technology
Trending UpSnapshot Window: 2026-10-11 03:00 UTC · ← Back to Non-Crypto Overview
Topic Engagement Summary
Total Engagement - Comments: 590, Retweets: 757, Likes: 3931, Impressions: 236943
Verbatim Citations & Public Discussions 135 source posts analyzed
· 2026-10-11T01:45:51+00:00【One-Click 180-Degree Turn Mouse Debuts】 "Razer Basilisk V4 Pro" ・Ergonomic wireless mouse for right hand ・Equipped with "Perfect 180" for turning around with one click ・Instantly turn to face directly behind with a single button ・Features and appeal tailored for
AI visual note: Two gaming mice are displayed against a dark background, with the black mouse resembling the Razer Basilisk V4 Pro featuring its distinctive scroll wheel illuminated in purple and ergonomic right-handed design, while a white and gray mouse with green scroll wheel lighting sits beside it. The lighting highlights the contoured grips and side buttons that align with the post's introduction of Razer'
· 2026-08-11T08:51:51+00:00The hive mind being constructed in low-orbit.
· 2026-10-10T18:00:32+00:00Introducing X Arena, fight as your favorite X character Yilong Ma, Nikita Boar, BoneGPT, The Zuck Name a @ i should add to the game Playable soon directly on the timeline
· 2026-10-11T02:35:14+00:00codebases have always contained slop. we have always been okay with this because code rarely needs to be perfect, outside of mission critical "hard" (eg embedded) software. swe is a unique form of engineering because you're often building something that constantly evolves and
· 2026-10-11T02:44:33+00:00Tesla vehicle light show and drone show at Tesla’s Giga Texas factory tonight. Cybercab in the sky
· 2026-10-11T02:53:08+00:00Model S/X Signature tribute
· 2026-10-11T03:01:00+00:00People involved in the discussions told the outlet that a large cyberattack is among the scenarios drawing the most concern. Read more:
· 2026-10-10T21:11:40+00:00Why do people buy $20k local machines just to run GLM 5.3 at 70 TPS? Bro I just used a random provider on openrouter ran it the entire night and paid $10.
AI visual note: A close-up of a weathered man smoking a cigarette with the caption "You wouldn't get it" — capturing the smug, dismissive tone of the poster bragging about cheaply running GLM 5.3 on OpenRouter for just $10, as if spending $20k on local hardware is incomprehensible.
· 2026-10-10T03:30:03+00:00at this rate i think the next major release of the cuda runtime will be done by vllm
· 2026-10-10T15:17:13+00:00Uses xtool under the hood, made by @kabiroberai
· 2026-10-09T11:14:25.037000+00:00South Korea and Japan are buffeted by hacks as AI lowers bar for cybercriminals
· 2026-10-11T01:00:20+00:008 Linux privacy tips I swear by (yes, you need them) - ZDNET
· 2026-10-11T01:29:27+00:00this is kinda embarrassing. i know it sucks that your field is getting automated, but just say that! it does suck for mathematicians! why do we insist on pretending that ai math is somehow less useful than human math
AI visual note: The image is a slide titled "A thought experiment on alignment and understanding" by Terence Tao, presenting a hypothetical scenario where an advanced AI produces a cancer-curing chemical cocktail validated through clinical trials, then asking whether humans would need to understand the underlying mechanism or mathematics before trusting such a cure. The slide is used to argue that even if AI generates valid proofs or solutions (like AI mathematics), humans still need to understand the reasoning behind them to build trust and verify correctness—so AI math cannot simply be treated as equivalent to human math in terms of reliability and insight.
· 2026-10-10T14:03:33+00:00Today, let’s talk about something that could become a serious problem as AI agents become more autonomous. How do you know which AI agent deserves your trust? An agent might promise to deliver a perfect result, but what happens when the work is incomplete? What if it misses the
· 2026-10-10T22:27:40+00:00This is critical. Local AI has a specific purpose and it's not trying to save money.
· 2026-10-07T16:16:09.555000+00:00Apollo Summit 2026 Keynote Livestream
· 2026-10-11T02:22:15+00:00what model does my @bot use? I have no idea and I don’t care, this is kinda the point
· 2026-10-11T02:13:56+00:00Impossible that this could happen!! Imagine that right now you can do (reverse engineering) for anything you like Website - app - tool - any software!!! You connect this repo to your agent, whether Codex or Claude, and give it the feature or tool or app, and it will shake the
AI visual note: A GitHub repository page for 'REA: Reverse Engineer Anything', described as 'One MCP for reverse engineering across binaries, applications, and runtime behavior.'
