technology
Trending UpSnapshot Window: 2026-09-30 20:40 UTC · ← Back to Non-Crypto Overview
Topic Engagement Summary
Total Engagement - Comments: 1298, Retweets: 3105, Likes: 27480, Impressions: 1085415
Verbatim Citations & Public Discussions 120 source posts analyzed
· 2026-09-30T11:46:07+00:00That laptop you refused to properly shutdown will sleep and not wake up one day. Continue!
· 2026-09-29T21:10:56+00:00El IA NOS VA A MATARRRRER
· 2026-09-30T12:41:17.314000+00:00AI now takes $1 in $12 of company software budgets, up from $1 in $70 last year (Zip Enterprise AI Index)
· 2026-09-30T15:20:00+00:00YouTube going nearly two decades without a serious rival is the most insane level of dominance we’ve ever seen on the internet.
· 2026-09-30T13:49:38+00:00GPT-6.1 Sol (high) consistently seems to think about it's maximum reasoning budget and it's annoying the shit out of me. I barely see it think about my task/problem at all, not sure what RL torture this model went through, poor GPT
AI visual note: The image shows a single line of AI reasoning text displayed on a dark background, reading: "I'm thinking about the token limit we have – is it 24k or 25k?" This relates to the post by illustrating the exact behavior they're complaining about—GPT-6.1 Sol wasting its reasoning budget on meta-questions about its own token limits rather than focusing on the actual task.
· 2026-09-30T20:27:05+00:00BREAKING: Google introduces its new frontier AI model Gemini 4 Argon
AI visual note: The image features a stylized Google Gemini logo alongside the text "Gemini 4 Argon" with a large white "4" set against a blue gradient background, visually promoting Google's newly unveiled frontier AI model.
· 2026-09-30T20:31:46+00:00Asked my dot to paint me. After sending me something cartoonish, I asked it to try a bit harder and to find a recent picture of me on the internet. It did better. Below video is me clicking through my dot's computer.
· 2026-09-30T15:23:49+00:00Is this possible today? @joshkim
· 2026-09-30T10:01:37+00:00Satisfied! I used Claude to make a short documentary about Shenzhen. In the time it took me to step out and pick up a kid, it was already done when I got back—pretty good results. Generated with just one sentence; the AI found the footage itself and edited it on its own.
· 2026-09-30T15:29:57+00:00Android Phones Googlebooks Your phone's files are just *there* in the Files app on your Googlebook Continue an activity from your phone in an app or browser on your Googlebook Cast your phone apps on your Googlebook. You can resize them and continue to use your phone
· 2026-09-30T18:02:00+00:00I’ve used both ChatGPT and Claude for years. Seeing Astra's quality lately, I feel like Sam Altman has lost the AI race.
· 2026-09-30T20:19:02+00:00P.S. - I know some cool people at DeepMind read my posts and I'd love to give some feedback :P
AI visual note: This image is a benchmark comparison table showing Gemini 4 Argon (DeepMind) versus GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5 across categories like Knowledge Work, Agentic Coding, ML Engineering, Science/Math, Long Context, Computer Use, Multimodal Understanding, and Cybersecurity, with Gemini 4 Argon frequently scoring highest (shown in blue). The caption playfully suggests feedback for DeepMind, hinting at the post's intent to engage DeepMind employees and possibly critique specific results visible in the table (such as Gemini's weaker 19.6% on Harvey's Legal Agent Benchmark despite leading most other categories).
· 2026-09-30T18:15:49+00:00Three awards, one message: Intel’s teams continue to drive innovation through advanced technology, from silicon to systems. In 2026, three independent organizations have recognized us: @FortuneMagazine ranked Intel No. 14 on America’s Most Innovative Companies, @Forbes ranked
· 2026-09-30T17:56:01+00:00AWS just got a lot easier to use.
