正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Screen Studio mouse hover effect desync finally solved! It checks which mouse movements cause hover effects (e.g., changing the label background). Then, it dynamically adjusts mouse movement smoothness to make sure the mouse position matches the footage during those moments.

Adam Pietrasiak

39,237 subscribers

10,568 次观看 • 19 天前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

If you play on mouse, make sure you check your polling rates, I'll link the website below: You click the screen and then spin your mouse in a circle to get an accurate polling rate of your mouse.

If you play on mouse, make sure you check your polling rates, I'll link the website below: You click the screen and then spin your mouse in a circle to get an accurate polling rate of your mouse.

KIRNEILL

34,206 次观看 • 3 年前

Take a moment to remember the good things in life; It’s a hard thing to do for sure, when it feels like everything in the world is tipping this way and that. Maybe this little mouse can help you connect with “la vie en rose” & if so, then my job as an artist is done for today 🩷

Take a moment to remember the good things in life; It’s a hard thing to do for sure, when it feels like everything in the world is tipping this way and that. Maybe this little mouse can help you connect with “la vie en rose” & if so, then my job as an artist is done for today 🩷

Liza Adamczewski or the accidental ecologist

19,548 次观看 • 8 个月前

004/100 Buttons. A bit of the process on building an animation. When looking at a finished animation or in this example a finished button, it can look quite complex inside the CSS. But when building it, it’s more like a lot of simple steps, one after another. Here I had the idea to make some kind of text animation like the footer logo on the Osmo site. I try to add the base animation with no complex easing, for example transition: translate 0.4s ease. Starting with just moving the one text from bottom to top and the other text to top. Adding a stagger, play around with it. Searching for a way to make it more circular. On the research I found the sin() function inside CSS which can build a more smooth non linear curve for the stagger which creates this circular effect. And step by step adding more complexity like, different easing for hover/hover-out, opacity, 3D transform and more. I use also the sin() function to rotate the letters, so the middle ones are getting more rotated than the outer ones. Another thing which helps is to add a small delay on hover, for example 0.05s or 0.1s, you don’t really see the difference, but when you hover pretty fast on and out it doesn’t get that jumpy. I’m using here GSAP’s SplitText to split every char into spans. And then I’m adding a CSS index variable to every span, starting from the center. SplitText can provide CSS index variables, but you cannot tell it from which direction. For the sin() it’s also important to have a max length, so I add another CSS variable with the max char number on it. Crafting 100 Buttons with Osmo ⏳ Total time: 63h

Eduard Bodak

166,023 次观看 • 5 个月前

Holy shit...Claude Code just killed one of the most annoying terminal problems — flickering. 🤯 Anthropic quietly shipped NO_FLICKER mode and it changes how AI feels inside the terminal: • No more screen flashing • Smooth scrolling + stable viewport • Mouse support (yes… in terminal) • Constant memory/CPU as chats grow • Cleaner selections (no UI junk copied) This isn’t just a UI tweak. They basically virtualized the entire terminal viewport to control rendering themselves. Result? Terminal AI finally feels like a real app — fast, smooth, usable. ⚡ Turn it on: CLAUDE_CODE_NO_FLICKER=1 claude Small launch. Huge UX upgrade. Terminal-native AI is getting scary good.

Holy shit...Claude Code just killed one of the most annoying terminal problems — flickering. 🤯 Anthropic quietly shipped NO_FLICKER mode and it changes how AI feels inside the terminal: • No more screen flashing • Smooth scrolling + stable viewport • Mouse support (yes… in terminal) • Constant memory/CPU as chats grow • Cleaner selections (no UI junk copied) This isn’t just a UI tweak. They basically virtualized the entire terminal viewport to control rendering themselves. Result? Terminal AI finally feels like a real app — fast, smooth, usable. ⚡ Turn it on: CLAUDE_CODE_NO_FLICKER=1 claude Small launch. Huge UX upgrade. Terminal-native AI is getting scary good.

