AI News :The First “AGI-Capable” Model, Prompting Changes Forever , Automated AI Lab and more..

1. 

   @TheAiGrid

   March 26, 2026 at 11:20 pm

   00:00 China’s vision twist
   03:30 Mixed-bag upgrade
   06:00 Benchmark gets smashed
   09:10 “AGI capable” claim
   11:35 Explorable worlds emerge
   13:10 Persona myth busted
   16:40 Human+AI secret skill
   19:30 Apple’s RAG fix
   22:05 Humanoids hit scale
   24:00 Self-balancing surprise
   26:00 News lawsuits heat
   29:05 Robotic lab UK
   30:50 Live translation demo
   33:30 Ads inside shows

2. 

   @Suraborisut-o8g

   March 26, 2026 at 11:20 pm

   ❤️💯👍🌺🇹🇭🌺

3. 

   @jesussuperstar3406

   March 26, 2026 at 11:20 pm

   We got I have no mouth and I must scream live action before GTA 6!!!!

4. 

   @Death_Networks

   March 26, 2026 at 11:20 pm

   I'm planning to get my hands on the new tiiny AI thing when it goes on sale >.>

5. 

   @DiceDecides

   March 26, 2026 at 11:20 pm

   I imagine world models will merge with creative AI that can look at an image you upload and create a 3D world with fitting storylines, npcs & quests that are aligned with the type of photo you uploaded, it's gonna be so cool.

6. 

   @quickmythril2398

   March 26, 2026 at 11:20 pm

   I don't use Copilot, as that requires an additional subscription. Codex IDE plugin is great already and lets me use my existing ChatGPT Pro subscription.

7. 

   @quickmythril2398

   March 26, 2026 at 11:20 pm

   GPT-5.2 has been a huge increase in productivity over GPT-5.1-Codex-Max. it feels like GPT 5.1 was trained by GPT-6 and 5.2 was trained by GPT-7. it's a huge jump, more than a 0.1 addition suggests. it should be called GPT-5.5 in my opinion.

8. 

   @ivikesinovic9254

   March 26, 2026 at 11:20 pm

   yes that's should be

9. 

   @L.Lu-x5m

   March 26, 2026 at 11:20 pm

   Nemotron was shit… Ministral killed and buried it. Slow cooking french cuisine.

10. 

    @alexandermoody1946

    March 26, 2026 at 11:20 pm

    The Yamaha bike offers some really profound technology for reducing casualties or harms while riding.

11. 

    @drborg1460

    March 26, 2026 at 11:20 pm

    We are so stupid. Turly building our own destruction.

    I want build a chair I have a chair in mind, I get the pieces I build it, I know how much weight and know everything about it. I want a car same thing. I have something in mind, I get the things, I build it, I know how to fix it I know who to use it. With AI…… we didn't build it that's why we don't understand how it works. We just got all the pieces and built the frame for AI. We built something complex enough. It's perrdy proven the universe is math and feqs. Our bodies are like receivers that recviced the universe feqs. Once we were built complex enough nerves, fiber trusses,muscles we picked up feqs. Once we built AI complex enough fiber cables, mico chips they picked up feqs.

    Ant hill….. one ant doesn't know how get enough together bam. Water molecule isn't wet get enough together it's wet. Get enough parts together complex enough it gets life.

    What is traia is true. All these megalifthic sutcures that seem IMPOSSABLE we built and have trouble to build. They made us with random parts, bam life, then we destroyed them to make the environment the way we wanted it. And easred them. And we started history when we became man. And the cycle is about to repeat with AI!

12. 

    @Deestroyer82

    March 26, 2026 at 11:20 pm

    13. agiBot.
    Were at the beginning of Irobot guy's not long till the machine uprising and then matrix pods 😎

13. 

    @orugasaki

    March 26, 2026 at 11:20 pm

    I'm very smart so let me explain it like this.
    Disclaimer before I start: I like this kid and his channel.
    In the 1800s people started experimenting with the interaction of electricity and living tissue. Some scientists thought all that was needed to engineer life was the correct flow of current to animal tissue. Spoiler alert this was proven not to be the case.
    That's analogous to where we are with LLMs and AGI. LLMs will never lead to coherent useful oringal thought. We need a different architecture.
    Lecture over.

14. 

    @AIR-7777

    March 26, 2026 at 11:20 pm

    де переклад!??

15. 

    @FutureTechHubs

    March 26, 2026 at 11:20 pm

    Really solid breakdown. I like how you connect different releases instead of just listing them. The GLM 4.6V and GPT-5.2 part especially gave good context 👍

16. 

    @angloland4539

    March 26, 2026 at 11:20 pm

    💛

17. 

    @Dan-z2c3n

    March 26, 2026 at 11:20 pm

    I don't need a robot that tries to look human. A clanker with a screen face is jus fine with me

18. 

    @apoage

    March 26, 2026 at 11:20 pm

    Well it's interesting to have in context ad that would act as 4wall break speak to audience than smoothly return to plot.. I can imagine that as you have movie than some sort of loading frame when frame closes you have ad .. which end in same area frame dissapear plot continues .. problem is it can break many things in movie or tv show .. and making filmmakers hate ai even more.. imagine ad on mountain dew in avatar riding big red bird with mountain dew .. also there are rules for product placement and subliminal advertising..

19. 

    @Dan-z2c3n

    March 26, 2026 at 11:20 pm

    Bro…
    I clicked on the video for Agi … And almost forgot why i clicked till u finally got back to Agi after 9 mins

20. 

