We’re going to add so many layers! https://t.co/lMLQ0hgPyu

BREAKING: @BAXUSco announces a decentralized price-mapping layer for real-world assets, only for @solanamobile 🥃 https://t.co/6dQSQ8ahKr

Global DePin. Decentralized Pricing Infrastructure. Real Time Real World Applications Coming exclusively to Seeker users. https://t.co/TzW1by3OL4

man, the last time i was on a boat this late in the persian gulf https://t.co/R85LlNKmW5

@Greg_Holliday1 @PatriotXV11 @elonmusk before you try to tell someone they are not a native american you better make sure they are not an actual native american lol ... https://t.co/4SHCa7PZzV

This is so cool. I just asked Grok in my Model Y to navigate to three different destinations, and it automatically added them all at once, in the correct order, to the navigation in my car. Grok can also remove or edit destinations. https://t.co/aRJ93V6MJP

「異世界で始める白猫少女との二人暮らし　ー　だったのに、いつの間にか美少女ハーレムの主になって世界を救ってました　ー　 / 46-7.狐釣り」投稿しました！ https://t.co/xFjsu0diNk #narou #narouN2301JC

@navarreynolds2 agreed but we can't expect much from a man named gary https://t.co/jSykz7maTY

@narindertweets Kaur anti British anti white racist POS. You really are a stupid cunt. https://t.co/bXPbVczgUx

“I have some native” well geeze, let them go! https://t.co/P09BKhbVmr

【 秋の小さな森 】 想像するの好きなんです この森に棲んですこの子たちを…w もちろん個体別に性格あるのだ( ･́∀･̀)ﾆﾔﾆﾔ #ARofTheAutumnleaves #GOsnapshot #GOsnapshotCelebration #ARofTheDay https://t.co/geWLyiuTVD

+2 more

Sergey Brin admits Google messed up by under-investing in the transformer architecture it invented Google was too scared to release chatbots that "say dumb things", so it under-invested in scaling compute "we didn't take it very seriously... and openAI ran with it"

Average pay looks wildly different in the world: Comparison of average monthly salaries across 69 major cities worldwide in 2025, and five-year cumulative changes since 2020, using data from Numbeo cost of living database visualised by @VisualCap via @SpirosMargaris. https://t.co/0oo0HuwZFX

Blockchain, FinTech, Krypto – kann die Schweiz ihren jahrelang aufgebauten Vorsprung halten? https://t.co/3uMtNixZKZ Der Bundesrat wird aktiv, weil Länder mit fortschrittlichen Regulierungen zum Überholen ansetzen. Reaktionen der Verbände auf die Pläne des Bundes. via @News_ZV https://t.co/QxERcSbCXF

Blockchain, FinTech, Krypto – kann die Schweiz ihren jahrelang aufgebauten Vorsprung halten? https://t.co/3uMtNixZKZ Der Bundesrat wird aktiv, weil Länder mit fortschrittlichen Regulierungen zum Überholen ansetzen. Reaktionen der Verbände auf die Pläne des Bundes. via @News_ZV https://t.co/QxERcSbCXF

Warren Buffett talking about the importance of patience 👀 https://t.co/KyPSNtkQr7

Warren Buffett talking about the importance of patience 👀 https://t.co/KyPSNtkQr7

Charlie Munger: "One of the reasons that Berkshire has been so successful is there's practically nobody at headquarters. We have almost no corporate bureaucracy." "Some of these [managers] that go into a stupid bureaucracy and fire a third of the people and then the place works better... They're doing the Lord's work, but you wouldn't think so if you were working there. There is a lot of horror and waste in bureaucracy — and it's inevitable."

American Hedge Fund billionaire Bill Ackman is a learning machine "You can learn investing by reading books." Here are his 9 favorite investing books everyone should read: https://t.co/FqFuyCUys0

AI investments reveal the new capitals of innovation. This map helps you understand where the market is moving today and why it matters for the future of business. Infographic by @StatistaCharts via @antgrasso #AI #BusinessStrategy https://t.co/tgPXM8E3ol

🇨🇳 CHINA RUSHES TO BUY NVIDIA AI CHIPS Nvidia is thinking about making more H200 AI chips because Chinese tech giants like Alibaba and ByteDance moved fast to buy them once the U.S. reopened the door on exports. The H200 is an extremely powerful AI chip made by Nvidia and built by TSMC, and right now it is the strongest option Chinese companies can legally access. The U.S. government said Nvidia can export the H200 to China but must pay a 25% fee for letting the chips leave the country. China’s government is uneasy because local chip makers still lag behind, and officials are debating rules like forcing buyers to bundle domestic chips. Nvidia also has limits since it is shifting to newer Blackwell and Rubin chips while competing with Google and others for scarce TSMC factory capacity. This is a global tech power struggle where AI performance matters more than politics, and China’s demand is way bigger than its supply. Source: Reuters

‘Ruined my Christmas spirit’: McDonald’s removes AI-generated ad after backlash https://t.co/3y3AyGzpmD

