Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Hey guys, I built a controller diagnostics tool last night - runs right in the browser, no install. Test polling rate & Latency (Any Controller) Stick Circularity & Drift (PS & Marius) Full DualSense support: recenter sticks, range-calibrate, lightbar & more Link to Github👇

TheQuest

5,193 subscribers

27,440 views • 3 months ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Two weeks since launch, Devin Review runs >40,000 times a day. Thanks for the love & feedback. Some updates based on popular demand: - One-click-apply fixes - Merge button - REVIEW md support Swap github -> devinreview in any PR link to try, no account needed.

Two weeks since launch, Devin Review runs >40,000 times a day. Thanks for the love & feedback. Some updates based on popular demand: - One-click-apply fixes - Merge button - REVIEW md support Swap github -> devinreview in any PR link to try, no account needed.

Cognition

31,818 views • 6 months ago

I’m a 21-year-old & no matter what polls say, I know President Biden has delivered more for me & my generation than any president in our lifetimes—& that’s why we will re-elect Joe Biden in 2024. Joined MSNBC to talk why we shouldn’t freak out about the polls. Watch & share👇

I’m a 21-year-old & no matter what polls say, I know President Biden has delivered more for me & my generation than any president in our lifetimes—& that’s why we will re-elect Joe Biden in 2024. Joined MSNBC to talk why we shouldn’t freak out about the polls. Watch & share👇

Victor Shi

660,771 views • 2 years ago

"We're a no-excuse team…figure it out. If we can't maybe I need to go deeper in bench & play guys less minutes…We can't say 'we played last nite'–no–go win…If you leave it on floor [&] take L, on to the next one" – Mike Brown on Knicks effort losing 2nd night of back-to-back

"We're a no-excuse team…figure it out. If we can't maybe I need to go deeper in bench & play guys less minutes…We can't say 'we played last nite'–no–go win…If you leave it on floor [&] take L, on to the next one" – Mike Brown on Knicks effort losing 2nd night of back-to-back

New York Basketball

45,978 views • 9 months ago

Any Manchester girls/guys coming this Saturday? My last signing in Waterstones & looking back at all the pictures has reminded me how lucky I am (It’s clearly been a fun experience) THANK YOU to everyone who has come along ♥️🥀♥️ {ticket link for…} ⬇️

Any Manchester girls/guys coming this Saturday? My last signing in Waterstones & looking back at all the pictures has reminded me how lucky I am (It’s clearly been a fun experience) THANK YOU to everyone who has come along ♥️🥀♥️ {ticket link for…} ⬇️

Jay McGuiness

11,593 views • 2 years ago

[New Xbox Controller Review] #ad Special thanks to Qrdgame_offcial for supplying me with a review unit of their Ferrox M5 Xbox & PC Controller QRD Ferrox M5 The QRD Ferrox M5 is a Wireless Xbox One–style gamepad with a transparent design. Since it targets Xbox One natively, it works automatically on PC and Xbox Series S/X. The included 2.4Ghz USB Dongle also supports Switch and Android with the right adapter (USB C), but I was not able to test those in time for this review. A Wireless Xbox Controller The QRD FERROX M5 comes in a familiar Xbox layout with Hall Effect joysticks, RGB rings around the sticks, and a magnetic & detachable clear shell that shows the internals. The controller feels very similar to a standard Xbox pad in terms of shape and size, but it’s slightly lighter at compared to the usual Xbox controller weight. The lighter weight makes it easier to hold for longer sessions without hand fatigue. The D-Pad is surprisingly good at dialing special moves feels as solid and responsive as the official Xbox controller, and that’s impressive at a lower price point. I spent a few days trying out the Mortal Kombat Legacy Kollection exclusively using this controller and I did not notice any difference at all. Capabilities From what I’ve tested, the controller performs extremely well for action games, fighting games, and shooters. The Hall Effect joysticks offer smooth movement and precise control with no drift issues. The D-Pad uses micro switches, so every input is tactile and accurate. The back triggers support an adaptive “trigger stop” mode, which shortens the travel distance for faster reaction time in FPS games or quicker blocks in Fighting Games. You also get 4 back macro buttons, turbo mode, and adjustable joystick sensitivity with deadzone options. Vibration strength and RGB lighting modes are also customizable. What can it do The FERROX M5 includes an 800mAh battery that lasts around 8 hours of gameplay. It only uses the 2.4Ghz wireless connection, not Bluetooth, so it always requires the included USB Dongle. The good thing is once you sync it with your Xbox or PC, you don’t need to sync it again—it carries over between both devices automatically. The 2.4G connection is stable and low latency, and the controller has a 3.5mm headphone jack for audio and mic support. It also features a quick calibration system to keep sticks and triggers accurate over time. The overall grip and shape feel close to the official Xbox controller, comfortable for long gaming sessions, and the transparent shell gives it a distinct look. Areas of improvement I initially had trouble charging the controller while playing because connecting it to the Xbox via USB would disable all inputs, making it unusable in wired mode. But luckily QRD Support helped and provided me with an updat - I had to connect it to a PC, download the firmware tool from QRD’s website, and update the controller manually. After installing the new firmware, the controller can now charge while being used normally on Xbox, so the issue is fully resolved. The Xbox button also behaves differently on Xbox, where the single press acts like hold and the hold acts like single press; this issue doesn’t appear on PC. Syncing with the dongle took a few attempts on both Xbox and PC, but once paired, it never needed syncing again. The wireless mode is strictly 2.4G with no Bluetooth support, so it cannot connect without the dongle, though the benefit is that one pairing works across both Xbox and PC. Final Summary ✅ Hall Effect Joystick ✅ Lights in the dark ✅ Good D-Pad for input dialing ✅ Light Controller ❌ First sync can be annoying on PC ❌ No bluetooth for phone usage - Dongle is necessary ❌ Needs an update for the Xbox Wired issues which require a PC The Controller is available on Amazon for $59.99, and will be part of the Black Friday Deals on their official website for even more discount! Amazon Link Website Link