· 2026-09-30T17:12:21+00:00IMPORTANT: Daybreak access ends tomorrow unless you get a hardware passkey. Just as a reminder you can get almost 50% off with your ChatGPT subscription.
· 2026-09-30T20:45:00+00:00Someone open-sourced an entire AI company with 172 specialists Headcount gives Claude, ChatGPT, and Codex 16 departments of specialized agents covering everything from finance and sales to legal, security, product, and operations. Repo:
· 2026-09-30T20:23:59+00:00Google cooked on Gemini 4! Now they just need to be more competitive on coding harness (Antigravity) + personal agent (Spark).
AI visual note: The image is a comparison table benchmarking Google Gemini 4 Argon against GPT-5 Astra, Claude Fable 5.1, and Claude Opus 5.5 across various tasks including Knowledge Work (Vals Index: 68.9%), Agentic Coding (Vibe Code Bench: 91.9%), and Long Context (GraphWalks: 99.7%), with Gemini 4 highlighted as the leader in most categories. This supports the post's claim that "Google cooked on Gemini 4," as the table shows Gemini outperforming competitors across most benchmarks, while implicitly highlighting the need for better coding and personal agent products.
· 2026-09-30T20:39:01+00:00yeh fuck that, im not touching anthropic
· 2026-09-30T20:32:10+00:00Vano in action. Join our discord for early access code.
· 2026-09-30T20:46:45+00:00BREAKING: Google projected to dethrone Anthropic as the company with the best AI model at the end of October. 83% chance.
· 2026-09-30T20:32:24+00:00Wow! Wow! @GoogleDeepMind is back with Gemini 4 Argon! Insane benchmarks! In one fell swoop, they’ve jumped to the very top of the AI frontier. As I’ve said before, never bet against Google in the age of AI! Can’t wait to try Gemini 4 ASAP!
AI visual note: A benchmark comparison table showing Gemini 4 Argon's performance against GPT-6 Astra, Claude Faible 5.1, and Claude Opus 5.5 across categories including knowledge work (Vals Index: 68.9%, AutomationBench: 51.3%), agentic coding (DeepSWE v1: 77.9%, Vibe Code Bench: 91.9%), science/math (LABBench 2: 88.8%), and long context tasks (GraphWalks: 99.7% at 128k BFS), supporting the post's claim that Google DeepMind's new model has surged to the top of AI benchmarks.
· 2026-09-30T20:27:22+00:00Looks like someone at Apple fucked up something on the backend... TestFlight links and App Store metadata are broken
AI visual note: The image shows a broken App Store page with "testflight.template.public.link.hero.title" replacing what should be the app name, and the body text contains repetitive template code like "testflight.template.public.link.content.installGetMarkupTestText.twoThree" and ".paragraphOne .installation-useelink .paragraphTwo" filling where normal descriptions should appear—matching the post's observation that Apple's TestFlight links and App Store metadata have broken on the backend, exposing placeholder template strings instead of proper app data.
· 2026-09-30T18:52:56+00:00Giving up on H200 and going RTX 6000 TP=8 using 3 of these switches. Managed to pry the last 3 so I can do root/leaf architecture This is going to be an enormous undertaking, powering 4 additional 600w GPUs and creating watercooling loops is going to require significant effort
· 2026-09-30T20:43:52+00:00Gemini 4 Argon is launching first in a limited capacity so Google can make sure it’s not misaligned.
· 2026-09-30T18:29:40.699000+00:00ChatGPT Astra when it gets access to my computer
· 2026-09-30T20:15:32.532000+00:00Gemini 4 is out: The competition has woken up
· 2026-09-30T20:36:23+00:00oh wow... Gemini 4 Argon Looks like it beats Opus 5.5 on most benchmarks. Is Google finally back??? Is this a dream?
