Insane AI Breakthroughs Nobody Is Talking About
Insane AI Breakthroughs Nobody Is Talking About
11 hours ago•Matt Wolfe•@mreflow
YouTube29 min 25 sec
Watch on YouTube
Note: AI-generated summary based on third-party content. Not financial advice. Read more.
Quick Insights
  • There is no high-conviction, stock-specific trade in these insights: the public-company examples are research-stage or lack evidence of commercial adoption, revenue impact, or a clear catalyst.
  • Monitor NVIDIA (NVDA) for progress in AI robotics, but treat Motion Bricks as a long-term theme rather than a near-term buy signal.
  • Avoid treating Block (XYZ), Tencent (0700.HK), or Adobe (ADBE)’s research demos as investment catalysts until products, adoption, and financial impact are confirmed.
  • Treat SubVocal’s glasses pre-order cautiously: the claims are unverified, and estimated shipping is not until Q1 2027.
Detailed Analysis

NVIDIA

  • NVIDIA’s research team introduced Motion Bricks, a model that generates character movement in real time and can also drive humanoid robots through its Groot system.
  • The model and code are available on GitHub, but the transcript says it is not yet available for direct use in Unreal Engine.
  • Sentiment: Positive about NVIDIA’s progress in AI-driven animation and robotics, but the discussion focuses on research and demos rather than commercial results.

Takeaways

  • NVIDIA’s work points to a potential growth theme in AI-powered robotics and animation tools.
  • The transcript does not provide revenue, adoption, valuation, or timing information to support a stock-specific recommendation.

Block

  • Block’s research lab developed Dessa, a voice model designed to carry on natural, overlapping conversations without converting speech to text first.
  • The model sounded highly human in the examples, but the transcript describes it as a research preview available only to a limited group of AI safety researchers.
  • The speaker highlighted the potential for scam calls as a reason to be cautious about wider release.

Takeaways

  • Natural-sounding voice AI could create new product opportunities, but the transcript does not establish how Block might commercialize Dessa or whether it will become a meaningful business.
  • The stated scam risk is a notable concern for the technology’s broader deployment.

Tencent

  • Tencent released an open world-generation model that can continue a scene as users direct it, while attempting to preserve consistency across camera views and changes.
  • The speaker said the model is free to use, including commercially, but described its output quality as not yet suitable for final production work.
  • The real-time setup reportedly used eight data-center GPUs; a slower, compressed version can run on one RTX 4090. Tencent had not published comparisons with other world models.

Takeaways

  • The release reflects the broader AI theme of interactive world models and lower-barrier access to generative tools.
  • The transcript offers limited evidence of commercial readiness: quality limitations and the absence of comparative performance data make the model’s practical value uncertain.

Adobe

  • Adobe and a university in Korea developed Movie Edit, a research model that can apply an edit—such as removing a person—across multiple shots in a clip.
  • The transcript says the project is currently a research paper and example videos, with no downloadable model or confirmation of integration into Premiere.

Takeaways

  • Consistent AI editing across scenes could be valuable for creative-software users if it becomes a reliable product feature.
  • There is no confirmed product release or commercial impact in the discussion, so this remains a technology to monitor rather than a demonstrated investment catalyst.

AI Voice and Customer-Service Software

  • Fish Audio’s Live 3 is a paid voice-AI product aimed at business use, including support calls and restaurant ordering. The transcript cites a limited-time price of about $0.035 per minute, set to double at year-end, and a $5 trial credit.
  • Haru 4B, from Japanese startup Higurashi, is an open model available on Hugging Face and GitHub. The speaker described it as a developer starting point, not a model expected to match ChatGPT- or Claude-level answers.
  • The transcript also describes Dessa as a research preview rather than a generally available product.
  • Sentiment: Encouraging about the pace of development, but the products differ in availability, capability, and business model.

Takeaways

  • Business voice automation is an emerging AI software theme, with Fish Audio’s per-minute pricing providing an example of a usage-based model.
  • The transcript does not identify a publicly traded security tied to Fish Audio or Higurashi, and it provides no customer adoption or revenue evidence.
  • The scam potential of highly realistic voice models, specifically raised in the discussion of Dessa, is a risk to consider for the sector.

AI Hardware and Robotics

  • The discussion highlighted growing links between AI models and physical systems: NVIDIA’s Motion Bricks research, Stanford’s Flash Aloha robot, and UCLA’s PhoneBot.
  • Flash Aloha reportedly learned tasks such as catching a ball from demonstrations, with a success rate above 80%. The speaker noted that this still means it could fail roughly one in five attempts.
  • PhoneBot uses an Android phone as its “brain” and about $400 in other parts, but its printable parts and assembly tutorial were not yet available at the time of recording. It was described as limited to flat ground and as having no arms.
  • These examples were presented mainly as research or DIY projects, not as commercially available robotics products.