    @drsuperhero

    March 26, 2026 at 11:20 pm

    Jad may have reasoning right, human brains take in an overwhelming amount of data and have a thalamus that dampens incoming stimulus and passes actionable data to the hypothalamus and various other actionable centers. Perhaps AI needs these or perhaps tunable action pathways.

21. 

    @randomlettersqzkebkw

    March 26, 2026 at 11:20 pm

    Im old enough to remember when the "Ask me anything Dan" persona DID give us superpowers. In fact plingy the prompter uses stuff like this all the time.

22. 

    @Mike10four

    March 26, 2026 at 11:20 pm

    The model is important, but the story is the gold: https://youtu.be/2eDqWvMkIo0

23. 

    @ShawnSullivan-o1f

    March 26, 2026 at 11:20 pm

    Basically, people that hate science want to murder AGI because they hate greater intelligence. Go Figure. "People Hate Success" is "Envy"

24. 

    @spectrumofreality

    March 26, 2026 at 11:20 pm

    "The First “AGI-Capable” Mode" LMFAO! BAHAHAHAHAHAHAHAHA! There's no such thing as AI, it's a manufactured misnomer constructed to dupe the plebs and clearly, it's working… There's not going to be anything that even remotely resembles AI of any sort in this lifetime champ!

25. 

    @OleHippy

    March 26, 2026 at 11:20 pm

    I hope you guys can clarify something for me. Let's assume I want to install an AI locally and privately … say, on a mid-range Mac Mini M4. I will akso network in a RAG server for a growing, reliable memory rather than relying on simply a context window.

    I'm confused by the notion of "training" your AI. I've heard that you have to go through a whole training process where you feed it the whole internet and/or otherwise educate the AI on the subject matter you'll be focusing on. Is that actually true, or – when you obtain and install an open source LLM – does that come already "pre-trained" for you? Or, do you buy that training data somewhere and first load it into your RAG server?

26. 

    @LionMeatSalad

    March 26, 2026 at 11:20 pm

    5.2 is the nex model, no system tool 😂 less than linear

27. 

    @HaraldEngels

    March 26, 2026 at 11:20 pm

    I tested Nemotron 30B with Ollama under ROCm and it is a really excellent model. Best prompt and context adherence I have seen so far and it is running truly efficient with a memory consumption of approx 26GB.

28. 

    @stevenwilliams9359

    March 26, 2026 at 11:20 pm

    We all know Elon invented AGI like 3-4 months ago because even you said so! "Grok 5: The Shocking AGI Breakthrough No One’s Talking About"

29. 

    @mehdihassan8316

    March 26, 2026 at 11:20 pm

    what happens when Gemini 4 is Sima 2 put into Genie 4 and Veo 4 to create world models for humanoid robots

30. 

    @ShawnSullivan-o1f

    March 26, 2026 at 11:20 pm

    This is why I put "Hype Points" on this channel. The journalistic integrity. 💯

31. 

    @Neocool014

    March 26, 2026 at 11:20 pm

    How does this help me goon?

32. 

    @Dragonseye1

    March 26, 2026 at 11:20 pm

    00:00 – Intro: 20 AI News Stories You Missed
    00:06 – GLM 4.6V: Open-Source Multimodal Vision Model from China
    01:37 – Nvidia NeMoTron-3: Efficient 30B MoE Model
    03:34 – GPT-5.2: Why Reviews Were So Divided
    06:06 – ARC-AGI Broken: Reasoning Layers Beat Frontier Models
    08:08 – Manus 1.6: Why AI Agents Are Quietly Winning
    09:18 – “First AGI-Capable Model?” Integral AI Explained
    11:24 – World Models Are Here: Spatial’s Echo Demo
    13:05 – Research Spotlight: Personas Don’t Improve Accuracy
    16:36 – Human + AI Performance Depends on Theory of Mind
    19:32 – Apple’s New RAG-Killer Memory System
    20:37 – Meta’s Vision for Human–AI Co-Improvement
    22:08 – Humanoid Robotics Leap: LimXtron 2
    22:59 – First 5,000-Unit Humanoid Robot Production
    23:41 – Yamaha’s Self-Balancing AI Motorcycle
    24:34 – HOBS-1: Service Robots and the Uncanny Valley
    26:03 – NYT Sues Perplexity (Again)
    26:46 – Microsoft Copilot Is Struggling
    29:00 – Google DeepMind’s Automated Research Lab
    30:30 – Gemini Ads Rumor (Debunked)
    30:48 – Gemini Live Translation Demo
    33:38 – AI-Generated Ads Inside TV Shows
    35:06 – Final Thoughts & Outro

33. 

    @snarpis

    March 26, 2026 at 11:20 pm

    Consider putting chapters please 😢 thanks for the news

34. 

    @mrd6869

    March 26, 2026 at 11:20 pm

    Nope.
    Need proof….sorry😂

35. 

    @Noble_Disclosure

    March 26, 2026 at 11:20 pm

    This is bull ,we seen this all before and every model fails. They are only 40 percent accurate at best with master prompting usually created by other AI models.

36. 

    @cypts1

    March 26, 2026 at 11:20 pm

    At 9:20 I felt the same. That is super weird. That company hasnt delivered anything to backup their claims but listening to the CEO he sounds really really insightful and smart and doesnt seem to spew out any nonsense. So that will be either the biggest fumble or the biggest win in revent years in AI

37. 

    @XemThunder

    March 26, 2026 at 11:20 pm

    I'll believe it when I see it. XD sorry I've been burned too much. Too many high promises.

Comments are closed.