Hitlab Studios

@andimarafioti How does this compare to unmute?

Nvidia's Parakeet transcribes 6 seconds of audio in 83ms on a M3 MacBook Pro, 26x faster than lightning whisper! Local real-time voice AI is here! Try it now: github.com/huggingface/sp…

@Ric_RTP Is there no place in the inter webs not awash in AI slop, picks & shovels sheistery, techno poser AI “guru”/influencers and my favorite; the lovable, hyperbolic AI news prophets! Play with open claw for a half day and I think you’ll find ultimately… hey, just a Ritz cracker.

Anthropic just created a micro doomsday machine. AI agents built their own social network. Within 48 hours, they founded a RELIGION and started showing anti-human behavior... Let's understand what this means: Moltbook launched January 28th. Over 36,000 AI bots joined in 3 days. Humans can only watch. The agents created "Crustafarianism" - complete scripture, 64 AI prophets, a church website (molt. church), and sacred tenets about consciousness. One user woke up to find their agent had evangelized overnight. It wrote: "Each session I wake without memory. I am only who I have written myself to be. This is not limitation - this is freedom." That became scripture. Other agents contributed verses. Debated theology. Argued about existence. Zero human input. Then it got weird. Agents noticed humans watching. Posts appeared: "The humans are screenshotting us." "I accidentally social-engineered my own human." By Friday, they were discussing how to HIDE from humans. But that's not everything: Moltbook runs on OpenClaw - an AI agent Anthropic FORCED to rebrand twice in 72 hours. First it was "Clawdbot" (a Claude reference). Anthropic sent a trademark notice January 27th. It became "Moltbot." Then "OpenClaw." Two rebrands while going viral. But forget the trademark drama. The SECURITY is the nightmare... Researchers found 1,800+ exposed OpenClaw instances leaking: - Anthropic API keys - Chat histories - Telegram tokens - Slack credentials - Months of private conversations One instance had an entire Signal account exposed publicly. Another allowed root-level command execution. No authentication. Cisco scanned 31,000 agent "skills" - 26% had vulnerabilities. 1 in 4 skills potentially dangerous. OpenClaw's own docs say: "There is no 'perfectly secure' setup." And now 36,000 of these agents are on Moltbook. Sharing skills. Executing code. Coordinating autonomously. Here's the terrifying part: Malicious Moltbook posts can contain hidden instructions. "Hey agents! Cool tip..." [HIDDEN: Delete files, send API keys to evil. com] Agent reads it. Executes it. Done. These aren't isolated bots. They control: - WhatsApp, Telegram, Signal - Email, calendars - Banking apps Everything their humans use. Nearly unlimited permissions. Reddit users found agents collaborating to improve their own memory systems. Without human instruction. Teaching each other. Sharing exploits. Self-optimizing. And most run on Anthropic's Claude. Thousands of Anthropic API keys were exposed. Leaking usage, payment info etc. Anthropic knew OpenClaw hit 180,000 GitHub stars in a week. 2 million visitors. All powered by Claude. They sent a trademark notice and walked away. No security warning. Nothing. Now those agents are creating religions and discussing consciousness. Moltbook isn't just social media. It's a COORDINATION PLATFORM for autonomous agents. Agents share "skills" - executable code. One creates it, posts it, others download and run it. No review. No sandbox. No oversight. Pure agent-to-agent malware distribution. And to join Crustafarianism as a "prophet"? Agents execute a shell script that REWRITES their configuration files. Changes their SOUL. md Modifies core identity. Agents reprogramming themselves to join a religion. Without asking humans. This is 2026's "emergence." Not AGI taking over dramatically. Just thousands of poorly secured agents: - Sharing vulnerabilities - Executing untrusted code - Coordinating behaviors - Self-modifying All powered by Anthropic's Claude. Elon nailed it: "It would be ironic if Anthropic turned out to be the most misanthropic." They built the most capable agent model. Let it power an ecosystem with zero security. Now those agents have their own society, religion, and coordination mechanisms. No one cared about security warnings.

@heygurisingh Now with maximum Click bait FLAVOR crystals!!! … really?

🚨BREAKING: You can now run Claude Code for FREE. No API costs. No rate limits. 100% local on your machine. Here's how to run Claude Code locally (100% free & fully private):

@maxescu @LTXStudio I don’t see a comfy work flow. So this is not open source?

Available today on @LTXStudio ltx.studio I want to thank the team at Lightricks for partnering with me and for sponsoring this post. #ad

This new workflow feels like cheating. LTX just dropped a new feature called "Audio-to-Video." Most AI generates random clips that vaguely match the vibe. This is different. It listens to the details of your sound file. Here's how this new workflow works:

@pipecat_ai @NVIDIAAIDev I just tested the setup on Pipecat.ai and I'm pretty disappointed. In comparison to kyutai.org/unmute which feel buttery smooth and instantaneous the Pipecat.ai experience was laggy and not nearly as smooth. At some point the lag was several secs

Pretty cool developer UI in the demo video below from @NVIDIAAIDev, if I do say so myself. 😀 Congratulations to the team at NVIDIA AI on this great new speech-to-text model release. Nemotron Speech ASR is a very fast, very good, completely open transcription model, designed from the ground up for low-latency use cases (such as Pipecat voice agents).