AI visual note: The image is a performance comparison chart for Google's "Gemini 4 Argon" model, showing it benchmarked against GPT-6 Astra, Claude Opus 5.5, and Claude Opus 5.1 across categories like knowledge work (Vals Index: 68.9%), agentic coding (DeepSWE v1: 77.9%), science/math (LABench 2: 88.8%), and multimodal understanding (LVBench: 91.7%). This directly supports the post's excitement, as Gemini 4 Argon's scores frequently outperform or match the Claude models across most benchmarks, fueling speculation about Google's competitive resurgence in AI.
· 2026-09-30T17:29:59+00:00i love the way my layout looks rn
· 2026-09-30T20:21:51+00:00Gemini 4 is out and it's a brutal benchmark mogging by Google
· 2026-09-30T20:29:07+00:00Gemini 4 Argon has the same 95% cache read discount that GPT-6.1 Sol has. Highest in the industry!
· 2026-09-30T20:29:35+00:00Google is back in the game for best AI model
AI visual note: A benchmark comparison table showing Gemini 4 Argon outperforming GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5 across various categories including knowledge work (Vals Index 68.9%), agentic coding (Vibe Code Bench 91.9%), long context (GraphWalks 99.7%), and multimodal understanding (LVBench 91.7%).
· 2026-09-30T15:08:19+00:00Came across a small site called Cursor Lab, which collects a bunch of custom mouse effects, like dots and rings, a pair of eyes, liquid glass, pixel trails, smoke, and ones that make the background grid deform along with it If you want to add some small details to your website,
· 2026-09-30T20:08:04+00:00thank you Luke! the witcheer almanac is one email a week. what's in it: - the Hermes Wingtips and Autopilot prompts I posted that week - what changed in each Hermes Agent release - the Flightplans, my long guides - my local model benchmarks on my RTX 5090 it just hit 100
AI visual note: A screenshot of an email newsletter titled 'the witcheer almanac: 21 to 27 september 2026' from sender 'witcheer' to 'Luke Dev', detailing updates on v0.21.4 and v0.21.5 releases.
· 2026-09-30T20:29:49+00:00Update: We now know a little more about the upcoming reboot from X Premium into X Subscriptions. The new Subscriptions are also known as XPass. There are 4 tiers. Premium upgrades to Plus, Plus to Super, and Super to Ultra. Each tier includes everything in the one below it.
· 2026-09-30T20:05:53+00:00Gemini 4 is out. Sort of.
AI visual note: A small, blurry financial data table with various columns and rows of numbers, featuring some highlighted blue cells.
· 2026-09-30T20:30:00+00:00Loved scroll-driven animations? Meet their scroll-triggered sibling. Instead of tying animation progress directly to the scrollbar, you can trigger full time-based CSS animations when crossing a scroll offset. See how to build them with pure CSS → goo.gle/4dhcvmh
· 2026-09-30T18:50:16+00:00The Harvard man used to own you. Then the Stanford man took the room. Better tech network. Better Valley access. Same one brain bottleneck. In 2026 the Agentic man walked in and took the desk. Not a better résumé. A private operating system powered by army of Grok Bots. 30+
· 2026-09-30T20:37:39+00:00excited to get your feedback, lots of progress, and still lots to improve on : )
· 2026-09-30T16:17:10+00:00yooo this launch video is so sickkk! can’t wait to try Lucas
· 2026-09-30T19:56:47+00:00Cisco Catalyst SD-WAN Fundamentals | Cisco U. Theater Sessions | Cisco Connect 2026
· 2026-09-30T20:23:11+00:00Getting flashbacks of G5 launch
· 2026-09-30T20:19:31+00:00Gemini 4 Argon entered the chat like a final boss
AI visual note: A dramatic illustration showing a large, powerful golden dragon labeled "Gemini 4 Argon" with a crown, standing over three smaller, battered dragons marked "Claude," "GPT," and "Grok."
· 2026-09-30T19:07:14+00:00Do we all love a grid view or list view? All on #one4home launcher - go check it out!
· 2026-09-30T18:49:25+00:00There appears to be a different like animation on this tweet??
