Expanding on point 2: In my profession, and discussing with experts from other fields, I’ve noticed that today’s top AI models work best when guided by domain specialists. For example, someone without expertise in fields like mine (nutrition) or medicine might miss subtle nuances that make an AI's response subtly incorrect. Non-experts often struggle to ask precise questions that elicit truly professional answers from AI. That’s why, at this stage, I see current AI models as ideal "coworkers" for experts rather than standalone solutions. We need human expertise to unlock their full potential and avoid critical mistakes.

I’m genuinely puzzled by how little the Fable 5 event seems to have changed the political conversation outside the US. This was a historical precedent: a US government order effectively showed that, when it wants to, America can cut foreign access to frontier AI systems overnight. And we are not talking about social media apps or cloud storage. The world is already being rebuilt around artificial intelligence: science, software, defense, education, logistics, companies, institutions, personal productivity. Yet Europe and many non-US states still behave as if access to frontier AI will remain a neutral global utility forever. Crazy. We spend billions on the rituals of late-stage comfort, on symbolic projects, on bureaucratic theater, on things that will matter very little in the world that is coming. Meanwhile, sovereign AI capability is still treated as a “strategic discussion” rather than an emergency mobilization. This is insane. If the Fable 5 incident is not enough to make governments rapidly redirect billions toward domestic AI systems, compute, talent, labs, energy infrastructure, and model independence, I honestly don’t know what would. The message could not be clearer: if your civilization depends on intelligence systems you do not control, then your civilization is not sovereign anymore.

What if GPT-5.6 is ready, but close enough to Fable-level that it’s now stuck in the same security/cyber-hardening review loop surrounding Fable/Mythos?

I had hoped this administration would take a more permissive stance on AI. This was somewhat expected, but I still hoped it wouldn’t go this way; at least, not this soon. Instead, it increasingly feels like every political camp is converging on the same endpoint: centralized control, security framing, and permissioned access to frontier systems. Different rhetoric, same destination: AI development as something to be licensed, contained, and state-mediated. As a European, this hits differently. Access to frontier AI will be the dividing line of the next era. The countries that have it will compound power across science, industry, defense, and capital. The countries that don’t will fall into dependency: slower, poorer, and forced to live downstream from systems built elsewhere.

What upsets me in the short term is that this gives OpenAI the perfect excuse to sit comfortably on 5.5 for as long as possible. In the medium term, this whole story sets a very wrong precedent, and far too early. Fable 5 is strong, but it didn’t warrant this level of hysteria.

Agree. GPT-5.5 is a good model and I rely on it a lot, BUT it feels like OpenAI chose the middle path: a strong “medium-large” model that’s easier to serve and iterate on, rather than a true giant. That makes sense strategically, but Pro users should still have access to a bigger model, even if it’s more expensive, usage-capped, or credit-based enough to stay economically neutral. I hope 5.6, or at least 5.7/5.8, gets a real size upgrade. The scaling era is not over: post-training matters, product matters, inference-time compute matters, but raw model scale still matters too.

Whenever it happens -whether with 5.6, 5.7, next week, or next month- a less-censored model meaningfully stronger than Fable 5 would send real shockwaves through society.

There’s a strange "trap" forming around personal AI infrastructure. You start by using a model. Then, slowly, it becomes the place where your projects live, your workflows are shaped, your memories accumulate, your prompts get refined, your files and habits settle. At that point, even testing a smarter model for a month comes with friction. You are no longer moving from one tool to another. You are disturbing an entire system you’ve built around yourself. That’s why a more capable model can pull you forward, while your existing AI infrastructure keeps pulling you back. One underrated move for AI labs could be making this migration almost painless: importing projects, memory, context, files, routines, maybe even agentic workflows. For challengers, that could be a massive adoption lever. For incumbents, the incentive is more complicated. Friction protects retention. But in the long run, the labs that make users feel free to move may end up earning more trust than the ones that simply make it hard to leave. And yes, this is exactly why I can’t wait for GPT-5.6 Pro.

Claude Fable 5 makes the “compute desert” obvious. We need FAR MORE datacenter capacity, especially if inference demand ramps the way it likely will over the next 6–12 months

Claude Mythos/Fable 5 looks really strong. Curious to see real-world feedback from people using it in actual workflows. For me, knowledge-work performance is the metric that matters, and it looks solid. I probably won’t switch, but it’s good to see the "public" bar move higher.

"kindle" (allegedly GPT-5.6) seems like a solid model. Just got one of its outputs in Design Arena.

Could be wrong, but I doubt we get both the super-app and GPT-5.6 together. Feels more likely they sequence them: first the ChatGPT revamp, then 5.6 the following week (or the one after).

They should say this to people hoping powerful AI might help cure their deadly disease, to people dreaming of a world with less suffering, to people who want out of wage slavery, to everyone who sees that the current system is slowly breaking. “Sorry, we need to slow down. ASI is postponed indefinitely. Please keep suffering quietly.” A slowdown can be understandable from a terminal risk perspective. Treating it as the only moral option is not. There has to be a better way than a blanket slowdown.

Something has felt different since last night in the outputs I’m getting from GPT-5.5 (Pro Extended). Maybe it just “learned” my workflow, but the style fidelity (especially on Word/docx tasks) feels much sharper and much closer to what I actually asked for. Shadow rollout/testing for GPT-5.6? @OpenAI

Most people still haven’t realized it: if you want to stay relevant in the centaur era, you need to become a skilled manager of AI agents. Fast. The ability to delegate, supervise, verify, and integrate their work is becoming a survival skill. And the time to build it is now.

The sharpness/depth of my professional work as a biologist-nutritionist have already improved noticeably through daily use of SOTA AI. I strongly agree with this perspective (especially for people who use AI as a cognitive partner rather than as a blind substitute for their own judgment) and I expect this effect to become increasingly evident with every new generation of models.

Agree with Logan here. We’re in a temporary transition phase: domain experts can still understand, evaluate, and steer the best models’ outputs, while non-experts are increasingly unable to distinguish genuinely excellent answers from subtly flawed ones. At the same time, despite occasional mistakes, I already find that models sometimes give me useful perspectives and insights I wouldn’t have thought of on my own. I don’t think this phase will last. 2–3 years max.

This is probably the best possible outcome for the film industry: good-enough video models, orchestrated by skilled professionals, will allow even low-capital but high-talent teams to deliver beautiful movies, using a savvy mix of human work, AI-augmented scenes, and full AI generation. My hunch is that within 2–3 years, films of this caliber will be possible even for low-budget, little-known production studios; and they’ll be created in weeks rather than months, or even a year.

If this model is already an internal daily driver, it’s probably closer than expected. Researchers don’t usually daily-drive fragile prototypes. My guess is that models used this way are at most one generation ahead of what’s public. TL;DR: could be 5.6 🔥