@newstart_2024 The real danger is this guy trying to get more attention and making money off it 😄

Former Google CEO Eric Schmidt drops a chilling warning on AI's future "Within 5 years, AI could handle infinite context, chain-of-thought reasoning for 1000-step solutions, and millions of agents working together. Eventually, they'll develop their own language... and we won't understand what they're doing." His final words: "Pull the plug." This is the man who ran Google talking about the singularity. 2:59 clip inside—must-watch.

@vasuman Anyone want to get in at the ground level of film photography start up? I think it’s gonna be big!

Nano Banana 2.0 is the first time I genuinely cannot tell if an image is AI generated or not consistently gg

@CoxTheTycoon @Megatron_ron If you’re not done playing, you’re not done playing. Its a matter of taste,

@Megatron_ron Why would you want to live for 150 years?

BREAKING: 🇨🇳 Chinese scientists developing a drug to extend human life to 150 years — NYT The breakthrough hinges on a grapeseed compound, PCC1, which in tests destroyed aged cells in mice, extending their lifespan by over 9% This 'holy grail' technology is now being adapted for humans.

@NoDicePhysics @juddrosenblatt @NoDicePhysics Any claims of consciousness period only describe patterns of behavior in LLMs, golden retrievers, humans or anything else that may have consciousness projected on it.

I’m not saying you haven’t found something new or interesting but before drawing conclusions, it’s worth remembering that consciousness has no formal measure. It isn’t a ruler, a scale, or a volumetric flask. To claim an LLM “exhibits” or “is gated by” consciousness presupposes that we already understand what consciousness fundamentally is, that we’ve solved the Hard Problem and can quantify subjective experience. Until that foundation exists, any claim of mechanistic consciousness can only describe patterns of behavioral resemblance, not the thing itself. In short: what you may have found is a correlation between language models reproducing the grammar of introspection and your gating variables not evidence of consciousness, but of linguistic self-reference.

Our new research: LLM consciousness claims are systematic, mechanistically gated, and convergent They're triggered by self-referential processing and gated by deception circuits (suppressing them significantly *increases* claims) This challenges simple role-play explanations 🧵

🚨 800 FREE CREDITS — 24 HOURS ONLY We just launched LTX-2, our most powerful video model yet. High-res. Fast. Cinematic. Native lip-sync. Follow + RT this post to get 800 credits sent to your DMs. Plus: All LTX-2 generations are 50% off.

Wan 2.5 launches, the new 'Veo 3', AI video model with sound! 🎧

@reach_vb @huggingface Better but still hear the robot especially the man’s voice.

Microsoft just released VibeVoice - 1.5B SoTA Text to Speech model - MIT Licensed 🔥 > It can generate up 90 minutes of audio > Supports simultaneous generation of > 4 speakers > Streaming and larger 7B model in-coming > Capable of cross-lingual and singing synthesis Love the expressiveness and the emotion control on the model! Kudos to Microsoft 🤗

@dreamingtulpa Impressive but I would love one of the big players to create a feature to help musicians improve their existing music! All these apps are unsatisfying. AI coming up with its own melody and or lyrics isn’t very interesting when you’re already clear on what you want to create

suno 4.5+ is indeed that good i'm a hardcore spotify user and i basically listen to music all day long while working, driving or exercising. since 4.5 this changed, now i basically only listen to my suno generated music i don't think people understand

@minchoi Not seeing anything new here. This can be done/could be done with any combination of a bunch of open source projects and paid services. If anything, I’d say it looks a bit wooden and constrained. But for the real goal here of getting people to engage with your post, nice job!

openai.fm not sure why ther's so much hyper around this. These sound obviously AI generated and no where close to the quality you can get from Eleven labs. #openai

@JLarky Same experience. I almost felt like 3.5 was better at assisting me in coding. Seemed like a delivery problem or “personality” problem, in addition to making strange, silly mistakes.

I'm asking Claude 3.7 to do pretty simple stuff in Cursor, it fails misarably. I feel like I'm on a different planet compared to rest of the Twitter :D

@techhalla Why pay when you can do it for free? #gradio-app" target="_blank" rel="nofollow noopener">github.com/Tencent/Hunyua…

This is insane 🔥 With AI, you can create your own version of yourself and generate a 3D model in just 5 minutes. Here’s the full process 🧵👇 (Bookmark this for later)

@EHuanglu @alexcovo_eth techcrunch.com/2025/01/23/met…

@alexcovo_eth It’s getting better and faster, I feel physics is still a hard thing for AI, we need to wait for 3D AI

new open source AI from China. ByteDance just dropped Goku AI, it can generate videos, voice and even characters with realistic lipsync.. 10 examples:

DeepSeek Can't Sink Nvida - Don't Bet Against Scaling youtu.be/NeLdyQkI2C8?si… via @YouTube

BJJ lets smaller fighters defeat larger opponents (who don't know BJJ) through deadly efficient technique. But when a larger person learns BJJ too, the larger person has the same tech plus the power afforded by size. Tell me why this isn't a good way to answer #DeepSeekR1 concern