Takeaways

  • AI-enabled robotics is a long-term investment theme, but the examples in the transcript remain early-stage and do not establish broad commercial deployment.
  • The stated task failures and practical limitations underline the gap between impressive demonstrations and dependable real-world use.

Early-Stage AI Products and Tools

  • Simulon offers an augmented-reality video app that adds responsive 3D characters to recorded footage. The speaker said it has a free starter allowance and then costs $15 per month; its interactive feature was iPhone-only at the time, with Android support planned.
  • SubVocal is developing glasses that claim to detect silently articulated words using radio sensing. The transcript cites a $199 refundable pre-order deposit toward a $499 price and estimated shipping in Q1 2027.
  • The speaker urged caution on SubVocal: its claims had not been independently tested, and published research on similar sensing was described as reaching a much smaller vocabulary than the company’s claim.
  • Other projects discussed, including Reshoot 4D and Budget 3D, were free or open-source tools rather than investment offerings.

Takeaways

  • These products illustrate potential consumer and creator markets for AI and augmented reality, but the transcript provides no evidence of public-market investment opportunities or established business performance.
  • Treat product demonstrations and company claims as preliminary where independent testing, availability, or commercial adoption is still limited.
Ask about this postAnswers are grounded in this post's content.
Video Description
12 AI breakthroughs you probably missed, from incredibly human AI voices to a walking robot powered by an old phone. These are some of the coolest AI tools, research projects, and tech demos I’ve come across recently that didn’t make it into my weekly AI news videos. We’re looking at AI characters that react to your movements, interactive world models, new ways to edit videos, and voice AI that invents an entire life for itself. I test several of these tools, explore what you can actually try today, and point out where the demos still fall short. Plus, glasses that claim to understand silent speech, NVIDIA’s AI character animation, and robots learning to move at human speed. Which of these would you want to try first? Let me know in the comments. Sharing the latest news and advancements in AI Discover More: 🛠️ Explore AI Tools & News: https://futuretools.io/ 📰 Weekly Newsletter: https://futuretools.io/newsletter Socials: ❌ Twiter/X: https://x.com/mreflow 🖼️ Instagram: https://instagram.com/mr.eflow 🧵 Threads: https://www.threads.net/@mr.eflow 🟦 LinkedIn: https://www.linkedin.com/in/matt-wolfe-30841712/ 👍 Facebook: https://www.facebook.com/mattrwolfe Resources From Today's Video: Simulon: https://simulon.com/ Simulon Launch Post: https://x.com/Simulon/status/2115519228564693170 Hy-Nomas: https://hy-nomas.hy.tencent.com/ Reshoot 4D: https://itsmetamike.github.io/reshoot-4d/ MoviEdit: https://multi-shot-edit.github.io/ Dessa: https://block.xyz/dessa Dessa Demo: https://x.com/RagavanThurair/status/2113671324400861645 Haru 4B: https://www.higurashi.ai/haru4b/ Fish Audio Live 3: https://fish.audio/live3/ Subvocal: https://subvocal.company/ Budget3D: https://budgetany.github.io/budget3d/ MotionBricks: https://nvlabs.github.io/motionbricks/ Flash ALOHA: https://johnyao93.github.io/flash-aloha/ PhoneBot: https://phonebot.dev Let’s work together! - Brand, sponsorship & business inquiries: business@mattwolfe.com #AITools #ArtificialIntelligence #Robotics Time Stamps: 00:00 AI Breakthroughs You Probably Missed 00:51 Simulan: AI Characters in Your Room 04:27 Tencent’s Hynomus World Model 08:13 Reshoot 4D: Change Your Video’s Camera Angle 10:26 Movie Edit: AI Edits Across Multiple Shots 11:26 Desa: AI Voice That Sounds Human 14:58 Haru 4B: Downloadable Conversational AI 16:56 Fish Audio Live 3: AI Voice Agents 18:48 Subvocal: Glasses That Decode Silent Speech 20:54 Budget3D: Turn Images Into 3D Models 23:15 NVIDIA MotionBricks: AI Character Animation 24:59 Flash Aloha: Robots That Catch and Clean 26:29 PhoneBot: A Walking Robot With a Phone Brain
About Matt Wolfe
Matt Wolfe

Matt Wolfe

By @mreflow

AI News Breakdowns every Saturday and other cool nerdy tech and AI stuff in between. Let's work together! - For brand ...