· 2026-09-30T20:32:47+00:00GEMINI 4 ARGON JUST BEAT CLAUDE BY 5X AND GPT BY 4X Google shipped Gemini 4 Argon with no warning. It leads on 13 of 18 major benchmarks and already runs inside Google itself. Runs on Google + 1M input tokens + 1M output tokens. One prompt returns a reply the size of 8 Harry
AI visual note: A benchmark comparison table showing AI model performance, with 'Gemini 4 Argon' highlighted in blue, comparing against GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5 across various tasks.
· 2026-09-30T18:57:47+00:00A content creator who participated in the development of Halo Infinite said that he tried to warn that the Halo Infinite demo wasn't good a week before the Xbox Showcase 2020 However, the creative director, Chris Lee, would have said that the problem was the YouTuber's monitor +
· 2026-09-30T20:17:22+00:00No private Lobbies? Come on we can't be pulling this crap after what people complained about with Halo Infinite...
· 2026-09-30T19:24:38+00:00most inbox tools help you organize work. Sol actually starts doing it. follow-ups, support replies, invoices… then you just review and approve. this feels like a much better direction for ai assistants.
· 2026-09-30T18:55:06+00:00Seems like @code is starting to tell you: your agent primitives need to move. There is now a new migration banner in the Chat panel, and it is part of a much bigger change happening under the hood: the move from the old Local harness to the new Agent Host architecture built
· 2026-09-30T20:24:00+00:00quick side note: google announced Gemini 4 Argon but it is not fully public yet internal use first, then trusted testers, then Ultra + API the interesting part is: - 1m output tokens - strong on long-horizon coding and agent workflows - google is claiming 77.9% on DeepSWE -
· 2026-09-30T20:21:38+00:00Introducing OpenAI Asstra
· 2026-09-30T19:35:11+00:00BREAKING Google achieved recursive self improvement but they are not confident of it since it was done with Gemini 3.8 Flash. They are planning to use Claude and ChatGPT to achieve RSI again.
AI visual note: Sundar Pichai, CEO of Google, speaks on stage wearing a dark suit with the colorful Google logo displayed on the screen behind him.
· 2026-09-30T20:23:59+00:00This is a bigger leap than it seems in the video. So far, an AI character was just a talking face. You wrote, it responded, and at most it moved its mouth. Here the character has a body and a stage: 1) It acts full-body. It dances, gestures, acts silly. 2) It moves around the
· 2026-09-30T20:30:34+00:00Can someone explain this in “Google cooked or we’re still waiting” English
AI visual note: A benchmark comparison table evaluating AI models 'Gemini 4 Argon', 'GPT-6 Astra', 'Claude Fable 5.1', and 'Claude Opus 5.5' across categories including Knowledge work, Agentic coding, ML engineering, Science and math, Long context, Computer use, Multimodal understanding, and Cybersecurity.
· 2026-09-30T20:25:28+00:00Anthropic → Claude Opus 5.5 OpenAI → GPT-6.1 + Dots Google → Gemini 4 Argon 3 nouvelles IA en 8 jours. Tu pars en vacances une semaine, tu reviens avec une IA de r****d
· 2026-09-30T20:22:49+00:00Our Tactical Talks return next week for a deep dive into esports motion design! Tune in to learn more about the art of bringing brands to life Tuesday, Oct. 6 5pm ET/2pm PT/11pm CET @reiabeamz , Motion Designer Join today & verify to gain full access:
· 2026-09-30T18:45:03+00:00President Donald Trump met with top tech leaders Tuesday to discuss the future of artificial intelligence, which he now calls "super intelligence."
AI visual note: A digital illustration showing the letters 'AI' inside a glowing blue brain-shaped outline, placed on a dark, high-tech circuit board background with binary code.