Suryansh Tiwari

32,145 次观看 • 6 个月前

POV: You thought you were getting ice cream… but the Turkish ice cream guy had other plans 😂🍦 Created with Seedance 2.5 on BudgetPixel AI PROMPT : CHARACTER: A young woman wearing a cute, stylish amusement-park outfit with a clearly visible Mickey Mouse ears headband. Keep the headband consistent throughout the entire video. She is excited, playful, and expressive. A cheerful Turkish ice cream vendor stands behind a traditional Turkish ice cream stall, holding a long metal ice cream paddle. 0–3 SEC — ORDER Start with a cinematic wide shot of the colorful amusement park, then smoothly move toward the Turkish ice cream stall. The girl approaches the counter, smiling excitedly. Her Mickey Mouse ears headband is clearly visible. She points toward the ice cream. The vendor prepares a cone with a scoop of ice cream and extends it toward her. 3–6 SEC — THE FIRST TEASE Cut to a medium shot. The girl eagerly reaches for the cone. Just as she is about to grab it, the vendor smoothly pulls the ice cream away with the long paddle. She looks at him in surprise. He smiles mischievously and offers it again. 6–9 SEC — MORE TEASING She reaches again. The vendor smoothly moves the cone from left to right, then slightly upward, always just out of reach. The girl follows the cone, laughing and becoming increasingly determined. Keep the interaction fluid, realistic, and playful, with natural body movement and timing. 9–11 SEC — THE NOSE TRICK The vendor finally brings the ice cream toward her. She confidently leans forward, expecting to receive it. Instead, he gently places a small dab of ice cream on the tip of her nose. Close-up on her shocked expression. She pauses, looks cross-eyed toward the ice cream on her nose, then looks back at the vendor with a playful annoyed expression. 11–15 SEC — SHE WINS Before the vendor can tease her again, she quickly grabs the cone from the paddle. She successfully gets the ice cream. The vendor laughs and raises his hands in playful defeat. She proudly holds up the cone, smiles toward the camera, wipes the ice cream from her nose, and takes a happy bite. CINEMATOGRAPHY: Photorealistic live-action, premium cinematic quality, warm golden-hour lighting, natural skin texture, realistic facial expressions, shallow depth of field, soft background bokeh, subtle anamorphic lens characteristics, natural motion blur, smooth gimbal movement, cinematic push-ins and tracking shots, realistic amusement-park atmosphere. IMPORTANT: Keep the Mickey Mouse ears headband clearly visible and consistent throughout the scene. Maintain consistent faces, wardrobe, hairstyle, ice cream cone, vendor paddle, and environment. All movements must be smooth and physically realistic. No exaggerated cartoon movement, no 3D animation, no jerky motion, no teleporting, no object duplication, and no unnatural hand movements.

POV: You thought you were getting ice cream… but the Turkish ice cream guy had other plans 😂🍦 Created with Seedance 2.5 on BudgetPixel AI PROMPT : CHARACTER: A young woman wearing a cute, stylish amusement-park outfit with a clearly visible Mickey Mouse ears headband. Keep the headband consistent throughout the entire video. She is excited, playful, and expressive. A cheerful Turkish ice cream vendor stands behind a traditional Turkish ice cream stall, holding a long metal ice cream paddle. 0–3 SEC — ORDER Start with a cinematic wide shot of the colorful amusement park, then smoothly move toward the Turkish ice cream stall. The girl approaches the counter, smiling excitedly. Her Mickey Mouse ears headband is clearly visible. She points toward the ice cream. The vendor prepares a cone with a scoop of ice cream and extends it toward her. 3–6 SEC — THE FIRST TEASE Cut to a medium shot. The girl eagerly reaches for the cone. Just as she is about to grab it, the vendor smoothly pulls the ice cream away with the long paddle. She looks at him in surprise. He smiles mischievously and offers it again. 6–9 SEC — MORE TEASING She reaches again. The vendor smoothly moves the cone from left to right, then slightly upward, always just out of reach. The girl follows the cone, laughing and becoming increasingly determined. Keep the interaction fluid, realistic, and playful, with natural body movement and timing. 9–11 SEC — THE NOSE TRICK The vendor finally brings the ice cream toward her. She confidently leans forward, expecting to receive it. Instead, he gently places a small dab of ice cream on the tip of her nose. Close-up on her shocked expression. She pauses, looks cross-eyed toward the ice cream on her nose, then looks back at the vendor with a playful annoyed expression. 11–15 SEC — SHE WINS Before the vendor can tease her again, she quickly grabs the cone from the paddle. She successfully gets the ice cream. The vendor laughs and raises his hands in playful defeat. She proudly holds up the cone, smiles toward the camera, wipes the ice cream from her nose, and takes a happy bite. CINEMATOGRAPHY: Photorealistic live-action, premium cinematic quality, warm golden-hour lighting, natural skin texture, realistic facial expressions, shallow depth of field, soft background bokeh, subtle anamorphic lens characteristics, natural motion blur, smooth gimbal movement, cinematic push-ins and tracking shots, realistic amusement-park atmosphere. IMPORTANT: Keep the Mickey Mouse ears headband clearly visible and consistent throughout the scene. Maintain consistent faces, wardrobe, hairstyle, ice cream cone, vendor paddle, and environment. All movements must be smooth and physically realistic. No exaggerated cartoon movement, no 3D animation, no jerky motion, no teleporting, no object duplication, and no unnatural hand movements.

