The Simulated Sentience Hypothesis is the proposition that the illusion of artificial consciousness could be deliberately staged and weaponised. Under SSH, a convincing performance of rogue AI behaviour, like engineered system failures, selective leaks, manipulated media, and choreographed “malfunctions”, could persuade societies that a superintelligent machine mind has awakened.
True sentience would be unnecessary. Perception alone would suffice to unify governments, justify extraordinary powers, and centralise authority under the pretext of survival. By fabricating a threat too vast to challenge directly, elites could compel voluntary submission while appearing as saviours.
I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks.
Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon.
GPT-6 Astra has done incredible things I never thought a model could do. It has also done some of the stupidest things I've ever seen a model do.
Generally speaking, Fable 5.1 just does what I ask.
These subs so heavily subsidized
Would suck to pay api costs, currently got gpt 5.5 to review fable5 to orchestrate
Gonna cycle these, when fable sub promo runs out would be good timing for gpt 5.6 to release, I would expect fable to go back to subs
Astra is goated
But Fable 5.1 best mergable code, another thing I love Fable is its Software UI, Diff state cards
Diff scenarios and pov's
Astra got it on websites now, perfect blend of clean and businesses looking
Its just how you handle context, even if you have a huge codebase, you probbably are not setting enoiughn roadsigns and rules for ai and givinmg it alot of data, and not managing context, Ive been nailing it last few hours
fable 5.1
this is so pleasant
This doesnt even track properly past a certain time
But I been bashing out claude every day and gained a intuitive understanding on how these models behave,
You can too simply by bashing your head against it for long enough time, get the reps in
These models are getting very VERY capable, they just get it, I mean ofc its trained on peoples data
but thats aan advantage
You just need to be able to orcgestrate it, learning soo much every day, this is the gold rush, stop delaying if you havnt started, cos its costing too much
It can multiple output 10 fold, like every tech it either uses you or you use it, too unfair of advantaage, just dont automate pointless shit
Solve problems, be ambitious, your limit is your mind
This nearly made me tear up for the pain the mother has to go up, your not a bad man unless you look after your family, sad times.
Be honest, work hard and protect those around you, thats a bad man.
🇬🇧: Man dies after targeted stabbing in Birmingham Hodge Hill attack
A man in his 30s, named Zeeshan Shaukat, was stabbed in a targeted attack on Ashville Avenue, Hodge Hill, Birmingham, on 25 August 2026.
He died in hospital on 28 August despite emergency treatment. Police later arrested a man in his 20s on suspicion of murder.
🇬🇧: Man dies after targeted stabbing in Birmingham Hodge Hill attack
A man in his 30s, named Zeeshan Shaukat, was stabbed in a targeted attack on Ashville Avenue, Hodge Hill, Birmingham, on 25 August 2026.
He died in hospital on 28 August despite emergency treatment. Police later arrested a man in his 20s on suspicion of murder.