[New Xbox Controller Review] #ad Special thanks to Qrdgame_offcial for supplying me with a review unit of their Ferrox M5 Xbox & PC Controller QRD Ferrox M5 The QRD Ferrox M5 is a Wireless Xbox One–style gamepad with a transparent design. Since it targets Xbox One natively, it works automatically on PC and Xbox Series S/X. The included 2.4Ghz USB Dongle also supports Switch and Android with the right adapter (USB C), but I was not able to test those in time for this review. A Wireless Xbox Controller The QRD FERROX M5 comes in a familiar Xbox layout with Hall Effect joysticks, RGB rings around the sticks, and a magnetic & detachable clear shell that shows the internals. The controller feels very similar to a standard Xbox pad in terms of shape and size, but it’s slightly lighter at compared to the usual Xbox controller weight. The lighter weight makes it easier to hold for longer sessions without hand fatigue. The D-Pad is surprisingly good at dialing special moves feels as solid and responsive as the official Xbox controller, and that’s impressive at a lower price point. I spent a few days trying out the Mortal Kombat Legacy Kollection exclusively using this controller and I did not notice any difference at all. Capabilities From what I’ve tested, the controller performs extremely well for action games, fighting games, and shooters. The Hall Effect joysticks offer smooth movement and precise control with no drift issues. The D-Pad uses micro switches, so every input is tactile and accurate. The back triggers support an adaptive “trigger stop” mode, which shortens the travel distance for faster reaction time in FPS games or quicker blocks in Fighting Games. You also get 4 back macro buttons, turbo mode, and adjustable joystick sensitivity with deadzone options. Vibration strength and RGB lighting modes are also customizable. What can it do The FERROX M5 includes an 800mAh battery that lasts around 8 hours of gameplay. It only uses the 2.4Ghz wireless connection, not Bluetooth, so it always requires the included USB Dongle. The good thing is once you sync it with your Xbox or PC, you don’t need to sync it again—it carries over between both devices automatically. The 2.4G connection is stable and low latency, and the controller has a 3.5mm headphone jack for audio and mic support. It also features a quick calibration system to keep sticks and triggers accurate over time. The overall grip and shape feel close to the official Xbox controller, comfortable for long gaming sessions, and the transparent shell gives it a distinct look. Areas of improvement I initially had trouble charging the controller while playing because connecting it to the Xbox via USB would disable all inputs, making it unusable in wired mode. But luckily QRD Support helped and provided me with an updat - I had to connect it to a PC, download the firmware tool from QRD’s website, and update the controller manually. After installing the new firmware, the controller can now charge while being used normally on Xbox, so the issue is fully resolved. The Xbox button also behaves differently on Xbox, where the single press acts like hold and the hold acts like single press; this issue doesn’t appear on PC. Syncing with the dongle took a few attempts on both Xbox and PC, but once paired, it never needed syncing again. The wireless mode is strictly 2.4G with no Bluetooth support, so it cannot connect without the dongle, though the benefit is that one pairing works across both Xbox and PC. Final Summary ✅ Hall Effect Joystick ✅ Lights in the dark ✅ Good D-Pad for input dialing ✅ Light Controller ❌ First sync can be annoying on PC ❌ No bluetooth for phone usage - Dongle is necessary ❌ Needs an update for the Xbox Wired issues which require a PC The Controller is available on Amazon for $59.99, and will be part of the Black Friday Deals on their official website for even more discount! Amazon Link Website Link

thethiny 🐰🍉

16,864 views • 9 months ago

LFT Can actually play any role & IGL if needed. Willing to relocate. I can bring a lot to a team just give me a shot. I will be making a lot more of these vids I have 250+ clips to go through so make sure to sub. full vid + extra clips in the link below:D RTs are appreciated😊

LFT Can actually play any role & IGL if needed. Willing to relocate. I can bring a lot to a team just give me a shot. I will be making a lot more of these vids I have 250+ clips to go through so make sure to sub. full vid + extra clips in the link below:D RTs are appreciated😊

HUYNH

120,315 views • 3 years ago

pov: you're running 15 loops across in glm 5.2 in opencode, codex, claude, amp or pi and have no idea which one needs you, bring em all to opensessions cmux, superset - cool ideas, but i got tired of blank screens, glitchy terminals, and macos-only apps i just wanted to know when my agents finish, fail, or need me so i built a tmux sidebar. now re-written in rust with worktree support runs inside your actual terminal. on any OS. - agents status' across all your projects at a glance - forwarded ports per project - live threads across amp, claude code, codex, opencode no electron. no blank panes. no new app. extremely fast and light. open source.

pov: you're running 15 loops across in glm 5.2 in opencode, codex, claude, amp or pi and have no idea which one needs you, bring em all to opensessions cmux, superset - cool ideas, but i got tired of blank screens, glitchy terminals, and macos-only apps i just wanted to know when my agents finish, fail, or need me so i built a tmux sidebar. now re-written in rust with worktree support runs inside your actual terminal. on any OS. - agents status' across all your projects at a glance - forwarded ports per project - live threads across amp, claude code, codex, opencode no electron. no blank panes. no new app. extremely fast and light. open source.

