i don't think most people even comprehend how far AI has come in just the last year
we went from 6 fingers, weird artifacts, and robotic voices to being able to become a full time youtuber without ever touching a camera
full prompt:
CAMERA: Handheld DV 16mm daily vlog footage. The video MUST begin with her holding the camera at arm's length in selfie mode, speaking directly to the lens while casually walking through the gym. The first 20–30 seconds are entirely handheld. Only later does she occasionally place the camera on a dumbbell rack, adjustable bench, stretching mat, gym bag, or water bottle for wider shots. Keep subtle handheld shake, drifting composition, autofocus hunting, rushed reframing, uneven zooms, exposure breathing, brief accidental face cropping, and imperfect framing throughout. The camera itself is never visible.
LOOK: Warm analog tape texture with gentle film grain, slightly softened sharpness, subtle halation around overhead lights, realistic skin tones, low contrast, tiny exposure shifts, and natural motion blur. The footage should feel authentic and completely unstaged, like a real creator documenting her workout.
STYLE: An intimate daily fitness vlog filmed in a mostly empty boutique gym. Casual, relaxed, and conversational. Small laughs, pauses, fixing her ponytail, wiping sweat with her towel, catching her breath, looking away while thinking, and talking naturally to the audience instead of delivering lines. **She speaks in short, natural sentences with frequent pauses. Never rush dialogue. Leave quiet moments between lines.**
CHARACTER: A beautiful white blonde fitness creator in her early 20s. Long blonde hair tied into a loose ponytail, blue eyes, glowing skin, and a slim athletic figure. She wears a fitted black long-sleeve workout top, high-waisted black leggings, white socks, white sneakers, and a white towel loosely resting around her neck.
SETTING: A modern boutique gym late in the evening. Dumbbell racks, kettlebells, medicine balls, mirrors, benches, stretching mats, resistance bands, a gym bag, shaker bottle, and warm overhead lighting. The gym is almost empty with only one or two people blurred in the distant background.
SCENES:
The vlog opens in selfie mode. She is holding the camera at arm's length while slowly walking through the dumbbell area, smiling naturally.
"...okay."
"I definitely earned dinner."
She laughs quietly and adjusts the towel around her neck.
"I almost skipped today."
She briefly turns the camera toward the nearly empty gym before pointing it back at herself.
"Glad I came."
Still holding the camera, she walks toward the stretching area.
"My legs are done."
She places the camera on a nearby bench for a wider angle and sits down on a stretching mat.
She stretches one leg and takes a deep breath.
"I really need to stretch more."
She smiles to herself.
She switches sides and rolls one shoulder.
"Tomorrow's gonna hurt."
She picks the camera back up and walks toward her gym bag.
She kneels beside it, pulls out a protein shake, and takes a sip.
"...that's actually good."
She laughs.
The camera sits on top of the gym bag while she packs away her towel, headphones, and lifting straps.
"Worth it."
She throws the bag over one shoulder and picks the camera back up.
Walking toward the exit in selfie mode, she smiles into the lens.
"See you next workout."
She gives the camera a small wave before ending the recording.
We're starting to leave the territory where you'd test an LLM by e.g. "create an svg of pelican on a bicycle". As one idea to generalize it, I was interested what Opus 5 would do if I gave it the first paragraph of the Lord of the Rings, a 1M token budget (~$10) and asked for three js render of it. Opus went off for ~2 hours and wrote 5500 lines of code that (procedurally) rendered the story. It's kind of janky but fun. But it's a bit mindboggling that the LLM has to place and orchestrate various polygon assets in (x,y,z) coordinates and write code that animates it all, and that it even does anything at all.
I also like this kind of examples because no one in their right mind would ever spend the time to write something this custom but LLMs have all the stamina and patience in the world, so it's an example where we go from "no one would ever do this" to "sure, why not, it's ~free". There might be a lot more. But I'm excited about creating hyper custom worlds that you can imagine dropping players into, e.g. here to participate in the LoTR story as a spectator NPC, or one of the characters, or etc. Something like an ephemeral GTA of X on demand.
Last thought is that the domain of worlds/games exposes a weakness in LLMs: they can't easily audit their work because they aren't able to efficiently and natively perceive videos or play games within them. Here, Opus 5 had to very slowly and painstakingly take screenshots at different points, and it messed up a few times and created a bunch of jank. An example of raw capability (multimodal, gameplay) that I think is still quite lacking.
Oh my god Opus 5 is ridiculously good at mobile apps. It might be better than Fable? wtf?
I tried my old "conversational calorie tracker" prompts. DAMN this is clean for a v1: