We just announced our Fusion API:
- Fable-level performance on deep research tasks, at half the cost
- Better-than-SOTA performance using panels
The future of AI is neurodiversity, not single-model takeovers.
Paul Verhoeven on whether the events that take place in "Total Recall" (1990) are real or just a dream:
"It is both. To be honest, that’s what I want. I made the movie in a way that it would be true on both levels, and I spent a lot of time to get that. If you want a scientific explanation, you know, of course, in quantum mechanics there is a very interesting principle, the principle of uncertainty, Heisenberg’s principle.
If you have a big object and if you try to measure the place of the object and the velocity of the object at the same time, the more precisely you measure velocity the less precise place gets. So that’s the principle. That means, of course, that there are different realities possible at the same moment. What I wanted to do in 'Total Recall' is to do a movie where both levels are true. I mean for me, of course, the film anyhow has to do with two realities, one being the reality of going as a secret agent to Mars and discovering that there is a problem, and solving the problem, which is starting the nuclear reactor and helping the guerrillas and destroying Cohaagen.
The second level of the movie, of course, is that from the moment that he goes into the Rekall chair ‘til the end it’s a dream, and I tried to make that second level work throughout the whole movie.
So there’s the dream level which starts when he gets into the chair and the thing is in his neck, and that would go throughout the whole movie, so in the next scene where they say, Oh, there’s a problem, there’s a big glitch here, that would be already the dream, of course. That’s where the dream starts. And the next scene where they are fighting and stuff would be part of his dream, convincing him that it is real, because there is a glitch but that would be part of the program. It would be built into the program to make him accept the fact that it’s real, but it’s a dream.
If you look at the movie, if you haven’t seen it, or for the second time, you’ll see that the whole program that’s set up at the beginning when he goes to the Rekall office and he talks to this guy who sells him the program on Mars, you’ll see that he gets everything that he wants: he gets the trip to Mars, he gets the girl, the exotic girl, he ki!!s the bad guys, and he saves the entire planet. That’s what he does. And that’s basically the dream.
Even halfway through the movie, you may remember, this other guy comes in, Dr. Edgemar, and tells him that he’s in a dream, that he’s still in the Rekall chair, and then Arnold says, “If I’m there, I can ki!! you.” And he puts a gun to his head and the guy says, “Sure, no problem for me, big problem for you, because you will be psychotic from now on because the walls of reality will fall apart. One moment you will be the savior of the rebel cause, the next moment you’ll be Cohaagen’s bosom buddy, but in the end—you will even have these strange fantasies about alien civilizations—but at the end you will be lobotomized.”
And then if you see the movie, you realize that all these things happen. I mean he is lobotomized at the end. That’s why at the last shot, when they are so happy and kissing each other, it slowly fades to white, which for me meant, “OK, there he goes. That’s the end-that’s the dream—they lobotomized him.” And all the other things happened, he finds the alien civilization, he rescues the planet, he finds the good girl, he k!!ls the bad guys, but it’s a dream. Now, of course you can see it as a reality, too. So at the end of the movie, getting to white means either it’s a happy ending or he loses his brains . . . which is probably also a happy ending, I don’t know.
That was basically what l wanted—that at the end there would be two possibilities, and they would be both true—for me they are both true—it’s not either one or the other. It’s not that either it’s a dream or it is a reality. It is a dream and it is a reality. And I think they’re both there."
(Paul Verhoeven's interview with Chris Shea & Wade Jennings, 1992)
P.S: On this day, 36 years ago, "Total Recall" (1990) premiered in Los Angeles, California, USA.
In 2000, Roger Ebert and Martin Scorsese got together to present their Top 10 Best Films of the 1990s, and it's really interesting to see what they loved, or didn't love, out of the entire decade 📽️
The Best 'Exposition Dump' in Film History. In 5 minutes, Spielberg lays out the characters, mythology, locations, artifacts and the entire narrative engine of the film. The audience discovers everything alongside Indy + each detail pays off later. A masterclass in storytelling.
In 1962 the BBC interviewed PETER O’TOOLE in Almería, Spain — where he was preparing to play the title role in a certain movie called
LAWRENCE OF ARABIA
Announcing Talkie: a new, open-weight historical LLM! We trained and finetuned a 13B model on a newly-curated dataset of only pre-1930 data. Try it below!
with @AlecRad and @status_effects 🧵
"how do you fit qwen 3.6 27b q4 on 24gb at 262k context" lands in my dms 5 times a week. here is the exact memory math.
model bytes at idle = 16gb (q4_k_m of 27b dense)
kv cache at 262k context with q4_0 for both k and v = 5gb
total = 21gb on the card
headroom = 3gb for prompts and tool call traces
the magic is the kv cache type. most people leave it at default fp16 or push to q8 thinking quality wins. on qwen 3.6 27b dense at 262k:
- fp16 kv cache = does not fit at all
- q8 kv cache = fits at 23gb but runs 3x slower (double penalty: more vram, less speed)
- q4_0 kv cache = fits at 21gb at full speed (40 tok/s flat curve, same speed at 4k or 262k)
most builders never test the kv cache type because tutorials never mention it. it is the single biggest unlock on consumer 24gb hardware.
flags i run:
./llama-server -m Qwen3.6-27B-Q4_K_M.gguf -ngl 99 -c 262144 -np 1 -fa on --cache-type-k q4_0 --cache-type-v q4_0
what they do:
-ngl 99 = offload everything to gpu
-c 262144 = 262k context window
-np 1 = single user slot (do not enable multi-slot, eats headroom)
-fa on = flash attention on (memory and speed both win)
--cache-type-k q4_0 --cache-type-v q4_0 = the unlock
if you are sitting on 24gb and not running this config, you are leaving 250k of context on the table. or worse, you are running q8 kv cache and burning 3x your speed for nothing.
q4 is not a compromise on consumer hardware. it is the right call.
LAWRENCE OF ARABIA (1962)
The best day-for-night shot in cinema history. Cinematographer Freddie Young underexposed the film and used military aerial reconnaissance filters + ND filters to help craft this shot.
@levelsio I see the responses you're getting - most are not seeing it, meaning there will be a rude awakening when the "SaaSpocalypse" eventually comes. What are we left with in the short term, financial markets?