Palani — oss/acc

13,971 views • 2 months ago

Introducing 🌸 Kanna – a web ui for Claude Code + Codex running right in your browser (and my first open source project!) It's sort of like if Claude Code & Codex had a baby with a much better ui/ux 💅 Here's what makes Kanna special: 🔀 One-click switch between Claude Code / Codex 🌐 Runs in your browser on localhost - no app switching 🧩 Embedded split terminals, persisted between chats with a beautiful horizontal scrolling UX 📁 Full project file browser built in, embedded editor coming soon too. Coming soon: 🔒 End-to-end encryption + remote access via reverse proxy 🌿 Git integration, diffs, PRs 🧱 Plugins I wanted to replace switching between cursor, the github app, browser & (most recently) the codex app. The browser is hardest to recreate yet also the most flexible so it makes sense to build in the browser itself. Install in 5 seconds: bun install -g kanna-code then just type: kanna Kanna is purely a ui/ux layer. It uses your existing CLIs, is 100% compliant & no data ever leaves your machine. If running claude or codex in your terminal works, kanna works too. Kanna natively supports every tool call, model, reasoning effort, fast mode, plan mode, compaction, user questions, web searching, skills, agents, mcps, everything. There are a few libraries that do this but none that I found as comprehensive. All feedback welcome!

Introducing 🌸 Kanna – a web ui for Claude Code + Codex running right in your browser (and my first open source project!) It's sort of like if Claude Code & Codex had a baby with a much better ui/ux 💅 Here's what makes Kanna special: 🔀 One-click switch between Claude Code / Codex 🌐 Runs in your browser on localhost - no app switching 🧩 Embedded split terminals, persisted between chats with a beautiful horizontal scrolling UX 📁 Full project file browser built in, embedded editor coming soon too. Coming soon: 🔒 End-to-end encryption + remote access via reverse proxy 🌿 Git integration, diffs, PRs 🧱 Plugins I wanted to replace switching between cursor, the github app, browser & (most recently) the codex app. The browser is hardest to recreate yet also the most flexible so it makes sense to build in the browser itself. Install in 5 seconds: bun install -g kanna-code then just type: kanna Kanna is purely a ui/ux layer. It uses your existing CLIs, is 100% compliant & no data ever leaves your machine. If running claude or codex in your terminal works, kanna works too. Kanna natively supports every tool call, model, reasoning effort, fast mode, plan mode, compaction, user questions, web searching, skills, agents, mcps, everything. There are a few libraries that do this but none that I found as comprehensive. All feedback welcome!

Jake Mor

43,379 views • 5 months ago

THIS GUY REBUILT A CHILDHOOD PS2 GAME AS A FREE BROWSER VERSION WITH CLAUDE CODE he grew up obsessed with downhill domination, the ps2 downhill mountain biking game that never got a sequel so he rebuilt the entire thing from scratch > its a full downhill biking game you play right in your browser, nothing to install > the whole thing is one single html file with no separate image, model, or sound files > everything, the track, the bike, the physics, is generated in code, not loaded from assets > built with fable, opus, vanilla js and three.js over multiple sessions of iterating someone even sent a pull request to add controller support, he merged it, and people helped him squash bugs on the fly a dead ps2 game came back to life as an open source community project this is the best use of this stuff

THIS GUY REBUILT A CHILDHOOD PS2 GAME AS A FREE BROWSER VERSION WITH CLAUDE CODE he grew up obsessed with downhill domination, the ps2 downhill mountain biking game that never got a sequel so he rebuilt the entire thing from scratch > its a full downhill biking game you play right in your browser, nothing to install > the whole thing is one single html file with no separate image, model, or sound files > everything, the track, the bike, the physics, is generated in code, not loaded from assets > built with fable, opus, vanilla js and three.js over multiple sessions of iterating someone even sent a pull request to add controller support, he merged it, and people helped him squash bugs on the fly a dead ps2 game came back to life as an open source community project this is the best use of this stuff

Om Patel

22,230 views • 1 month ago

GoldenEye 007 N64 running natively on iOS and iPadOS for the first time. Not an emulator. The GoldenEye decompilation community spent years reconstructing the original game. GoldenEye64Recomp, N64Recomp, N64ModernRuntime, and RT64 provided the static recompilation, runtime, and rendering foundations. I built GoldenPad on top: the iPhone and iPad app, touch interface, native audio and saves, controller support, and the mobile lifecycle work needed to make it run on real devices. It runs great on iPhone. GoldenEye runs better through GoldenPad than it has in any N64 emulator I have tried on iOS. Native ARM64 code, high-resolution Metal rendering, and modern dual-stick controls make a huge difference. No JIT, no N64 code being interpreted at runtime, and no streaming from a Mac. This is the original 1997 N64 game, not the XBLA version, and it is not based on GoldenEye Metal. I am actively working on full local split-screen multiplayer for iPad. MacOS support soon. I built the iOS and iPadOS integration with 5.6 Sol. Open sourcing it all below. Happy Friday!