Donald Trump threatens federal funding cuts for states with ‘onerous’ AI laws https://t.co/gyN1xcoD6J @joemillerjr @ft

AI’s errors may be impossible to eliminate what that means for its use in health care https://t.co/7ldAMfaHqi @ConversationUS

Disney wants you to AI-generate yourself into your favorite Marvel movie https://t.co/iJ6GYGPez5

we challenged Claude Code and Codex to fine-tune an open source LLM. Claude Code won by surpassing the base model score on its fourth run in 138k tokens. We gave both agents the same task: - Fine-tune Qwen3-0.6B on OpenR1/codeforces-cot - Beat the base model on HumanEval - 5 jobs max, unlimited compute and tokens The results: - Claude Code matched the base model score a few times and eventually beat it. - Codex struggled to get jobs to start at first and under-trained a few models. It wasn’t able to surpass the base model’s score. - In general, Claude was better at consuming tokens before starting jobs and maximising its context by ingesting the most amount of information from the skill.

3000 Reachy Mini on their way https://t.co/K4Qen1Dkgf

3000 Reachy Mini on their way https://t.co/K4Qen1Dkgf

Looking forward to what @AboreanFi is preparing for us 👀 “When the Forge inevitably heats up, it will be an amazing value add to ABX through early launch access and massive fee generation.” - @nativeaborean

The PAUSE Act is simple: stop the abuse, shut down the magnets, and restore order. Corporate elites and the Left broke the system through H-1B and “diversity” visas. We must PAUSE immigration so America can reset the conversation. https://t.co/KmzTetPEip

I might have just found an OpenAI employee’s wallet on Polymarket User 'pony-pony' is up $80,000+ betting on internal company timelines that nobody else could possibly know Look at the precision: - GPT 5. Knew it wouldn't be released by Aug 5, but was guaranteed by Aug 10. And placed an exact bet on August 7th (3/3 hits) - OpenAI Browser. Knew August wouldn't happen, but Q4 was confirmed. Loaded up "YES" on Oct 31 for a $50k profit (3/3 hits) - GPT 5.2. Two high-conviction bets: "No" by Dec 9, "Yes" by Dec 13 (2/2 hits) After all this, it becomes obvious that he is somehow connected to the company and knows their exact release dates He bets small amounts, probably to stay under the radar. No way this is his only wallet Found wallet here: https://t.co/7lk30PCVbT

# Big Post On The Oddities of Frame Scheduling The mouse across all OSes have special hardware & GPU paths dedicated to it. This ensures that even when everything freezes, your mouse moves stay smooth. It’s one of the last thing to ever lag on your display. When you see the red square lag behind the cursor, you’re seeing the (minimum!) usual webapps lag vs the ground truth hardware-accelerated cursor. But even the ground truth lags. macOS actually has a higher input latency (both iOS and macOS are triple buffered, as opposed to double for Windows/Linux) because it prioritizes smoothness “correctness” over raw latency (ask chatgpt). Triple buffered = 3 frames, which is a lot, but with 120fps refresh rate like ProMotion, it’s less noticeable. On the other hand, webapps are in theory infinite frames behind because many use a timeout for coordinating certain logics instead of using frames. So 50ms timeout means 4 frames on 60hz, 7 frames on 120hz, and your favorite helper library for e.g. virtualization or whatever probably has some timeout somewhere so anything you do will pathologically have at least that much latency. cc @andy_matuschak Regarding scroll: yes the black items are actually lagging, just that there’s no reference on screen for comparison. The browser scroll event is fired async; they're the "most" async on Safari (thus on every iPhone, since every browser on iPhone is actually just a UI around Safari). This is why you see some random mobile site with a floating button/menu that jitters vertically as you scroll; those read the stale scrollTop value from the async scroll event and calculate an offset that's unsynced vs the scrolling view behind the floater. If you want to prioritize smoothness over raw latency, you'd in theory hand-update both the floater and your scroll UI's position (e.g. read from an invisible view's scrollTop and update the real visible view's scrollTop. And since the floater UI also uses the same wrong scrollTop, at least everything's wrong at the same time). This is very complex; plus Safari still caps JavaScript updates to 60fps by default. Btw scrollTop value is also wrongly truncated across different browsers by spec. I have a bigger tangent on scrolling in this messy half-finished repo here: https://t.co/XjJUqGAlpi cc @jordwalke. People flee to CSS to use one-off fixes but those don't cooperate with any of the JS for the above reasons Also requestAnimationFrame likely will add drawing problems. Chrome skips repaints: https://t.co/3LD2OPR7Hv and rAF scheduling is weirder than you think: https://t.co/fPHv5MbqTd. More info at https://t.co/oD5EOml3T9 And then React itself adds complexities. Here's a thread I've barely found back. Feel free to read or ignore it but I myself forgot half of the nuances there already: https://t.co/ZWTM6Le9pt https://t.co/ScHUR55KDs And on top of this, folks throw around useEffects for data syncing and now we're an arbitrary number of frames behind Related, some posts by @trishume on keyboard latency: https://t.co/ctXbxhlSZa https://t.co/a1J7dDkvXX

+1 more