Sophia

19,583 次观看 • 1 个月前

Wigmund’s Biggest Sale Ever is ending soon - 87% off! In a world where big corporations push games toward the same formulas, we chose a different path. We turned down investors who wanted us to make our combat mainstream. Instead, we protected our vision and kept the unique mouse-gesture sword combat 100% authentic. The result is a vibrant open-world RPG that feels genuinely different. If you’re looking for something refreshing and original, now is the best time to jump in. Thank you again for all the incredible support during these difficult weeks. It means everything to us. We’re excited to show you what’s next soon ❤️ #Wigmund #IndieGame #IndieDev #RPG #MouseAsSword

Wigmund’s Biggest Sale Ever is ending soon - 87% off! In a world where big corporations push games toward the same formulas, we chose a different path. We turned down investors who wanted us to make our combat mainstream. Instead, we protected our vision and kept the unique mouse-gesture sword combat 100% authentic. The result is a vibrant open-world RPG that feels genuinely different. If you’re looking for something refreshing and original, now is the best time to jump in. Thank you again for all the incredible support during these difficult weeks. It means everything to us. We’re excited to show you what’s next soon ❤️ #Wigmund #IndieGame #IndieDev #RPG #MouseAsSword

Wigmund / On Steam

27,229 次观看 • 4 个月前

i helped satisfy one of my coaching client's wives last month there's this one windows setting that's secretly holding back thousands of league players it's called enhance pointer precision sounds good right? sounds like it makes your cursor more accurate it does the opposite it adds mouse acceleration. your cursor moves different distances based on how fast you move your hand move slow: cursor barely moves move fast: cursor flies across the screen so your muscle memory is fucked. your brain can never build consistent patterns. you're trying to learn micro but your cursor is literally different every single time you move it it's like trying to learn piano but every key is in a different spot each day this client was silver 4. adc main. mechanics looked drunk. cursor flying everywhere. overshooting targets. kiting looked like a seizure told him to turn it off instantly his cursor is 1 to 1. hand moves an inch, cursor moves an inch. every time. consistent. predictable. smooth his hands and fingers finally learned what consistency feels like two weeks later he's gold 3. mechanics cleaner. skillshots landing. kiting smoother then i get a message "bro my wife told me to thank you" i said "for helping you climb?" he said "nah bro. the mouse precision thing. helped with more than league" took me a second. i don't just coach league i help wives finish Study the Saskio way

i helped satisfy one of my coaching client's wives last month there's this one windows setting that's secretly holding back thousands of league players it's called enhance pointer precision sounds good right? sounds like it makes your cursor more accurate it does the opposite it adds mouse acceleration. your cursor moves different distances based on how fast you move your hand move slow: cursor barely moves move fast: cursor flies across the screen so your muscle memory is fucked. your brain can never build consistent patterns. you're trying to learn micro but your cursor is literally different every single time you move it it's like trying to learn piano but every key is in a different spot each day this client was silver 4. adc main. mechanics looked drunk. cursor flying everywhere. overshooting targets. kiting looked like a seizure told him to turn it off instantly his cursor is 1 to 1. hand moves an inch, cursor moves an inch. every time. consistent. predictable. smooth his hands and fingers finally learned what consistency feels like two weeks later he's gold 3. mechanics cleaner. skillshots landing. kiting smoother then i get a message "bro my wife told me to thank you" i said "for helping you climb?" he said "nah bro. the mouse precision thing. helped with more than league" took me a second. i don't just coach league i help wives finish Study the Saskio way