GoldenEye 007 N64 running natively on iOS and iPadOS for the first time. Not an emulator. The GoldenEye decompilation community spent years reconstructing the original game. GoldenEye64Recomp, N64Recomp, N64ModernRuntime, and RT64 provided the static recompilation, runtime, and rendering foundations. I built GoldenPad on top: the iPhone and iPad app, touch interface, native audio and saves, controller support, and the mobile lifecycle work needed to make it run on real devices. It runs great on iPhone. GoldenEye runs better through GoldenPad than it has in any N64 emulator I have tried on iOS. Native ARM64 code, high-resolution Metal rendering, and modern dual-stick controls make a huge difference. No JIT, no N64 code being interpreted at runtime, and no streaming from a Mac. This is the original 1997 N64 game, not the XBLA version, and it is not based on GoldenEye Metal. I am actively working on full local split-screen multiplayer for iPad. MacOS support soon. I built the iOS and iPadOS integration with 5.6 Sol. Open sourcing it all below. Happy Friday!

Kahris

539,756 views • 11 days ago

Introducing GameSir Tarantula Pro for Xbox🟢 Tarantula Pro for Xbox | The Best Xbox Controller in Symmetry ⚖️ -Highly-refined ergonomic mould. Providing the perfect grip that many players have been searching for.🍃 -GameSir Anti-stick drift Mag-Res™ TMR Sticks. Designed for pro-level performance, deliver smooth, linear control with ultra-low power consumption. 🕹️ -Motion Control on PC. Built-in 6-axis gyroscope enables motion control and gyro remapping on PC, with customizable settings via GameSir Nexus software.🔬 -Dual-Mode Trigger. Change between Micro Switch triggers / Hall Effect analog triggers. From FPS to racing, enjoy seamless performance across different games.🏁 -1000Hz polling rate, mechanical D-pad, and micro-switch ABXY deliver instant feedback. With nine remappable buttons and GameSir Nexus software support, customize your ultimate setup as you wish. 🎮 Be competitive. Be Pro. Pre-order & learn more:

Introducing GameSir Tarantula Pro for Xbox🟢 Tarantula Pro for Xbox | The Best Xbox Controller in Symmetry ⚖️ -Highly-refined ergonomic mould. Providing the perfect grip that many players have been searching for.🍃 -GameSir Anti-stick drift Mag-Res™ TMR Sticks. Designed for pro-level performance, deliver smooth, linear control with ultra-low power consumption. 🕹️ -Motion Control on PC. Built-in 6-axis gyroscope enables motion control and gyro remapping on PC, with customizable settings via GameSir Nexus software.🔬 -Dual-Mode Trigger. Change between Micro Switch triggers / Hall Effect analog triggers. From FPS to racing, enjoy seamless performance across different games.🏁 -1000Hz polling rate, mechanical D-pad, and micro-switch ABXY deliver instant feedback. With nine remappable buttons and GameSir Nexus software support, customize your ultimate setup as you wish. 🎮 Be competitive. Be Pro. Pre-order & learn more:

GameSir

45,426 views • 4 months ago

Your coding agent can run all night. It still can't tell if what it built actually works. Today we're open-sourcing the TestSprite CLl (Apache-2.0) A tool your agent calls on its own to test your app end-to-end like a real user, fix what broke, and re-check everything it ever got right. It's the same engine 100,000+ teams already use. We proved it in public, on a public leaderboard: Most correct app on the board:89% Built by the cheapest model in the field At half the cost of the priciest one You no longer need the biggest, most expensive model to ship software you can trust. Setup is 2 commands: npm install -g @testsprite/testsprite-cli testsprite init That's the last command you'll ever type - from there, your agent runs the tests itself.

Your coding agent can run all night. It still can't tell if what it built actually works. Today we're open-sourcing the TestSprite CLl (Apache-2.0) A tool your agent calls on its own to test your app end-to-end like a real user, fix what broke, and re-check everything it ever got right. It's the same engine 100,000+ teams already use. We proved it in public, on a public leaderboard: Most correct app on the board:89% Built by the cheapest model in the field At half the cost of the priciest one You no longer need the biggest, most expensive model to ship software you can trust. Setup is 2 commands: npm install -g @testsprite/testsprite-cli testsprite init That's the last command you'll ever type - from there, your agent runs the tests itself.