· 2026-10-11T02:54:51+00:00She turned a cluttered entrance into a clean, organized space. From scattered shoes to a perfectly arranged entryway, every little detail makes the transformation satisfying to watch. Made using Seedance 2.5 Prompt: ↓
· 2026-10-11T01:58:47+00:00AI replaces you. SI supercharges you.
· 2026-10-11T01:03:57+00:00Went from master suite at @reductoai ’s offsite, 6x size of my room in sf, then back to reality
AI visual note: A view from inside a darkened room looking out through large glass windows at a sunny beach and ocean, with a cloudy blue sky and a small red boat on the water.
· 2026-10-11T01:29:18+00:00A third Yandex data center attack last night, causing large-scale outage.
· 2026-10-11T02:20:00+00:00it's weird that we can tell pretty often which model was used just by looking at the dashboard design
· 2026-10-11T02:30:44.396000+00:00I understand why people are downgrading from 5090 to 5080...
· 2026-10-11T03:00:47+00:00Currently in the country, the way to directly experience Starlink in the air is through Qatar Airways, with routes from Doha to Beijing, Shanghai, and Guangzhou all gradually being equipped with Starlink as standard. Another option is to transfer via Seoul Incheon, then take a
AI visual note: A Korean Air mobile app screen for flight KE751 from Seoul/Incheon to Tokyo (20:30-22:50), displaying information in Chinese that passengers can use free Starlink Wi-Fi during the flight.
· 2026-10-11T02:03:45+00:00YOLO11 + ByteTrack for Estimating Apple Yield and Ripeness Detects fruit from camera footage, tracks it to prevent duplicate counting Aggregates colors from multiple frames to estimate ripeness Achieves approximately 40 FPS using a laptop PC's GPU Occlusion by
· 2026-10-11T01:35:00+00:00An account that posts 5 pieces a day, but each one has its own unique perspective. An account that posts 30 pieces, all just rewriting other people's content. After a year, who do you think will have a more loyal community?
· 2026-10-11T00:40:10.387000+00:00Holy crap. Look what it can do now that they added the new dynamic UI
· 2026-10-11T01:37:27+00:00Excited to finally see this out!!! Glad to have contributed :D Good homage to Zatom-1 and lots of fun easter eggs and newer results/capabilities Try it out, let us know how you plan to use nucleus and zevals
AI visual note: A technical diagram of a multitask transformer architecture for molecular and material modeling, showing conditioning inputs, a transformer backbone with atom encoder/decoder, and task-specific readouts including Euclidean geometry, force vectors, and energy.
· 2026-10-11T02:22:53+00:00I wanted to see if FSD Supervised could find a space in Melbourne Airport's busy & complex short-term car park. It tried the wrong spot first, then snaked up and down the rows, found a bay and reversed straight in. Really impressive. I've linked my full airport video in the
· 2026-10-10T05:10:16+00:00《我用AI》第 3 期上线了。 这期嘉宾阿伟 @zwzw3888 他用AI做了两个微信小程序，全流程跑通、上架，然后没人用。 其实现在做产品已经非常简单了，反而如何把产品卖出去才是最值得思考的问题。 然而在卖出去之前，我们首先应该去验证需求。 阿伟复盘了他的整个流程，也提出了两个解决方案。
· 2026-10-11T01:59:04+00:00also opus 5.5 in the captain's chair with mostly 6.1 sol subagents are incredible
AI visual note: A screenshot of a Mac desktop displaying an email client with a sidebar of accounts (utiltest, transitof) and threads visible, showing technical content about software testing.
· 2026-10-11T01:15:02+00:00Good Reddit posts lead with the tech question, not the promo. TWiT is there for AI, security, Apple, Microsoft, privacy, and what changes next. reddit.com/user/TWiT_tv/ #TWiT #TechDebate
· 2026-10-11T01:53:55+00:00ピーチmodのあちちです！ marshmallow-qa.com/messages/cfd7a …
· 2026-10-11T02:30:00+00:00right now, Sonnet 5.5 is 9 points behind Opus 5.5 and still beats GPT-6 Astra on Image-to-WebDev Arena
AI visual note: A ranking chart for the 'Image-to-WebDev Arena' by Arena.ai, showing 'Claude Opus 5.5: Ranked #1' with a score of 1,749, followed by other AI models.