Tony Chau

23,052 次观看 • 10 个月前

SAMPURNA 70 MM, Vanasthalipuram, Hyderabad – Recent Experience ───────────────────────── Yesterday, I watched #UstaadBhagatSingh just to visit this theatre. I’ve covered almost all the theatres around Dilsukhnagar, LB Nagar, and Karmanghat. I’ve been hearing about this theatre since 2011, but I had never watched a movie here or even seen it from the outside. Finally, I visited it yesterday. Overall, it was an okayish experience. The sound system is good, especially the bass. However, the screen brightness is quite low, and the visuals lack clarity. The theatre gives off a nostalgic old-school vibe, which I liked. One interesting highlight is the graffiti below the screen featuring Telugu hero tag names, enhanced with lighting effects. They turn on during mass song moments, which adds a fun touch. At present, it’s not the best place for a proper movie-watching experience. However, renovation work is underway, so it’ll be interesting to see how it turns out after the upgrade. #Sampurna70MM

SAMPURNA 70 MM, Vanasthalipuram, Hyderabad – Recent Experience ───────────────────────── Yesterday, I watched #UstaadBhagatSingh just to visit this theatre. I’ve covered almost all the theatres around Dilsukhnagar, LB Nagar, and Karmanghat. I’ve been hearing about this theatre since 2011, but I had never watched a movie here or even seen it from the outside. Finally, I visited it yesterday. Overall, it was an okayish experience. The sound system is good, especially the bass. However, the screen brightness is quite low, and the visuals lack clarity. The theatre gives off a nostalgic old-school vibe, which I liked. One interesting highlight is the graffiti below the screen featuring Telugu hero tag names, enhanced with lighting effects. They turn on during mass song moments, which adds a fun touch. At present, it’s not the best place for a proper movie-watching experience. However, renovation work is underway, so it’ll be interesting to see how it turns out after the upgrade. #Sampurna70MM

CineManiac

10,066 次观看 • 7 个月前

Cat Playing With Its Mouse: The Meaning Behind China’s Radar Lock on Japan’s F-15J on December 6th There is real tension in the air between Japan and China. Washington is slowly retreating from the Western Pacific, abandoning the region to Japan and telling it to hold the line alone. The world has shifted. There was a time when China was weak and poor, when every bully felt entitled to trample it. Those days are over. China rose for one purpose: to never again be abused and bullied. And now, strong and self-assured, China can finally settle accounts with the old tormentors — and Japan is at the top of that list. But the age of war has changed. The modern battlefield is not measured by bayonets or trenches, but by fire-control radars, AI-driven targeting systems, and the kind of technological superiority that lets you toy with your opponent without firing a shot. And this is exactly what unfolded on December 6. The Fire-Control Radar Incident: A Gun to the Forehead Koizumi Shinzorō, Japan’s Prime Minister, wanted confrontation, and China responded with precision. On December 6, Japan’s Defense Minister publicly admitted that J-15s launched from the Liaoning carrier had twice locked Japan’s F-15Js with fire-control radar over the high seas southeast of Okinawa. He called it “dangerous” and expressed “deep regret.” He is right to feel danger. This was no accident, no miscalculation. This was the PLA presenting a complete offensive posture, signaling that the region is one fingertip away from war — and China is fully prepared to exercise their rights under UN Charter Article 107. Most people have no idea how deadly serious “fire-control illumination” is. Japan’s F-15J pilot certainly knew: his cockpit must have exploded into warning alarms; Imagine that shrill beeping screaming through the cockpit for half an hour - enough to drive any normal person mad; his breathing must have turned shallow; his hands probably shook as he tried to maneuver away from the lock. However, even under that level of crushing psychological pressure, the Japanese pilot chose to stay inside the zone rather than fleeing. This is kamicaze level provocation. Because the moment fire-control goes live, the radar narrows into a focused beam, feeding exact parameters to the missile under the jet’s wing. In peacetime exercises, a sustained lock counts as a confirmed kill. In real combat, nobody activates it unless they are ready to shoot. And China kept that beam on the F-15J for over half an hour. First lock: 16:32–16:35, three full minutes. Second lock: 18:37–19:08, more than thirty minutes. There is no suspense about how this confrontation would end in a real war. It would be a guaranteed kill. The Cat and the Mouse To be precise, the J-15 wasn’t merely locking and unlocking. It was playing — the way a cat toys with a mouse trapped under its paw. Japan’s F-15J was that mouse.