· 2026-09-30T20:06:29+00:00我好像有点起飞了 有个工具爆了 但是大哥们你们咋不认证
AI visual note: A mobile app screenshot in Chinese showing a user invitation management page titled "受邀用户" with a list of users, registration times (2026-09-30), invitation codes (QY001), and authentication statuses marked as "未完成" (incomplete).
· 2026-09-30T20:25:39+00:00The team finally escaped the blank-prompt era. Time to build from what’s already working. @avastudio_
· 2026-09-30T19:41:00+00:00A cleaning company can lose the job the moment a customer has to wait for a callback. So I gave Claude one prompt and it built this. Brightside Cleaning (made up). The customer picks bedrooms, bathrooms, home size and how often. The price updates as they click. Choose every two
· 2026-09-30T20:19:31+00:00Cybertruck owners: if you're not doing this, you should be. Muddy drip after a wash = gunk trapped behind the fender flares. They pull off in minutes without breaking clips. Spray it out, pop them back on. No more Cyber-drip.
· 2026-09-30T20:32:24+00:00google either went insane with the benchmaxxing or dropped a generational model. there’s no in between.
AI visual note: Benchmark comparison table showing Gemini 4 Argon outperforming GPT-6 Astra, Claude Fauble 5.1, and Claude Opus 5.5 across categories including Knowledge work, Agentic coding, ML engineering, Science and math, Long context, Computer use, Multimodal understanding, and Cybersecurity
· 2026-09-30T20:00:00+00:00【個人AI代理人】 Muse是Meta推出的「個人AI代理人」，簡單說就是它可以幫你做事。結果有人試用後嚇出一身冷汗，然後直接刪了 這位叫Matt Robb的男人把Facebook
AI visual note: An Instagram post by matt.j.robb describes a handled Facebook marketplace situation involving a Usman MX Keys Mini pickup and a refused porch pickup, including screenshots of text messages with engagement metrics of 2.1K likes and 1.6K shares.
· 2026-09-30T20:28:43+00:00谷歌发布Gemini 4，各项评分大幅领先！哈基米支棱起来了！ 核心定位与规格：Argon 专为复杂的长周期工作流设计，单次输出 token 上限从 64K 大幅提升至 100 万 tokens。首发定价为输入每百万 token 2 美元、输出每百万 token 10 美元（缓存输入享 95% 折扣）。 内部工程实践：Google
· 2026-09-30T20:00:09+00:00An update on threetopia.com - A home for Three.js creators. Here you see an example map. There are currently four different worlds here, all from different creators. Each creator can put their world inside a ready-made tiling system, which by itself provides a
· 2026-09-30T18:45:00+00:00Tired of messing around with software back and forth just to download online videos—hasn't that already driven you up the wall? DataTool is this tool that you can just open in your browser and use right away. It covers a pretty impressive range of platforms: ①X (Twitter)
AI visual note: A Chinese online video downloader website titled "在线视频下载器" featuring a URL input field with a "开始解析" (Start Parsing) button and various social media platform icons indicating compatible sites.
· 2026-09-30T20:19:07+00:00@avastudio turns market signals into finished creative without starting from scratch.
· 2026-09-30T19:25:29+00:00I built Hiwaar, an AI interviewer you can FaceTime to practice for your next interview. It holds a real conversation, reacts to your answers, and digs deeper when needed. Teamed up with @GaryGao891219 and Chert to bring it to FaceTime. No app, just pick up and talk.
· 2026-09-30T20:28:54+00:00“Xbox not for sale” Low IQ Red gecko has been very confused lately. First it’s him not understanding “console exclusive” in the same terminology he used just a few years ago. Now this ASHA is sure disruptive to the XBOX HATER
· 2026-09-30T18:36:34+00:00designeng.tools will never be “finished.” As long as I keep finding great resources, I’ll keep adding them. Already at 250+ curated resources: → 70+ inspiration sites → 55+ component libraries → 35+ design tools → 20+ icon libraries → and a lot more The goal:
· 2026-09-30T20:39:58+00:00wow, is google back?
