanation and rarely pads the answer. I usually have to ask Fable to explain things more simply and/or more concise. > Its front-end taste has improved, but the default is predictable. Left alone, it turns websites into PowerPoint decks with huge statements and hard section breaks. The good news is that it takes design direction well and can revise without destroying the parts that already work. > It still makes confident mistakes. I asked it to rebuild parts of a system, and it told me the job was finished. Later, I found out it wasn't. Bits of its internal process also leak into the answer occasionally. > Claude Fable is more naturally autonomous on large, open-ended projects. GPT-5.6 is easier to reach for. I don't need to invent a huge project to justify using it. It works just as well for a small edit or browser chore. > GPT-5.6 is also cheaper. Sol costs $5 per million input tokens and $30 per million output tokens. Fable costs $10 and $50. Cached input is cheaper too. Still, cost per finished task matters more than cost per token. > GPT-5.6 isn't the best at everything, and it still needs supervision. But it generates faster, wanders less, works at almost any scale, and wastes less of my time. It's the model I have the most confidence in to get the job done right the first time. I put together a full breakdown with all the tests, prompts, and examples on a site. You can read it here: signals.forwardfuture.com/gpt-5-6-review/