Your curated collection of saved posts and media
@lukas_m_ziegler The prices of everything in technology always start real high and come down real quick. I remember when a color printer was $45,000 in 1989, and today a $79 printer is way, way better. But how does https://t.co/JuDkLaml2o sell Neo for $20,000?
Introducing β¨VideoRAGβ¨ - Extremely Long Video Understanding. Chat with your videos like never before. Accepted by KDD 2026! π π Fully Open Source: https://t.co/QrNajpOby4 Current AI struggles with long videos. Most systems can't handle hours of content or understand connections across multiple videos. β¨ VideoRAG breaks these barriers! - Easy-to-Use: Just drag & drop your videos and start chatting - No Length Limits: Handle everything from 30-second clips to 100+ hour documentaries - Cross-Modal Understanding: Understands visual content, audio, and context together --------------------------------------------------------- For Developers & Researchers: - π Process hundreds of hours on a single RTX 3090 - π§© Multi-modal knowledge graphs for structured understanding - π LongerVideos benchmark with 134+ hours across lectures, docs, and entertainment - π Open source and extensible

@din0s_ @typelessdotcom Here is @blevlabs talking to his agents. And he is a genius. https://t.co/pyxvrissKW
Reachy Mini conversation app γγγγγδΈ https://t.co/fRb4jlANIH
Reachy Mini conversation app γγγγγδΈ https://t.co/fRb4jlANIH
The .0003 kWh measure for a standard prompt. is very consistent. For an independent assessment, here is ML Energy Leaderboard, which uses a collection of 500 human prompts for testing, and is not particularly optimized for energy consumption, again finding right around .0003 kWh. https://t.co/BHzkTX7PfZ
When AI learns βwhy,β it creates better content https://t.co/TnAkei1Q8x @FuturityNews
Before you ask AI another dumb coding question⦠watch this. https://t.co/QDoviX0grP
Before you ask AI another dumb coding question⦠watch this. https://t.co/QDoviX0grP
π¨The AI agent handbook Google just dropped a 46-page playbook on how to build and use agents. This is what you need to know (and how to get it 100% free): https://t.co/MDYOBIGy9d
Maybe I should really print these shirts for @ManusAI Used our new design mode to edit them and Iβm in love at how ugly they are π€£π€£ https://t.co/MnIZrwiRUt

π§ββοΈγ―γͺγΉγγΉγ€γγ«ιζ³ε°ε₯³γ«γͺγε€’γ Manusγ§εΆγγ‘γγγΎγγοΌwwβ¨ ιζ³γ§ζ΅γεγγ΅γ€γγγ Manusγγγγ°γ©γγ³γ°γε ¨θͺεγ§ζγγ¦γγγγ γγͺγγιζ³ε°ε₯³γ«γͺγ£γ¦ζ¦γγοΌw (γγ²δ½ι¨γγ¦γΏγ¦)π https://t.co/YXqWE3abqr https://t.co/58Ain8f3VL
Apple: We are working on our first foldable screen Huawei: https://t.co/gXitJzHXyu
A man made a swimming harness for his fish that cannot swim due to its illness. https://t.co/SQJd2fVT1T
In fact the average ChatGPT query takes almost exactly as much energy as a Google search in 2008 (that is the last time Google clearly indicated the electrical consumption of a search). https://t.co/SHAa8hthNT
this tawan for this newwiee https://t.co/jVmk3XWsC1

AQUΓ POR ANILPIN, Y TΓ? ANILPIN WE ARE THE ONE #ΰΈΰΈ΄ΰΉΰΈΰΈ ΰΈ±ΰΈΰΈΰΈ΄ΰΉEP6 https://t.co/0AMQEkEUBh
Scoop: Mattel's collaboration with OpenAI won't emerge this year https://t.co/kDFXKjcZMb @axios @inafried #GenAI Cc @mclynd @BrettKing @UrsBolt @helene_wpli @SpirosMargaris https://t.co/TnX2ozrWn3
2026 predictions: AI agents take over finance β’ 75% firms invest in autonomous AI for global ops β’ S&P 500 to 7,800β17% EPS boom from AI efficiency β’ Stablecoins go mainstream amid AI infra surge β’ China 5% GDP pivot fuels Asia-tech overcapacity Great chart below will save you hours of reading β #fintech #tech #finserv #AI @BetaMoroney @efipm @BrettKing @spirosmargaris @jasuja @enricomolinari @mikeflache https://t.co/Zc1qcaYCNY