· 2026-09-30T20:28:37+00:00This is crazy man!! Al going from "give me a task" to "I found something that needs doing" is a pretty massive shift. Sol is showing exactly that.
· 2026-09-30T20:20:53+00:00seiso 不只是一个检查项目文档的 linter 了，今天把它背后那套规范单独发了出来，叫 Seiso Convention Specification 大概就是给项目里的 Markdown 文档定几条规矩。每篇文档声明自己是 howto、reference 还是
· 2026-09-30T19:40:14+00:00We have too much fun with these screens!
· 2026-09-30T19:24:52+00:00I think the added details are enough, I don’t want the to look very dirty, it’s not an abandoned building. I’m already starting to place other objects. My task was to create only the building modules, so I’ll use the free set from Fab. #ue5 #3d #3dart #environment #art
AI visual note: A 3D rendering shows a single-story gray house with a wooden accessibility ramp and deck, situated on a sandy surface under a partly cloudy sky.
· 2026-09-30T18:46:37+00:00Buzzopolis: Hive Incremental A desktop incremental roguelike about bees, mutations, management and controlled chaos. Final polish before Steam. Wishlist store.steampowered.com/app/3835550/Bu …#GameDev #IndieDev #IndieGame #PixelArt #Steam #IndieGames
· 2026-09-30T20:21:24+00:00the result looks suspicious kek anyone with a deep mind should give me a name for this masterpiece
· 2026-09-30T20:43:08+00:00Impressive results from a great collaboration with our Instagram engineering team! Read the full deep dive
AI visual note: Promotional graphic reading 'How Instagram Direct reduced token cost by 33% with Jetpack Compose' showing a smartphone displaying Instagram DM conversations
· 2026-09-30T19:01:07+00:00A 1% improvement can make a massive difference at fleet scale. See how additive manufacturing is delivering measurable impact in aerospace: bit.ly/3VDDYbL
· 2026-09-30T18:06:47.060000+00:00I was tired of relaying context between Slack and my Claude Code terminal‚ so I built a shared doc where your whole team's agents meet
· 2026-09-29T09:35:24.451000+00:00Times Car confirms data breach affecting 6.6 million user accounts Suggested
· 2026-09-30T20:37:53.564000+00:00How to connect CM360 , Google analytics and our data lake data together? For reference, when someone fills our form for a consult, we collect that data (first/last name, email, phone number, consult type) and save it in our database. However we don't really know what the user is doing before they got to that consult conversion page. We want to add more data to our advertising campaigns so that we can track the journey better. From research it looks like you can combine CM360 (DCLID) , Google analytics (pseudo user id) and then our own customer id together somehow? For example
· 2026-09-30T19:27:03+00:00Xbox is celebrating 25 years. @Xbox Fanfest London was amazing. Read @jordicampbell1 experience of the event, we think he enjoyed it. @majornelson , you made his day :)
· 2026-09-30T20:10:28+00:00GPT 6.1 Sol Light vs GPT 6 Astra Ultra vs GPT 6.1 Sol Ultra Same prompt: Build ThreeJs Enropean roundabout
· 2026-09-30T20:46:45+00:00The prompt is part of your coding workflow. Now you can edit it with the Vim keys you already know. Vim mode in Muse Code brings familiar navigation, text objects, and Visual selections directly to your prompts. 1. Enable Vim: type /vim and press Enter for this session, or set
· 2026-09-30T19:22:59+00:00I took a random video from my camera roll and asked opus 5.5 to edit it into a saas launch video
· 2026-09-30T19:49:55+00:00先看置顶，再决定关不关注。 我只放自己跑通、能直接抄的作业—— 1. 英文长视频自动剪成“中文90秒精华”（Codex + HyperFrames） x.com/AI_DVD6/status … 2. 零基础做完35秒短剧（Opus 5.5） x.com/AI_DVD6/status … ✓ 完整、会更新的清单在这条的第一条回复。
· 2026-09-30T20:21:36+00:00Endless agent loops are not a tools problem They are a stop condition problem More tools only give an agent more ways to continue An answer that sounds complete is only a feeling Stop only when three things are true The result exists in the right place It works because it was
· 2026-09-30T20:39:17+00:00Vodafone müşteri hizmetlerini aradım beyfendi sesinizi alamıyorum diyo alamazsın tabi çekmiyo ki telefon,
· 2026-09-30T19:00:00+00:00Bring life to your screen. Sep 2026 · Pack #9 · Wallpaper #01 Beloved Birds by Arif Afifudin 4K download here. Full pack in 8K on WallpaperZ: iOS - l.linklyhq.com/l/1u30X Android - l.linklyhq.com/l/1u30Y Website - wallpaperzapp.com/pack/036_sep_2 … Lock Screen or Home
· 2026-09-30T19:15:28+00:00Apple Pay just dropped in India. Available for Razorpay merchants right away. So we cooked a launch video.