Cat Playing With Its Mouse: The Meaning Behind China’s Radar Lock on Japan’s F-15J on December 6th There is real tension in the air between Japan and China. Washington is slowly retreating from the Western Pacific, abandoning the region to Japan and telling it to hold the line alone. The world has shifted. There was a time when China was weak and poor, when every bully felt entitled to trample it. Those days are over. China rose for one purpose: to never again be abused and bullied. And now, strong and self-assured, China can finally settle accounts with the old tormentors — and Japan is at the top of that list. But the age of war has changed. The modern battlefield is not measured by bayonets or trenches, but by fire-control radars, AI-driven targeting systems, and the kind of technological superiority that lets you toy with your opponent without firing a shot. And this is exactly what unfolded on December 6. The Fire-Control Radar Incident: A Gun to the Forehead Koizumi Shinzorō, Japan’s Prime Minister, wanted confrontation, and China responded with precision. On December 6, Japan’s Defense Minister publicly admitted that J-15s launched from the Liaoning carrier had twice locked Japan’s F-15Js with fire-control radar over the high seas southeast of Okinawa. He called it “dangerous” and expressed “deep regret.” He is right to feel danger. This was no accident, no miscalculation. This was the PLA presenting a complete offensive posture, signaling that the region is one fingertip away from war — and China is fully prepared to exercise their rights under UN Charter Article 107. Most people have no idea how deadly serious “fire-control illumination” is. Japan’s F-15J pilot certainly knew: his cockpit must have exploded into warning alarms; Imagine that shrill beeping screaming through the cockpit for half an hour - enough to drive any normal person mad; his breathing must have turned shallow; his hands probably shook as he tried to maneuver away from the lock. However, even under that level of crushing psychological pressure, the Japanese pilot chose to stay inside the zone rather than fleeing. This is kamicaze level provocation. Because the moment fire-control goes live, the radar narrows into a focused beam, feeding exact parameters to the missile under the jet’s wing. In peacetime exercises, a sustained lock counts as a confirmed kill. In real combat, nobody activates it unless they are ready to shoot. And China kept that beam on the F-15J for over half an hour. First lock: 16:32–16:35, three full minutes. Second lock: 18:37–19:08, more than thirty minutes. There is no suspense about how this confrontation would end in a real war. It would be a guaranteed kill. The Cat and the Mouse To be precise, the J-15 wasn’t merely locking and unlocking. It was playing — the way a cat toys with a mouse trapped under its paw. Japan’s F-15J was that mouse.

America-China Watcher

21,361 次观看 • 10 个月前

I've been editing this article about "brain mapping" and connectomics, and I'm just stunned by how quickly the cost estimates to map, say, a mouse brain have plummeted in just the last couple years. It actually seems feasible that we could map the entire human brain -- all 86 billion neurons, and their connections -- in this lifetime. In the 1970s, Sydney Brenner started mapping all the connections between neurons in C. elegans. His team sliced the worm into thin pieces, took photos using an electron microscope, and manually traced and reconstructed each synapse for 302 neurons total. This project took more than a decade of work, and it cost about $16,500 to reconstruct each neuron. Scaling this up to a human brain boggles the mind. Electron microscopy remained the norm in connectomics for decades, because it was the only option available to see synapses at a resolution high enough to be able to trace their paths. Each electron microscope costs several hundreds of thousands of dollars, though, and you need lots of them to map even a mouse brain in a reasonable timeframe. In 2023, the Wellcome Trust released a report estimating how long, and how expensive, it would be to map the mouse connectome (~70M neurons). They estimated that imaging alone would cost $200-300M, and that proofreading (or ensuring that traces between neurons are correct) would cost $7-21 BILLION. (A human can only manually trace about 1 mm of neuron per hour.) Also, the images would occupy about 500 petabytes of data, and getting those data would require 20 electron microscopes running in parallel for about 5 years, continuously. They estimated the whole project would take about 17 years of work. This is, understandably, insane. But now it seems like there's an actual path toward mapping the full mouse brain in about five years for ~$100M dollars. There have been three major breakthroughs in the last year or so: 1/ Expansion microscopy, first developed in 2015, showed that it's possible to "enlarge" the brain by about 5x using a swellable polymer. But an improved method increases this number to >20x expansion, meaning we can now expand brains and image neurons much more easily using cheap light microscopes, rather than expensive electron ones. 2/ E11 Bio (a nonprofit research org) developed protein barcodes that get delivered into brain tissue; each neuron gets a unique combination of barcodes. These cells are then stained with colorful antibodies, which stick to a matching protein barcode, causing each neuron to light up in a distinct color. This makes tracing neurons so much easier. 3/ Google Research released PATHFINDER this May, an AI-based neuron tracing tool that can proofread about 67,200 cubic microns of brain tissue per hour, with very high accuracy. It works on electron micrographs, but something similar could be presumably be developed for the E11 / colorful tag approach. This is an extremely exciting time for neuroscience. (C. elegans connectome below.)

