Your curated collection of saved posts and media
Yes I was blessed to #ThisOurLand #Native https://t.co/NS5rd9LDKQ
Did y'all ever get to meet y'all Great Grandma? Seriously tho....I wanna know
BREAKING: Tesla Model Y Robotaxi with literally no one in the car spotted driving itself on public roads in Austin, Texas! A new age has begun ๐ค https://t.co/BtqN7QfDe7
@Peaceke81017283 @bananacafe1212 @4mYeeFHhA6H1OnF @SSSHeart89537 @mishafumisha @SatoshiMasutani @Armelle_Kiyo @CdfbDtYAxmbMuRZ ๅฆใชใใจใซๆฐใฅใใใใงใใ๐คใใฎๆนๅธฐๅใใใๆฅๆฌๅจไฝๆญด้ทใ่จญๅฎใงใใใญ๐คๆผขๅญใใใไฝ่ฃใงๆใฆใใฏใใ ใๆๆณใ็ฅใฃใฆใพใใใญ๐คใใจใๅคๅฝไบบใฏใ ใใใๆฏๅฝ่ชใฎใญใผใใผใไฝฟใใใใใใใใฒใใใชๆ็จฟใฎๆนใใใใฉใใใใฏใ๐คๆผขๅญๅคๆใใใใฏ่ฑ่ชใง"Arigato"ใใฎใพใพใซใชใใฏใ๐คใตใผใ๐ค https://t.co/jWl8kAvjZE
Introducing the first NATIVE Now! Cover Star: @davido. 1 PM โฐ https://t.co/W4ed9qls8t
Go behind the scenes at our #NATIVENow Cover Shoot with @davido โจ https://t.co/KH0jnjAnSO
yk this is exactly why locals dont like us https://t.co/AaebxFrb2X
Microsoft AI boss Suleyman opens up about his peers and calls Elon Musk a 'bulldozer' with 'superhuman capabilities to bend reality to his will' https://t.co/Xp3EK7gOGQ @microsoft @mustafasuleyman @MicrosoftAI @fortunemagazine
NEW research from Google on effective agent scaling. More tool calls don't always mean better agents. The default approach to scaling tool-augmented agents today remains throwing more resources at the problem such as more search queries, API calls, and more budget. But agents lack budget awareness and quickly hit a performance ceiling. This new research introduces BATS (Budget Aware Test-time Scaling), a framework that makes agents explicitly aware of their resource constraints and dynamically adapts planning and verification strategies based on remaining budget. Standard agents don't know how much budget they have left. Without explicit signals, they perform shallow searches and fail to utilize additional resources even when available. Simply granting more tool calls doesn't help because agents terminate early, believing they've found sufficient answers or concluding they're stuck. Budget Tracker is a lightweight plug-in that surfaces real-time budget states inside the agent's reasoning loop. At each step, the agent sees exactly how many tool calls remain and adapts accordingly. Results: Budget Tracker achieves comparable accuracy to ReAct with 10x less budget (10 vs 100 tool calls), using 40.4% fewer search calls, 21.4% fewer browse calls, and reducing overall cost by 31.3%. BATS goes further by making budget awareness shape the entire orchestration. A planning module adjusts exploration breadth and verification depth based on remaining resources. A self-verification module decides whether to dig deeper on a promising lead or pivot to alternative paths. On BrowseComp, BATS with Gemini-2.5-Pro achieves 24.6% accuracy versus 12.6% for ReAct under identical 100-tool budgets. On BrowseComp-ZH, 46.0% versus 31.5%. On HLE-Search, 27.0% versus 20.5%. All without any task-specific training. Budget-aware design produces more favorable scaling curves and pushes the cost-performance Pareto frontier, achieving higher performance while using fewer resources. It's all about wise-spending. Paper: https://t.co/hqZQoXDnX7 Learn to build effective AI Agents in our academy: https://t.co/JBU5beIoD0
In April this year, we acquired a small (but mighty) robotics company from Bordeaux, France. This weekend, thanks to them, we're shipping 3,000 Reachy Minis all over the world (half of them in the US). That's the power of international collaboration and open-source robotics (and hard work haha)!