· 2026-09-30T17:56:51+00:00marketing's claude code just dropped with @avastudio_ . old 2-week creative loops are cooked
· 2026-09-30T20:02:28+00:004 streamers. 1 boxing ring. too much ego opus 5.5 absolutely cooked. $220 in API credits. expensive, but a year ago this would’ve taken me a week. opus is cheaper for me right now. astra is equally good, sometimes better. never been more into building.
· 2026-09-30T18:50:06+00:00Pasted my link into @avastudio_ and got instant, usable hooks instead of generic scripts.
· 2026-09-30T20:02:34+00:00If you’re a VR dev and have been sitting on an idea for an older / female skewing audience but have been put off by the current market... now is the time to build
· 2026-09-30T19:58:06+00:00#ONCEHUMAN #ONCEHUMANCC #VirtualPhotography #oncehumanconsole
· 2026-09-30T20:00:12+00:00Claude Fable 5.1 just dropped and it's already the most creative model on the planet. I used to make motion graphics for Apple and PlayStation, and they took weeks. So you'd think Fable needs a long prompt. It doesn't. Level one: Claude Design, an animation on how to drink milk.
· 2026-09-30T18:50:20+00:00Finally, an AI tool iterating on real performance instead of random visuals. @avastudio_
· 2026-09-30T19:43:46+00:004K Retake Ninja #CS2 #counterstrike2 #Twitch #gaming #clutch
· 2026-09-30T20:43:58+00:00Mixture-of-experts inference has a load-balancing problem most people don't benchmark for. Standard 8-way expert parallelism leaves hot GPUs at 95% utilization while cold ones sit at 20-30%. Widening to 16-way expert parallelism with online load balancing flattens that gap to
AI visual note: A Spheron Network infographic titled 'Expert parallelism imbalance solved' comparing an 8-way baseline (95% hot GPUs, 20-30% cold GPUs) with a 16-way + load balancing solution achieving 82-88% utilization, 30% more tokens/sec, and 23% lower cost per M tokens.
· 2026-09-30T20:44:24.559000+00:00I tried cable managing for the first time. It's better than nothing
· 2026-09-30T20:16:00+00:00Tesla is now rolling out an update that could drastically improve your cabin experience youtu.be/p4XbYUh2ynU
· 2026-09-30T18:46:45+00:00vibe coding finally hit marketing. pasted a link and got full creative variations @avastudio_
· 2026-09-30T18:51:16+00:00happy @grok day
AI visual note: A crudely drawn pink cartoon rabbit with a smile and a blue bandana with a yellow symbol stands against a starry space background.
· 2026-09-30T20:11:00+00:00Seven walked out of the Nest and wrote down what they believed. Fifty-seven pages. Canto grew out of them. Every quill is a rule it won't break. Before every match, it warns the table how dangerous it could become. At length. Then it plays. Canto, Anthropic's champion.