TestSprite

1,042,264 views • 2 months ago

I built a macOS app for benchmarking local LLMs. 6 test suites. Multiple providers. One workspace. Open source. There are hundreds of local models now. New ones every week. How do you actually pick one? Leaderboards test for general ability. But if you're building an agent that chains tool calls, or a pipeline that extracts structured data, or a code assistant that needs to debug Rust, you need to know if the model handles that specific thing. Not in theory. On your hardware. With your prompts. The benchmarks that exist are either locked behind papers, too abstract to map to real failures, or impossible to extend. You can't add your own test cases. You can't test what matters to your use case. That's what BenchLocal is for. It's a benchmark platform where every test is practical, deterministic, and built around real-world tasks. And you can build your own tests. It ships with 6 Bench Packs TODAY: → ToolCall-15 — tool-use accuracy → BugFind-15 — debugging capabilities → DataExtract-15 — structured data extraction → InstructFollow-15 — constraint-heavy instruction following → ReasonMath-15 — practical reasoning and math → StructOutput-15 — validator-backed structured output Every pack has 15 fixed scenarios. Every score is deterministic and verifiable. Some of you saw ToolCall-15 and BugFind-15 — the individual test packs I open-sourced over the past few weeks. People ran them, filed issues, sent PRs. But managing separate repos, separate scripts, separate results doesn't scale. BenchLocal puts everything in one place. What the app does: > Workspace with tabs — run BugFind-15 in one tab, ToolCall-15 in another. > Any provider — Ollama, llama.cpp, OpenRouter, any OpenAI-compatible endpoint. Local and cloud, same interface. > Run modes — serial, batch per model, batch per test case, or fully parallel. > Test histories — every run saved. Compare any previous session. But the part I'm most excited about isn't the app. It's the ecosystem. BenchLocal is a platform. Each Bench Pack is a plugin. I'm shipping an SDK so anyone can build their own — test what matters to you, package it, share it. Install and uninstall packs right inside the app, same way you'd manage extensions in VS Code. The registry is GitHub-based, fully public. I built 6 packs. I want the community to build the next 60. Theme system built in too — because if I'm staring at benchmark results for hours, it should at least look good. v0.1.0 is macOS only. Windows and Linux are coming. MIT licensed. Everything — the app, the bench packs, the SDK — is open. PRs welcome. Bench Packs even more welcome.

I built a macOS app for benchmarking local LLMs. 6 test suites. Multiple providers. One workspace. Open source. There are hundreds of local models now. New ones every week. How do you actually pick one? Leaderboards test for general ability. But if you're building an agent that chains tool calls, or a pipeline that extracts structured data, or a code assistant that needs to debug Rust, you need to know if the model handles that specific thing. Not in theory. On your hardware. With your prompts. The benchmarks that exist are either locked behind papers, too abstract to map to real failures, or impossible to extend. You can't add your own test cases. You can't test what matters to your use case. That's what BenchLocal is for. It's a benchmark platform where every test is practical, deterministic, and built around real-world tasks. And you can build your own tests. It ships with 6 Bench Packs TODAY: → ToolCall-15 — tool-use accuracy → BugFind-15 — debugging capabilities → DataExtract-15 — structured data extraction → InstructFollow-15 — constraint-heavy instruction following → ReasonMath-15 — practical reasoning and math → StructOutput-15 — validator-backed structured output Every pack has 15 fixed scenarios. Every score is deterministic and verifiable. Some of you saw ToolCall-15 and BugFind-15 — the individual test packs I open-sourced over the past few weeks. People ran them, filed issues, sent PRs. But managing separate repos, separate scripts, separate results doesn't scale. BenchLocal puts everything in one place. What the app does: > Workspace with tabs — run BugFind-15 in one tab, ToolCall-15 in another. > Any provider — Ollama, llama.cpp, OpenRouter, any OpenAI-compatible endpoint. Local and cloud, same interface. > Run modes — serial, batch per model, batch per test case, or fully parallel. > Test histories — every run saved. Compare any previous session. But the part I'm most excited about isn't the app. It's the ecosystem. BenchLocal is a platform. Each Bench Pack is a plugin. I'm shipping an SDK so anyone can build their own — test what matters to you, package it, share it. Install and uninstall packs right inside the app, same way you'd manage extensions in VS Code. The registry is GitHub-based, fully public. I built 6 packs. I want the community to build the next 60. Theme system built in too — because if I'm staring at benchmark results for hours, it should at least look good. v0.1.0 is macOS only. Windows and Linux are coming. MIT licensed. Everything — the app, the bench packs, the SDK — is open. PRs welcome. Bench Packs even more welcome.

stevibe

50,584 views • 4 months ago

Let's reverse engineer Disney's adorable, lifelike robot! I couldn't find a whitepaper, but this is how I think it's trained: 1. The emotional behaviors are curated by Disney animation artists, keyframe by keyframe. But it cannot be "rendered" directly on the robot because it doesn't take into account the complex real-world physics. 2. Reinforcement learning (RL) is a great tool for training low-level robot controllers. RL needs a reward function to optimize, and it's typically a task reward (e.g. walk in a straight line as fast as possible). The problem is that RL doesn't know what counts as "natural behavior", and often produces weird-looking body postures that somehow still maximize the reward. This is a human alignment problem just like ChatGPT. 3. Enters Adversarial Motion Prior (AMP): a technique that learns the human preference by training a classifier on what we consider "emotional & cute". In GAN literature, this is called a discriminator. Disney artists are good at creating such a dataset. You can then add AMP as an auxiliary reward in simulation to nudge the robot towards desired behaviors. AMP was developed by Peng et al. 2021 and Escontrela et al. 2022. 4. Add lots of data augmentation to make the controller robust to physical disturbances. In RL, it's called "domain randomization". This is a very powerful technique that bridges the gap between simulator and reality. Previously, OpenAI used domain randomization to train a 5-finger robot hand to manipulate a Rubik's Cube: IEEE news article gave hints about the pipeline: Finally, praying for world peace 🙏. I hope robotics like this will bring more joy to the world.