· 2026-10-11T02:08:08+00:00Tracking players and the ball from tennis footage, and quantifying their positions on the courtTracking players with YOLOv8 + ByteTrack Estimating the ball's trajectory and bounce position with TrackNetV2 Converting footage to actual coordinates using OpenCV's
· 2026-10-11T02:31:31+00:00For context, I don’t use Hermes for coding Just for my personal and work assistant So Grok Bot does that for me now
· 2026-10-11T02:09:02+00:00Most Claude Max subscribers will never touch their new API credit because they don't know where to start. So here's the start. Paste this into Claude Code and it builds your first AI employee: an inbox sorter that runs every weekday at 07:00 on Haiku 5.5, logs what it costs,
AI visual note: A screenshot of a digital advertisement from 'Zephyr' promotes setting up an AI employee, with text reading '$0 payroll. Day one.' and a code block for a 'First Hire Setup'.
· 2026-10-11T02:50:39+00:00𝐆𝐌, 𝐄𝐕𝐄𝐑𝐘𝐎𝐍𝐄 𝐋𝐈𝐕𝐄 𝐒𝐏𝐀𝐂𝐄 𝐈𝐒 𝐎𝐍 Join Connect Engage Grow Follow Support Stay Active 𝐒𝐏𝐀𝐂𝐄 𝐋𝐈𝐍𝐊 Join Now & Invite Your Friends.
· 2026-10-11T01:56:05+00:00AI is out of control! But funny AF!
· 2026-10-11T01:32:21+00:00> Anthropic's Claude Anthropic*
· 2026-10-11T01:44:51+00:00Google Gemini is the smartest AI right now. But 90% of people prompt it like ChatGPT. That's why I made the Gemini Mastery Guide: → How Gemini thinks differently → Prompts built for Gemini → 2000+ AI Prompts Comment "Gemini" and I'll DM it free.
AI visual note: A logo graphic displaying "Google" and "Gemini" separated by a vertical line on a dark navy background.
· 2026-10-11T02:45:42+00:00一间屋子换了家具，人走进去看一眼，就知道该从哪里绕过去。 可机器人要适应这些变化，需要有办法获取环境信息。想到这里，就能理解，为什么教机器人认识现实世界，还需要不断采集真实场景的数据。 @vangrid_io 做的，就是让更多人用手机参与这件事。 Vangrid 面向 Physical
AI visual note: An advertisement by VANGRID with Chinese text '用手机，为 AI 采集真实空间' (Use your phone to capture real space for AI) showing a hand holding a smartphone that projects a blue 3D point cloud of the room.
· 2026-10-11T02:08:14+00:00It isn't gaming hardware if it doesn't have LEDs
· 2026-10-11T01:01:44+00:00embeddings are all you need
· 2026-10-11T02:00:01+00:00We are excited to be part of the MIX Fall Game Showcase! See you then, Timehackers! #slashzero #roguelike #action #gameplay #wishlist #demo #indie #TheMIX
· 2026-10-11T01:44:00+00:00my new airpods won't work and i called apple support and they kept "escalating" me to more and more senior staff and apparently at some point i yelled "oh are you connecting me to steve jobs now???"
AI visual note: Two women stand talking inside an Apple Store, with a 'one to one' promotional poster and other customers visible in the background.
· 2026-10-11T01:57:45+00:00As autoresearch agents run longer, do they keep exploring new approaches, or mostly refine the ones they found early? Turns out, runs settle into different "idea basins" and stay there. We introduce fork-and-flush, a simple way out
AI visual note: This image shows the title page of an academic paper titled "FORK-AND-FLUSH: ESCAPING IDEA BASINS IN AUTORESEARCH AGENTS" by authors including Ziyang Cai, Christos Ziakas, Vasilis Kontonis, Tim Pearce, Siddhartha Sen, Akshay Krishnamurthy, Shivam Garg, and Dimitris Papailiopoulos from Microsoft Research. The abstract explains that autoresearch agents often plateau at suboptimal solutions, with different runs settling into distinct "idea basins," and introduces the fork-and-flush method—periodically forking trajectories to escape these basins—which outperformed baseline methods by 66.0% and 44.4% across 13 research and engineering tasks. The title and abstract directly illustrate the accompanying post's point that autonomous research agents tend to get stuck in local "idea basins" rather than continuously exploring new approaches, presenting fork-and-flush as the proposed solution for escaping these thinking patterns.