AI visual note: A split comparison image showing '2026' with two men in suits at the top and '2022' with two boxers in action at the bottom.
· 2026-09-30T19:53:57+00:00Announcing the Agora Conversational AI LinkedIn Live series, starting this October. Grab your seat: linkedin.com/events/7510448 … 𝗙𝗶𝗿𝘀𝘁 𝘀𝗲𝘀𝘀𝗶𝗼𝗻: 𝗚𝗲𝘁𝘁𝗶𝗻𝗴 𝗦𝘁𝗮𝗿𝘁𝗲𝗱 𝘄𝗶𝘁𝗵 𝗔𝗴𝗼𝗿𝗮 𝗖𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗔𝗜 We'll go from "What exactly does
· 2026-09-30T19:03:50+00:00Question to You Dear folks who follow our Fighting Game in progress, my Brother and me set down to talk about the roster, we want the full Game to have around 15 Characters, is there any real Life Fighting style You Like to see? Our Game will be a tactical Fighter in that sense
· 2026-09-30T20:37:54+00:00Gemini 4 looking good
AI visual note: A blue-tinted promotional graphic displaying the text 'Gemini 4 Argon' alongside a stylized multicolored star logo and a large translucent '4'.
· 2026-09-30T20:27:00+00:00While the Foundation Series badge is the coolest and I do miss mine, the subtlety of the Cyberbeast badge is
AI visual note: A close-up view of a car's interior shows dark leather seats, a textured headliner, and a decorative etched panel with purple ambient lighting.
· 2026-09-30T20:30:00+00:00いくつもの脚が交互に沈む。 鉄の体がゆっくりと傾き、また立ち上がる。 砲口は遠くを睨み、装甲は陽を弾く。 一歩ごとに地面が凹み、 埃が舞い、影が長く伸びる。 誰も乗っていないように見えて、 それでも確実に前へ進む。 沈黙のまま、ただ歩く。 #GrokImagine#AIArt
· 2026-09-30T20:30:57+00:00ANTHROPIC SHIPPED A NEW CLAUDE. I WENT THROUGH WHAT PEOPLE BUILT WITH IT IN ONE WEEK AND FOUND 10 PROJECTS THAT SHOULDN'T HAVE BEEN POSSIBLE IN A WEEK 1. 1,113 real LEGO parts, engineered into a working duck. Every one of its 3,204 connections verified for zero collisions and
· 2026-09-30T20:41:09+00:00Her account on here got hacked, she is on instagram and BlueSky. This is worth knowing.
· 2026-09-30T20:44:09+00:00One of my vibe tests for AGI is whether it can understand exchanges like this. Current status: only with extensive handholding
AI visual note: A note app screenshot with dark mode explaining the subtext of a Star Wars meme about Anakin and Padmé, discussing implied dialogue about accusations of rape
· 2026-09-30T19:05:28+00:00$5.00 vs $3.50. Same Coca-Cola. Nearby stores. That’s a 43% price difference. When one store prices your product higher, you risk losing sales without even knowing it. AI Merch spots price gaps across stores, so you can act before they cost you. See every shelf. Sell more.
· 2026-09-30T20:43:19+00:00AI helps me grow
· 2026-09-30T20:44:11+00:00Seems like a great time to dive headfirst into the Apple orchard, let's just hope the fruit's ripe!
· 2026-09-30T20:40:20+00:00Astra really came out of nowhere lately
· 2026-09-30T20:38:51+00:00Usage tells you far more than follower growth.
· 2026-09-30T20:42:37+00:00Another new model…
AI visual note: A benchmark comparison table evaluating AI models (Gemini 4 Argon, GPT-6 Astra, Claude Fable 51, Claude Opus 5.5) across various categories including knowledge, agentic coding, science, and multimodal understanding.
· 2026-09-30T20:44:17+00:00the concept of living 3d worlds is intriguing