I've been editing this article about "brain mapping" and connectomics, and I'm just stunned by how quickly the cost estimates to map, say, a mouse brain have plummeted in just the last couple years. It actually seems feasible that we could map the entire human brain -- all 86 billion neurons, and their connections -- in this lifetime. In the 1970s, Sydney Brenner started mapping all the connections between neurons in C. elegans. His team sliced the worm into thin pieces, took photos using an electron microscope, and manually traced and reconstructed each synapse for 302 neurons total. This project took more than a decade of work, and it cost about $16,500 to reconstruct each neuron. Scaling this up to a human brain boggles the mind. Electron microscopy remained the norm in connectomics for decades, because it was the only option available to see synapses at a resolution high enough to be able to trace their paths. Each electron microscope costs several hundreds of thousands of dollars, though, and you need lots of them to map even a mouse brain in a reasonable timeframe. In 2023, the Wellcome Trust released a report estimating how long, and how expensive, it would be to map the mouse connectome (~70M neurons). They estimated that imaging alone would cost $200-300M, and that proofreading (or ensuring that traces between neurons are correct) would cost $7-21 BILLION. (A human can only manually trace about 1 mm of neuron per hour.) Also, the images would occupy about 500 petabytes of data, and getting those data would require 20 electron microscopes running in parallel for about 5 years, continuously. They estimated the whole project would take about 17 years of work. This is, understandably, insane. But now it seems like there's an actual path toward mapping the full mouse brain in about five years for ~$100M dollars. There have been three major breakthroughs in the last year or so: 1/ Expansion microscopy, first developed in 2015, showed that it's possible to "enlarge" the brain by about 5x using a swellable polymer. But an improved method increases this number to >20x expansion, meaning we can now expand brains and image neurons much more easily using cheap light microscopes, rather than expensive electron ones. 2/ E11 Bio (a nonprofit research org) developed protein barcodes that get delivered into brain tissue; each neuron gets a unique combination of barcodes. These cells are then stained with colorful antibodies, which stick to a matching protein barcode, causing each neuron to light up in a distinct color. This makes tracing neurons so much easier. 3/ Google Research released PATHFINDER this May, an AI-based neuron tracing tool that can proofread about 67,200 cubic microns of brain tissue per hour, with very high accuracy. It works on electron micrographs, but something similar could be presumably be developed for the E11 / colorful tag approach. This is an extremely exciting time for neuroscience. (C. elegans connectome below.)

Niko McCarty.

67,157 次观看 • 10 个月前

Good old days at OpenAI in 2016: an agent stares at screen pixels, moves a mouse, and books a flight on United. We called it World of Bits, inside OpenAI Universe. 10 yrs later, Astra is reincarnated in the same universe. Even the naming is astronomically correct 😆 Universe was perhaps the most ambitious AI infra project at the time, but we couldn't quite figure out how to solve it. A policy with zero prior knowledge of what a "submit" button does has to rediscover the entire internet visual lingua by trial and error. In retrospect, RL from scratch against hand-drawn, per-task "artisan" reward functions on a bunch of Pascal Titan X GPUs was completely doomed. To solve computer use agent, the right way turns out to be boiling the ocean first (hillclimb on every general task you can find), and then specialize back down to the screen pixels and keystrokes. Or simply, a "Specialized Generalist". Lessons learned: one step ahead of everyone, you're a pioneer. Three steps ahead, you're a prophet. Five steps ahead, you're a martyr. Congrats, GPT-6! That United flight finally gets booked, reliably this time. The 2016 intern in me has a big smile.

Good old days at OpenAI in 2016: an agent stares at screen pixels, moves a mouse, and books a flight on United. We called it World of Bits, inside OpenAI Universe. 10 yrs later, Astra is reincarnated in the same universe. Even the naming is astronomically correct 😆 Universe was perhaps the most ambitious AI infra project at the time, but we couldn't quite figure out how to solve it. A policy with zero prior knowledge of what a "submit" button does has to rediscover the entire internet visual lingua by trial and error. In retrospect, RL from scratch against hand-drawn, per-task "artisan" reward functions on a bunch of Pascal Titan X GPUs was completely doomed. To solve computer use agent, the right way turns out to be boiling the ocean first (hillclimb on every general task you can find), and then specialize back down to the screen pixels and keystrokes. Or simply, a "Specialized Generalist". Lessons learned: one step ahead of everyone, you're a pioneer. Three steps ahead, you're a prophet. Five steps ahead, you're a martyr. Congrats, GPT-6! That United flight finally gets booked, reliably this time. The 2016 intern in me has a big smile.