Let's reverse engineer Disney's adorable, lifelike robot! I couldn't find a whitepaper, but this is how I think it's trained: 1. The emotional behaviors are curated by Disney animation artists, keyframe by keyframe. But it cannot be "rendered" directly on the robot because it doesn't take into account the complex real-world physics. 2. Reinforcement learning (RL) is a great tool for training low-level robot controllers. RL needs a reward function to optimize, and it's typically a task reward (e.g. walk in a straight line as fast as possible). The problem is that RL doesn't know what counts as "natural behavior", and often produces weird-looking body postures that somehow still maximize the reward. This is a human alignment problem just like ChatGPT. 3. Enters Adversarial Motion Prior (AMP): a technique that learns the human preference by training a classifier on what we consider "emotional & cute". In GAN literature, this is called a discriminator. Disney artists are good at creating such a dataset. You can then add AMP as an auxiliary reward in simulation to nudge the robot towards desired behaviors. AMP was developed by Peng et al. 2021 and Escontrela et al. 2022. 4. Add lots of data augmentation to make the controller robust to physical disturbances. In RL, it's called "domain randomization". This is a very powerful technique that bridges the gap between simulator and reality. Previously, OpenAI used domain randomization to train a 5-finger robot hand to manipulate a Rubik's Cube: IEEE news article gave hints about the pipeline: Finally, praying for world peace 🙏. I hope robotics like this will bring more joy to the world.

Jim Fan

314,694 views • 2 years ago

Introducing PhoneLLM, an open model for voice agents. GPT 5.6 Terra performance on typical voice agent tasks at 1/3 the latency and 1/18 the cost. For voice agents, we need models that are both very low latency and very good at tool calling and instruction following. There's a trade-off here, and we often have to compromise on either latency or capability when building voice agents. With PhoneLLM (and the training and data stack that made this model possible) we're fixing this problem. For the last couple of years, most of the effort in frontier model development has gone towards leveraging test-time compute. Which is awesome! Models of all shapes and sizes are available that perform really, really well ... if you have "thinking" turned on for your model. But if you need your agent to respond at voice conversation speed, you can't use thinking models. PhoneLLM is a full-weights fine-tune of NVIDIA Nemotron Nano 30B. We trained on a wide range of real-world telephone and customer support use cases. The training focused on taking the excellent Nano 30B base capabilities and teaching the model to do typical voice agent tasks with thinking disabled. The results are really good: accurate tool calling and concise, on-topic responses in long conversations. And fast: TTFAT measured server-side is <100ms if you run PhoneLLM on a lightly loaded B200. :-) But seriously, when we characterize model latency, we do it with full, end-to-end, batched request simulations using real Pipecat voice agent pipelines. You can serve more than 80 concurrent agents on a single B200 with P95 end-to-end TTFAT <600ms. Including network overhead. That's an LLM cost-per-minute around $0.0025. (1/4 of a cent.) At a latency lower than any third-party API offers today. More details about this model, including weights on Hugging Face, how to spin it up with one click on Modal, and a starter project repo you can clone, are in the thread ...

Introducing PhoneLLM, an open model for voice agents. GPT 5.6 Terra performance on typical voice agent tasks at 1/3 the latency and 1/18 the cost. For voice agents, we need models that are both very low latency and very good at tool calling and instruction following. There's a trade-off here, and we often have to compromise on either latency or capability when building voice agents. With PhoneLLM (and the training and data stack that made this model possible) we're fixing this problem. For the last couple of years, most of the effort in frontier model development has gone towards leveraging test-time compute. Which is awesome! Models of all shapes and sizes are available that perform really, really well ... if you have "thinking" turned on for your model. But if you need your agent to respond at voice conversation speed, you can't use thinking models. PhoneLLM is a full-weights fine-tune of NVIDIA Nemotron Nano 30B. We trained on a wide range of real-world telephone and customer support use cases. The training focused on taking the excellent Nano 30B base capabilities and teaching the model to do typical voice agent tasks with thinking disabled. The results are really good: accurate tool calling and concise, on-topic responses in long conversations. And fast: TTFAT measured server-side is <100ms if you run PhoneLLM on a lightly loaded B200. :-) But seriously, when we characterize model latency, we do it with full, end-to-end, batched request simulations using real Pipecat voice agent pipelines. You can serve more than 80 concurrent agents on a single B200 with P95 end-to-end TTFAT <600ms. Including network overhead. That's an LLM cost-per-minute around $0.0025. (1/4 of a cent.) At a latency lower than any third-party API offers today. More details about this model, including weights on Hugging Face, how to spin it up with one click on Modal, and a starter project repo you can clone, are in the thread ...