Larry Ellison's (Net Worth: $300B) No.1 advice to Entrepreneurs https://t.co/rhBanWUJMW
Larry Ellison's (Net Worth: $300B) No.1 advice to Entrepreneurs https://t.co/rhBanWUJMW
No, you canβt tell when something was written by AI https://t.co/eAJDDfqKY5 @ElaineDMoore @ft
Tech groups shift $120bn of AI data centre debt off balance sheets https://t.co/KS55NWIRxL @Tabby_Kinder @ft
Why AI Agents Need Their Own Identity, Not Yours: https://t.co/BDLlBxPmcd @thenewstack
ELON MUSK: βThe fundamental moral flaw of the left is empathy for the criminals and not empathy for the victims. There's been immense judicial overreach that is unconstitutional. It's undermining the people's faith in the legal system. It needs to stop. It's gone too far." https://t.co/Pb6fFWuot1
This man had 8 arrests in just one jail in 2025, 4 in 2023, 1 in 2020, a misdemeanor conviction in 2024, and a felony assault conviction in 2012 All for assault. He refuses to stop hurting people But if we locked him and other 10+ criminals away, violence would decline ~20% https://t.co/Emf4WvJONb
After βThe Bail Projectβ NGO posted bail for repeat offendersβ¦ Donnie Allen killed Benjamin McComas. Travis Lang killed Dylan McGinnis. Marcus Garcia killed Christie Holt. Samuel Lee Scott killed Marcia Johnson. Itβs not just Judges. NGOs are involved too. How many more? https://t.co/Y0T5X7gvjG
The violent man who slashed an elderly woman across the face with a nail and destroyed her eye after years of being allowed to punch random women in the face without consequence identifies as a woman, according to his defense attorney. https://t.co/zpgHr6nvsN
10 hours into Tesla FSD v14.2.2 and here are my thoughts: Zero lane change hesitation at all. Extremely refined, excellent lane choices and some parking improvements are a few things to note. Really stoked about the lane changes on this build. There is ZERO lane indecisiveness, no hesitation, no stuttering. Nothing. I had never really seen much before, but this build you can tell its decision making process is much quicker and it commits. Mad Max mode is definitely mad again and I love it. Seems even quicker than v14.2.1.25 which was an improvement, but the lane changes + choices are better now. Hurry mode seems quicker on the highways and larger roads than before and very close to Mad Max, it was perfect before on v14.2.1.25. The new addition on the release notes for arrival pin being adjusted based on arrival option works. It changes the route based on if youβre going curbside, or to the lot or street parking. It behaved that way before usually but now itβs visualized. Parking is improved and impressive this build. Every single spot I parked in, it performed excellent. Centered in the lines, quick one and done parking jobs and they were great, improvement over early v14.2 builds. In parking garages, it now pulls up exactly at the ticket dispenser where in earlier v14.2 builds it would stop 9 inches short like I mentioned in previous v14.2+ build reviews. Thatβs fixed and worked great- couldnβt have asked for it to be any better. Perfect distance and approach- will do a lot more of this testing tomorrow. Canyon roads were insanely impressive. Did the snake 3 times which is a super narrow and twisty road and each time it performed great. Speeds were good, took the proper lines and even accelerated out of corners very well. These were all in the dark too very late at night- will post some more videos of it. Stop and go traffic performance in rush hour was awesome as before, with some added notes for how well it waited to enter an intersection if there was stopped traffic on the other side. It waited 3/3 lights I encountered where if it went and the light changed, it would be sticking into traffic so it didnβt go. Exactly like a human wouldnβt but the car judged it better than I did. One time it creeped a bit slow, but other than that fine since. Saw one instance of red hands but before I could react, it had already recovered itself. Still so impressed with how well it handles lane splitting motorcycles. It sees them approaching before I even do, and moves over just enough to be courteous to the rider as well as the car or shoulder next to me. Nice touch, and glad to see it perform well. One thing I noticed is how smooth its steering inputs always are. It just takes corners with such precision and smoothness that no human could ever come close to it. Especially noticeable on twisty roads- it drives better than almost every human on them. And I encountered a power outage tonight and the car still drove perfectly even with pitch black around. Overall, loving what Iβm seeing with this build. Will report back in the day time with more but I am going to head home to get some sleep. Another huge thank you to all the magicians @Tesla_AI for working so hard during the holiday season to get these updates out.
https://t.co/K0nlZAJyVK
Grok Rankings Update β Dec 23 π₯ #1 Overall on OpenRouter Leaderboard π₯ #1 in Programming Market Share π₯ #1 in Python-Specific Tasks π₯ #1 in Categories Token Share π₯ #1 on Kilo Code Leaderboard π₯ #1 on BLACKBOXAI Leaderboard π₯ #1 on Roo Code Leaderboard π₯ #1 on Cline Leaderboard π₯ #1 on EQ-Bench3