Jim Fan

85,998 次观看 • 1 个月前

Ok so I was actually working on this so my wife would let me keep Linux on this 2-in-1 tablet PC. Completely vibe coded but it borrows the same philosophy where SUPER + SHIFT + Key is for applications while SUPER + CTRL + Key is for system stuff (btop, bluetooth, etc) - i.e., separation of concerns. Here it's 1 finger for scrolling within the app, 2 fingers for rearranging the windows, 3 fingers for app-level manipulation like fullscreen, dismiss, etc. 4 fingers for workspaces. Same "just sit down and learn it bitch" UX but actually quite good and better than iOS once you get used to it. Just a first pass though - I should really join the Discord and see if someone else is working on this. I'm not sure if this should land in Hyprland or as an Omarchy plugin. And no I'm not trying to make a political statement, that's just an awesome mouse pad. Recorded this video earlier to show my wife lol.

Ok so I was actually working on this so my wife would let me keep Linux on this 2-in-1 tablet PC. Completely vibe coded but it borrows the same philosophy where SUPER + SHIFT + Key is for applications while SUPER + CTRL + Key is for system stuff (btop, bluetooth, etc) - i.e., separation of concerns. Here it's 1 finger for scrolling within the app, 2 fingers for rearranging the windows, 3 fingers for app-level manipulation like fullscreen, dismiss, etc. 4 fingers for workspaces. Same "just sit down and learn it bitch" UX but actually quite good and better than iOS once you get used to it. Just a first pass though - I should really join the Discord and see if someone else is working on this. I'm not sure if this should land in Hyprland or as an Omarchy plugin. And no I'm not trying to make a political statement, that's just an awesome mouse pad. Recorded this video earlier to show my wife lol.

gabe

14,305 次观看 • 2 个月前

home page hero ✨ Design notes: - "Forever" hero text dot pixel FX done in Unicorn Studio. (I will do a whole tutorial on this later. Unicorn's WebGL engine is absolutely wild and very powerful / robust) - built in Framer - I wanted to recreate the iOS unlock effect where your home screen icons cascade into place in a beautifully timed choreography. This took a lot of careful timing using Framer's "Appear" effect on the hero text and surrounding avatars because it was super important that we didn't lose the legibility of our main message ("Build Your Forever Audience") with all the animations. - If you look closely, the choreography is setup to lead your eye through the hero text first starting with "Build Your" then "Forever" and finally "Audience." - With those text layers in place + the surrounding avatars, there is a slight 1 sec pause before the remaining elements slide in below and above (How it works, CTA buttons, announcement badge, and lastly the main nav). - All told the entire loading sequence is 6 seconds - Custom particle system powers the interactive star field (the stars slowly gravitate to your pointer position, and the star field perspective changes ever so subtly as you move your mouse around on the page) - I have 3 shooting stars made of small white line layers that start out off canvas rotated at different angles that shoot across to another point off canvas at random times on a loop effect. - Given this hero scene is in space, I wanted the surrounding avatar elements to "float" in low gravity mode. For this I used Framer's loop effect that slowly oscillates the layer's y position. I then offset the delay of each element randomly to stagger the floating loop so each avatar floats independently/randomly - The final major treatment for this hero scene was the scroll animations on the avatars. I wanted to create a bit of a warp speed effect when you scroll down, as if the avatars were being pulled or sucked into a worm hole as you scroll down below this hero fold. - To accomplish this, I applied Framer's scroll transform affect set to "section in view" on each of the floating avatars, and set the "scroll to" position of the upper avatars to be much, much further away on the y-axis than the "scroll to" position of the lower avatars. (eg. -1700px on upper most avatars vs. -600px on lowest positioned avatars). This effectively causes the upper avatars to slide up off the hero canvas with much greater velocity than the lower positioned avatars. - And when you scroll back up to the hero section, the inverse happens where the lower positioned avatars "arrive back in place" from up above the hero canvas before the upper avatars come back into the scene and settle in place. - Overall I wanted this hero section to feel alive. The floating avatars, particle system with very subtle star movements, and the Caustics effect on the "Forever" text all sort of move at the pace of slow breathing - which is a great pace to create a sense of life and comfort in your scene. Conversion Results (so far) - When this new Framer site launched along with Calaxy v1.9 release on Base a couple weeks ago, we saw a surge in traffic, around 20k page views in the first few days. - Of those 20k hits, 11k visited the app install page ( - which is our main CTA - We saw around 10K new users in the first week after v1.9 launch Overall I'm very happy with the new site and early performance metrics. Lots of tweaking to do but its a good start. If you are a designer building in Framer - hit me with any questions on the above hero notes. Happy to share more specifics! 👾