https://t.co/28N6CMC1wN
https://t.co/28N6CMC1wN
The public has a distorted view of science, because children are taught in school that science is a collection of firmly established truths. In fact, science is not a collection of truths. It is a continuing exploration of mysteries. - Freeman Dyson https://t.co/cLDtcaY68u
The more I work on agents, the more I realize this uncomfortable truth You can't ship a production grade agent without first understanding production grade code in that exact domain Agents don't replace depth, They punish the lack of it Every extra line of AI code quietly multiplies the slop until momentum collapses
It is time to say goodbye to Egypt. Unfortunately I caught a mild cold and had to spend some time resting in bed. Thank you to the organizers and attendees for the warm hospitality and passion. Here are a few photos from my trip. #GoogleDeveloperExpert #AI https://t.co/dKSNguzPje
ใ#ๆป่ณใฌใคใฏในmathhisใใใคใผใ ใ7 ๆธธ ่พๅใPGใใฏ่ช็ถ๏ผ ใ0 ใถใใฏ๏ผZ๏ผ C/PFใใจใ4 ใฉใคใขใณ PFใใฏใใฃใใงใณในใงใ่ฒข็ฎใใฆใใพใ๏ผ index https://t.co/pkEmDsgTj8 https://t.co/YGynFbWjIc
Most won't do this right Link - https://t.co/twLWInZasQ #maths #iteachmath
@_BarnyO Mathematics been answering this question! https://t.co/N18u1KYjJ5
@RachelReevesMP The mathematical genius of rachel reeves. https://t.co/NWC20M5rzY
๐ค New Position Paper !!๐ค๐๐ค @j_foerst and I wrote a position piece on what we think is the path to safer superintelligence: co-improvement. Everyone is focused on self-improving AI, but (1) we don't know how to do it yet, and (2) it might be misaligned with humans. Co-improvement: instead, build AI that collaborates *with us* to solve AI faster, and to help fix the alignment problem together. More details in the paper! Read it here: ๐:https://t.co/peiPnLHHXG
@SpaceX @Starlink Congrats SpaceX! This marks 120 Starlink missions with 3,095 satellites deployed so far for the year โ an incredible 1,133 more than full-year 2024! ๐๐ฐ https://t.co/myjix5shGj
GROK TEXT-TO-VIDEO โ WORDS IN, CINEMATIC MAGIC OUT Grok Imagine upgrades are magical! No base image. No warm ups. Just pure text straight into video. Quick Hit: โขโ โ Type a prompt, Grok spits out video like itโs directing a trailer โขโ โ Explosion โ rocket launch? Cosmic portal? It handles the wild stuff โขโ โ 10-second clips will roll out like magic soon โขโ โ Rivals like Runway, Pika, and Sora are holding on to their hats xAI went from chatbot to images to full video in under a year. Imagine the progress in year two. Actually, donโt - Grok will generate it! Source: @elonmusk, @xAI
Cherry blossom reflections, Kyoto Japan. https://t.co/SjFkdWaCIL
China is *all* of the growth in global electricity per person.
China's electricity is now 30% more than Europe's โ PER CAPITA. More than double in total.
THE GRID WILL BREAK FIRST Texas just received requests for 226 gigawatts of data center power. The entire state runs on 85 gigawatts. Here is what nobody is telling you. ERCOT's interconnection queue nearly quadrupled in twelve months. From 63 gigawatts in December 2024 to 226 gigawatts by November 2025. Seventy-three percent is data centers. The AI infrastructure buildout is now the largest in American history. But here is the number that changes everything: 3%. Of the 226 gigawatts requested, only 7.5 gigawatts are actually connected and operating. Fifty-seven percent of the queue has not even submitted planning studies. Joshua Rhodes at UT Austin called it "laughable." His estimate for what actually gets built by 2030? Twenty to thirty gigawatts. Ten percent of the queue. The rest is vapor. Meanwhile, Oracle just delayed OpenAI data centers from 2027 to 2028. The company carries $108 billion in debt. Negative $10 billion free cash flow. Credit default swaps at 126 to 141 basis points, the highest since 2020. And it owes OpenAI 4.5 gigawatts of capacity under a $300 billion contract. OpenAI loses $9 billion per year. It must grow revenue fivefold in two years just to pay Oracle. The circular financing everyone fears? We verified it. The $610 billion figure circulating through financial media is inflated tenfold. Actual executed investments: $63 to $70 billion. Still unprecedented. Still concentrated risk. But not the imminent collapse some predict. The real constraint is not financial. It is physical. Electrons cannot be printed. Transmission lines take a decade. Generation capacity takes five to seven years. The hyperscalers have the capital. They do not have the kilowatts. This is not a bubble that pops. It is a ceiling that binds. Data center vacancy: 1.9%. Pre-leasing: 84%. The demand is real. The timeline to meet it is not. Position accordingly. Read the full article - https://t.co/PRUjv6zVI5
Block all AI browsers for the foreseeable future: Gartner Analysts worry lazy users could have agents complete mandatory infosec training, and attackers could do far nastier things https://t.co/fY7ECeszTD @Gartner_inc @theregister
SpaceX just completed its 550th booster landing Falcon booster B1093 stuck a perfect touchdown on OCISLY after its 9th flight, delivering more Starlink satellites from Vandenberg From the first landing in 2015 to 550 today https://t.co/qUwwtXZFSl
Neuralinkโs first patient in the UK, Paul, who is paralyzed by motor neuron disease, has received a brain implant. Just hours after surgery, he was able to control a computer with his thoughts and is now using it to play games and regain independence. https://t.co/OqqiqwDqmh
Falcon 9 launches 27 @Starlink satellites from California https://t.co/eyYDpBFamK
https://t.co/pikmydpaPh