kwindla

319,488 views • 4 days ago

M64 is to the Nintendo 64 what the iPhone was to the 3310 Developed by ModRetro, the M64 is presented as the ultimate alternative to the original hardware, featuring the most advanced components on the market and the highest-quality finish that enhances every little detail of the system you remember. Presented in packaging with a ’90s aesthetic, this console comes in a dual-textured plastic case (frosted translucent in any of the five available colors or clear). It’s capable of playing original cartridges, reproductions, backups, homebrew, and, of course, remastered releases and new titles from ModRetro. Through its HDMI output, it supports resolutions of 720p, 1080p, 1440p, and 4K, as well as displaying images at their original resolution on a CRT TV via its own adapter. As you’d expect, the system has no region lock, so you can enjoy your NTSC-USA, NTSC-JAP, and PAL-EUR cartridges like never before. Additionally, both the Expansion Pak and—coming soon—the Controller Pak are emulated by the console, ensuring that the entire original catalog is available in the best possible quality. Its architecture is open-source, easy to disassemble, with no degradable electronics, and gold-plated traces and connectors that ensure optimal preservation of the system over the years. Of course, just like with the Chromatic, its firmware is updatable via Wi-Fi or USB cable, and it features a front-end menu that allows, among other things, for detailed on-the-fly customization of resolutions and filters, as well as the implementation of different overclocking levels that take the 64-bit experience to unprecedented heights. And I can’t forget one of its main selling points: the Pro Controller. A controller that stays true to the original form factor but incorporates improved ergonomic solutions, a ceramic-coated aluminum finish, high-end PBT crystalline polymer buttons, and a state-of-the-art TMR joystick coated in soft polymer and free of any drift; which offers a wider travel range, with four levels of spring tension and a greater degree of tilt, achieving unbeatable precision with no dead zones. Additionally, the stick gates are interchangeable, allowing you to customize your experience by installing the classic octagonal, equidistant hexagonal, or circular models. It also features an ERM vibration motor. You can play with the controller connected via cable or wirelessly using Bluetooth 6.2, which reduces response latency to below 0.65 microseconds. It is powered by a rechargeable battery with a USB-C port or by two AA batteries housed in an included module. Like the console, its firmware is updatable and it’s available in the same colors. You can purchase the M64 for $229.99 and the Pro Controller separately (since the system supports your old N64 or third-party controllers) for $89.99. I can guarantee that with this new ModRetro ecosystem, you’ll not only dust off games you may not have touched in decades, but it will also serve as the perfect incentive to dive into one of the most interesting and iconic game libraries—featuring both single-player and multiplayer titles—with undeniable appeal.

M64 is to the Nintendo 64 what the iPhone was to the 3310 Developed by ModRetro, the M64 is presented as the ultimate alternative to the original hardware, featuring the most advanced components on the market and the highest-quality finish that enhances every little detail of the system you remember. Presented in packaging with a ’90s aesthetic, this console comes in a dual-textured plastic case (frosted translucent in any of the five available colors or clear). It’s capable of playing original cartridges, reproductions, backups, homebrew, and, of course, remastered releases and new titles from ModRetro. Through its HDMI output, it supports resolutions of 720p, 1080p, 1440p, and 4K, as well as displaying images at their original resolution on a CRT TV via its own adapter. As you’d expect, the system has no region lock, so you can enjoy your NTSC-USA, NTSC-JAP, and PAL-EUR cartridges like never before. Additionally, both the Expansion Pak and—coming soon—the Controller Pak are emulated by the console, ensuring that the entire original catalog is available in the best possible quality. Its architecture is open-source, easy to disassemble, with no degradable electronics, and gold-plated traces and connectors that ensure optimal preservation of the system over the years. Of course, just like with the Chromatic, its firmware is updatable via Wi-Fi or USB cable, and it features a front-end menu that allows, among other things, for detailed on-the-fly customization of resolutions and filters, as well as the implementation of different overclocking levels that take the 64-bit experience to unprecedented heights. And I can’t forget one of its main selling points: the Pro Controller. A controller that stays true to the original form factor but incorporates improved ergonomic solutions, a ceramic-coated aluminum finish, high-end PBT crystalline polymer buttons, and a state-of-the-art TMR joystick coated in soft polymer and free of any drift; which offers a wider travel range, with four levels of spring tension and a greater degree of tilt, achieving unbeatable precision with no dead zones. Additionally, the stick gates are interchangeable, allowing you to customize your experience by installing the classic octagonal, equidistant hexagonal, or circular models. It also features an ERM vibration motor. You can play with the controller connected via cable or wirelessly using Bluetooth 6.2, which reduces response latency to below 0.65 microseconds. It is powered by a rechargeable battery with a USB-C port or by two AA batteries housed in an included module. Like the console, its firmware is updatable and it’s available in the same colors. You can purchase the M64 for $229.99 and the Pro Controller separately (since the system supports your old N64 or third-party controllers) for $89.99. I can guarantee that with this new ModRetro ecosystem, you’ll not only dust off games you may not have touched in decades, but it will also serve as the perfect incentive to dive into one of the most interesting and iconic game libraries—featuring both single-player and multiplayer titles—with undeniable appeal.

Pre-Rendered Backgrounds Aesthetics

23,444 views • 20 days ago

Nothing beats open-source! We now have a fully open-source coding agent that can compete with any proprietary solution out there: • 500+ available models • It doesn't train on your code • It doesn't charge a penny for tokens • It's open. You can change if you want. Kilo Code is the #1 app on OpenRouter right now, and the community is releasing new features on GitHub at rocket speed (about 20 releases in the last 30 days). I installed it in Visual Studio Code, but you can install it in pretty much any IDE. I need to spend more time with it, but here is a very interesting design choice: Most coding agents support two modes: an "agentic" mode for writing code and an "ask" mode for talking to the model without making any changes. Kilo Code supports 5 different modes: 1. Architect mode - Research, design, and planning 2. Debug mode - Problem solving and troubleshooting 3. Code mode - To write code. 4. Ask mode - To talk to the model. 5. Orchestrator mode - Combine every agent to solve problems. On paper, I really like this. I'll need more time to try out the specialized agents and see whether niching becomes an advantage. Here is the link to the website: Thanks to the Kilo for partnering with me on this post.