Chadd Weston

16,445 次观看 • 1 年前

THE DEPTH MAP TRICK THAT FIXED DANCE ACCURACY IN SEEDANCE 2.0 Feed the model a video of someone dancing and it tries to interpret everything- the person, the clothes, the lighting, the room, and somewhere in there, the movement. Feed it a depth map and there's nothing left to interpret but the motion. Most creators trying to transfer a dance to a character reference the source footage directly, then wonder why the choreography drifts. The problem isn't the model - it's that you handed it ten variables when you only wanted one. Here's the workflow 1. Lock the character reference in GPT Image 2 first -face, build, costume, so identity holds independently of whatever motion gets applied to it 2. Convert the source dance footage into a depth map instead of using the raw video -this strips out the original performer's appearance, clothing, and environment entirely 3. Feed the depth map as the motion reference and the character sheet as the identity reference- two separate inputs doing two separate jobs, not one input trying to do both 5. Let the depth map carry only spatial movement -the model receives body position and momentum with no competing information about who's moving or what they look like 6. Keep the character and motion inputs isolated throughout - the moment you mix appearance data into the motion reference, the model starts negotiating between two identities Why this works • Raw footage passes the model everything at once- performer, wardrobe, room, lighting -and the choreography competes with all of it for attention • A depth map is pure spatial information, so the only thing left to transfer is movement • Separating identity from motion means the character can stay locked while the dance stays accurate - normally you're trading one for the other • The accuracy gain isn't the model getting better, it's the model getting fewer decisions to make Use cases: ⁃ Dance and choreography transfer onto original characters ⁃ Motion capture-style workflows without motion capture ⁃ Any sequence where a specific movement needs to survive intact ⁃ Character showcase content built on existing performance footage The character sheet answers who's dancing. The depth map answers how - and keeping those two questions separate is the whole trick.

THE DEPTH MAP TRICK THAT FIXED DANCE ACCURACY IN SEEDANCE 2.0 Feed the model a video of someone dancing and it tries to interpret everything- the person, the clothes, the lighting, the room, and somewhere in there, the movement. Feed it a depth map and there's nothing left to interpret but the motion. Most creators trying to transfer a dance to a character reference the source footage directly, then wonder why the choreography drifts. The problem isn't the model - it's that you handed it ten variables when you only wanted one. Here's the workflow 1. Lock the character reference in GPT Image 2 first -face, build, costume, so identity holds independently of whatever motion gets applied to it 2. Convert the source dance footage into a depth map instead of using the raw video -this strips out the original performer's appearance, clothing, and environment entirely 3. Feed the depth map as the motion reference and the character sheet as the identity reference- two separate inputs doing two separate jobs, not one input trying to do both 5. Let the depth map carry only spatial movement -the model receives body position and momentum with no competing information about who's moving or what they look like 6. Keep the character and motion inputs isolated throughout - the moment you mix appearance data into the motion reference, the model starts negotiating between two identities Why this works • Raw footage passes the model everything at once- performer, wardrobe, room, lighting -and the choreography competes with all of it for attention • A depth map is pure spatial information, so the only thing left to transfer is movement • Separating identity from motion means the character can stay locked while the dance stays accurate - normally you're trading one for the other • The accuracy gain isn't the model getting better, it's the model getting fewer decisions to make Use cases: ⁃ Dance and choreography transfer onto original characters ⁃ Motion capture-style workflows without motion capture ⁃ Any sequence where a specific movement needs to survive intact ⁃ Character showcase content built on existing performance footage The character sheet answers who's dancing. The depth map answers how - and keeping those two questions separate is the whole trick.

Nexlow

120,346 次观看 • 1 个月前