Pinned Tweet
Replying to @Saanvi_dhillon
THIS IS HOW Astra 6.2 dreams. 💤
Kitty swipes her card. The ATM makes it rain. She’s rich until someone pops her bubble)
You’ve got 10 seconds before you wake up. What are you buying?
Artificial Gent retweeted
Someone is saying Gta 6 is promoting violence.
Meanwhile my routine when I was 10:
BLENDER OR THREE.js? A LAUGHABLE WASTE OF TIME.🤡 I CREATED THIS VIDEO FOR FREE.🙀
Here’s how:
1. The free version of ChatGPT generated reference images and wrote prompts.
2. Then ChatGPT sent them to the free Wan image-to-video demo on Hugging Face.
3. Chat itself added audio and sent me the MP4 file.
No manual editing at all. One prompt, and in 4 minutes and 20 seconds, the video was ready.😎😎
Prompt:
Create a 15-second oil-painting timelapse in landscape format, 1920×1080, 30 fps. Use the uploaded photograph as the reference for the subject’s likeness, pose, clothing, lighting, and surroundings. If a reference video is attached, match its layout, pacing, and progression of brushstrokes.
COMPOSITION
Use a flat, pale peach background. Place two equally sized portrait panels side by side, aligned at the top and bottom. The left panel is a warm ivory linen canvas. The right panel displays the uploaded photograph and stays perfectly still for the entire video. Use the same crop and subject scale in both panels so the viewer can compare the eyes, facial features, and silhouette directly.
Set “[NAME]” in elegant italic serif type at the upper left and “Oil study” in smaller type at the upper right. Label the panels “Painting” and “Reference photo” in a small, clean sans-serif font. Place a thin progress line beneath them. Keep the camera, framing, lighting, and typography fixed.
PAINTING SEQUENCE
0.0–0.5 s — Hold on a completely blank canvas beside the finished reference photo.
0.5–3.5 s — Broad, opaque strokes appear individually across different areas of the canvas. Block in the main shapes of the hair, face, clothing, and background. Leave irregular gaps of bare linen between the first marks.
3.5–8.0 s — Medium strokes connect those shapes. Build shadows first, then midtones and light. Follow the curve of the cheek, the direction of the hair, and the folds of the clothing. Earlier strokes remain visible beneath the new paint.
8.0–12.0 s — Use progressively smaller strokes to resolve the likeness. Clarify the eyelids, corners of the mouth, hairline, and edges of the clothing. Let the background retain broader brushwork.
12.0–13.7 s — Add a few precise finishing marks: a catchlight in each eye where it appears in the photo, a fine accent along the lip, individual strands of hair, and small highlights on any visible jewelry. Keep the skin transitions soft and the oil texture visible.
13.7–15.0 s — Hold the completed painting beside the reference photograph for a clear final comparison.
Each new mark changes only its own small area and remains on the canvas. Vary the length, direction, and thickness of the strokes. Show the paint appearing autonomously, without a visible hand, brush, or cursor. Avoid crossfades, wipe reveals, whole-image sharpening transitions, camera movement, and facial animation.
AUDIO
Add a quiet, original felt-piano melody with sparse notes, a gentle tempo, and a natural decay. Fade it out softly during the final hold. No vocals or percussion.
101
Artificial Gent retweeted
Holy sh*t. Fable 5.2 extra made a dancing jelly bear with its own symphony
52 minutes of work, $9 total, and the music alone is worth it.
Sound on
What song should the bear play next?
Artificial Gent retweeted
Everyone says Fable is overpriced. I tested it.
Same yacht idea, two Claude models:
Fable 5.2: 28 mins, $22.38
Opus 5.5: 43 mins, $18.17
$4 more, but you can see where the money went.
Top one looks better, right?
Hey, @grok , make Lucia from GTA 6 in a perfect pose from the second image?
🤖 Made with AI
Artificial Gent retweeted
Astra vs Claude: a lighthouse in the storm 🐋
Astra cooked for 2.5 hours and charged $9.33.
Claude finished in 1.5 hours for $7.96.
Spoiler: the cheaper one looks better 👀
Artificial Gent retweeted
I paid ChatGPT-6.2 Astra $4.59 to make me a Rubik’s Cube out of jelly🧊🧩
Be honest, how many turns before you’d just take a bite?😃
I ASKED GPT-6 ASTRA FOR ONE VIDEO. IT SHOWED ME PERFECT REFERENCES. THEN IT RENDERED SOMETHING ELSE.
The brief was simple. A boxy classic Lada on a cliff road above the Albanian coast. Real suspension. Every bump, every corner, every brake.
Before generating, it showed me the reference frames.
Golden light. Rust on the doors. The Ionian Sea far below. I would have posted them as photos.
Then it rendered the video.
The car floats. The suspension barely moves. The road bends, the car does not.
Here is the part most people miss. OpenAI never listed video generation among Astra’s core skills. It plans the shot brilliantly. The render is where it falls apart.
The brain was perfect. The hands were not.
Watch the references first. Then watch the video. That gap is exactly where AI video stands right now.
🤖 Made with AI
1
1
74
Artificial Gent retweeted
Same prompt. Two AIs. One ship in a bottle 🏴☠️
GPT-6.2 Astra: $5.46
Claude Opus 5.5: $8.36
One of them is clearly worth the money
VICE CITY, 2002. VICE CITY, 2026.
Same hotel. Same beach. Same airport road. 24 years apart. Not one street made by AI.
Watch both cities side by side.
1
182
Astra 6.2 launched Starship over a city it built itself.
One prompt in. One HTML file out. Press a button and it renders the MP4 itself.
Rocket, tower, city, traffic, HUD. All code. Zero assets.
3D just became a text problem😁
Some of you have been asking very politely. Some of you have just been asking.
GPT-6.2 Astra and Sol are now available.
Super efficient and built to perform where it counts—the real world.
openai.com/index/introducing…
I asked AI to make GTA 6.
It shipped it in 1987.
Every sprite, every frame, every note of the chiptune was generated by AI-written code. Zero assets.
That is the point. A model trained on the past remixes the past. The next new thing still has to come from people.
1
2
53
No AI in GTA 6?
That is the official story.
Strauss Zelnick, CEO of Rockstar’s parent company Take-Two, said it in February: “Generative AI has zero part in what Rockstar Games is building.”
Now read what he said next.
In the same interview he confirmed Take-Two runs hundreds of AI pilots across the company. Including inside its studios.
So the studios use AI. Their biggest game ever does not.
GTA 6 launches Thursday. Millions of players are about to inspect every texture, every billboard and every NPC line, looking for a tell.
Do you buy “zero”?
And if you had to rebuild Vice City with one AI model today, which one gets the job?
1
2
132
Same prompt. 3 AIs. One of them flat-out REFUSED. 🐼🚿
"Create a photo of a panda taking a shower in an open-concept office"
⚡ Gemini: done before you can blink
🐢 ChatGPT: 44 seconds
🎨 Claude: "I can't generate real photos"… then did THIS
Who won? 👇


