Nothing beats open-source! We now have a fully open-source coding agent that can compete with any proprietary solution out there: • 500+ available models • It doesn't train on your code • It doesn't charge a penny for tokens • It's open. You can change if you want. Kilo Code is the #1 app on OpenRouter right now, and the community is releasing new features on GitHub at rocket speed (about 20 releases in the last 30 days). I installed it in Visual Studio Code, but you can install it in pretty much any IDE. I need to spend more time with it, but here is a very interesting design choice: Most coding agents support two modes: an "agentic" mode for writing code and an "ask" mode for talking to the model without making any changes. Kilo Code supports 5 different modes: 1. Architect mode - Research, design, and planning 2. Debug mode - Problem solving and troubleshooting 3. Code mode - To write code. 4. Ask mode - To talk to the model. 5. Orchestrator mode - Combine every agent to solve problems. On paper, I really like this. I'll need more time to try out the specialized agents and see whether niching becomes an advantage. Here is the link to the website: Thanks to the Kilo for partnering with me on this post.

Santiago

73,062 views • 9 months ago

Every morning I open my laptop and i get served with Alpha in Crypto and Ai.. basically, i got tired of checking a dozen tabs every morning just to find what's actually moving in AI and crypto. so i built my own radar using GPT 5.6 sol ( cc: Sam Altman ) And i have named it Signal Radar.. it sweeps 4 platforms at once.. will add more.. - X - hacker news - github - and reddit. throws out the noise, keeps only AI and crypto, and ranks what's left by what's ACTUALLY pulling.. loudest posts by likes, trending repos by the stars they got today, the reddit threads bubbling in the subs my people actually sit in.. and it doesn't just link them.. it reads them. click any card and the full thing's already pulled in, the whole thread, the whole post. so i'm not opening multiple tabs or social media platforms to catch up.. it did that overnight. this sweep alone: 77 signals in, 49 flagged as actually worth the click.. runs on my own machine. free. no api keys, no subscription, nothing. running on my own setup for now. want it? say so below and i'll make it open source

Every morning I open my laptop and i get served with Alpha in Crypto and Ai.. basically, i got tired of checking a dozen tabs every morning just to find what's actually moving in AI and crypto. so i built my own radar using GPT 5.6 sol ( cc: Sam Altman ) And i have named it Signal Radar.. it sweeps 4 platforms at once.. will add more.. - X - hacker news - github - and reddit. throws out the noise, keeps only AI and crypto, and ranks what's left by what's ACTUALLY pulling.. loudest posts by likes, trending repos by the stars they got today, the reddit threads bubbling in the subs my people actually sit in.. and it doesn't just link them.. it reads them. click any card and the full thing's already pulled in, the whole thread, the whole post. so i'm not opening multiple tabs or social media platforms to catch up.. it did that overnight. this sweep alone: 77 signals in, 49 flagged as actually worth the click.. runs on my own machine. free. no api keys, no subscription, nothing. running on my own setup for now. want it? say so below and i'll make it open source

Axel Bitblaze 🪓

17,126 views • 1 month ago

This is the most insane story of 2025 The number 1 game in the entire world right now was built and shipped by a single person Made $20,000,000+ legit over night TylerTVGS built a Breaking Bad simulator by himself and now it has more players than games 1000+ devs worked on This is the power you have in 2025 Literally ANYONE can do this AI gives the power for ANY single person to come up with an idea, build it, and ship it no matter how little skills you have If you're thinking "oo you gotta be special and talented to do something like that" I promise you're wrong I did the same thing. I built and shipped an AI tool for X by myself and now it's my full time business I truly believe this is the single greatest economic opportunity of the last 1,000 years If you haven't started using Grok yet to plan and build apps, even if you have 0 experience, you're letting down your entire blood line Carve out your calendar, copy the prompt below, and run it today: "I have 0 programming experience at all but would like to build an app. Here is a list of my 5 interests. What are challenges than I can build with software? Choose the easiest challenge to solve then build me a plan for building it even though I've never programmed before in my life." Don't waste this opportunity.

This is the most insane story of 2025 The number 1 game in the entire world right now was built and shipped by a single person Made $20,000,000+ legit over night TylerTVGS built a Breaking Bad simulator by himself and now it has more players than games 1000+ devs worked on This is the power you have in 2025 Literally ANYONE can do this AI gives the power for ANY single person to come up with an idea, build it, and ship it no matter how little skills you have If you're thinking "oo you gotta be special and talented to do something like that" I promise you're wrong I did the same thing. I built and shipped an AI tool for X by myself and now it's my full time business I truly believe this is the single greatest economic opportunity of the last 1,000 years If you haven't started using Grok yet to plan and build apps, even if you have 0 experience, you're letting down your entire blood line Carve out your calendar, copy the prompt below, and run it today: "I have 0 programming experience at all but would like to build an app. Here is a list of my 5 interests. What are challenges than I can build with software? Choose the easiest challenge to solve then build me a plan for building it even though I've never programmed before in my life." Don't waste this opportunity.

Alex Finn

1,430,662 views • 1 year ago