· 2026-10-11T01:35:48+00:00AI is good at math but bad at art. The corollary is that writing is art, but music is math.
· 2026-10-11T02:36:44+00:00legends not just an influential eval but also a wonderful methodology! “formalizing the gap between what agents can do and what humans need” is 100% the most important skill right now!
AI visual note: An arXiv paper page titled 'SWE-bench: Can Language Models Resolve Real-World GitHub Issues?' by Carlos E. Jimenez et al., submitted on 10 Oct 2023 (v3).
· 2026-10-11T01:52:57+00:00You’re using Jev wrong. Without a decision graph, you pay a model call for every row, even the obvious ones. rulesmith has DSPy write the graph over your fields, so Jev only judges the hard cases. Job fraud, fine-tuned 4B judge: +6.7 pts accuracy, 29% of the calls.
AI visual note: A bar chart titled 'Same 4B judge, same 100 examples, a fraction of the calls' comparing accuracy and model calls per row between 'rulesmith' and 'DSPy typed classifier' for job postings and clothing reviews.
· 2026-10-11T02:06:32+00:00My agents have reached bottleneck
· 2026-10-11T02:17:54+00:00it was quite negligent for claude code to ship with such aggressive history selfcensorship. great loss to science.
· 2026-10-11T01:40:01+00:0050M token 的长记忆，关键不在让模型一次看完，而在别把看过的内容反复重算。 arxiv.org/abs/2610.10845 galahad-kv 将每约 16K token 的 KV 状态加密存入本地 NVMe，需要时再逐块恢复。单张 H100 上，恢复比重算快 2.8–4.3 倍，GPU 能耗低 8.8–12.3 倍。但这不是 50M token
AI visual note: The title page of an academic preprint by Sietse Schepke from Corbetic AI titled 'Real Long-Term Memory for AI: A 50-Million-Token Window That Is Faster and Cheaper Than Recompute' (arXiv:2610.10845v1, dated 7 Oct 2026), including the abstract describing an NVIDIA H100 memory layer system.
· 2026-10-11T01:07:26+00:00mistral large 4 just flopped the text arena. rank 115. after the code arena pareto post and the agent arena top 15, this one HURT europe's flagship is a code model wearing a chat model's clothes (via @dummerspast39 )
AI visual note: A leaderboard ranking AI services shows 'mistral-large-4' by Mistral Proprietary ranked at 115 with a score of 1429, highlighted by an orange box.
· 2026-10-11T01:51:10+00:00MEM OFF IN SAN FRANCISCO Compare different memory solutions or bring your own! @Claude_Memory @cognee_ @helixdb Next Saturday at the Pebblebed office
· 2026-10-11T02:40:23+00:00AI（Claude Code）にHoudiniとDaVinci Resolveを操作させて、20秒の映像を作成する簡単なスタディ Houdini：64種のオブジェクトをSDFで造形し、パーツがねじれたり変形したりする1秒ループを作成 DaVinci：スクリプトAPIで、64本の壁の上をカメラが旅する編集を組んでレンダー #Houdini
· 2026-10-11T01:56:24+00:00The more I spend on here the more I realise there is an issue with reading comprehension. Either bots haven’t mastered nuance, or people are being deliberately obtuse. The number of times someone posts something perfectly clear, only for the comments to invent a meaning that
· 2026-10-11T02:25:07+00:00guys you don't understand the microcenter had one on CLEARANCE open. box. I SAVED $1800 I *HAD* TO GET THE SECOND GPU!!!
· 2026-10-11T02:31:05+00:00This VFX took 3 years #RobloxDev #RobloxDevs #RealtimeVFX #RobloxVFX
· 2026-10-11T02:00:00+00:00“they don't have that context” @OfficialLoganK on why context matters for useful AI products—and what glasses could add. From our July 2025 interview.
· 2026-10-11T02:33:19+00:00I am hand-typing raw english human words and my google docs shows me this Lord save me
AI visual note: A simple graphic showing a text area containing the words 'This section resembles AI text' next to a toggle switch.
· 2026-10-11T02:51:35+00:00Good morning dear, the more capable robots become, the more important their view of the real world gets. a simulation can teach a robot plenty, but real streets, buildings and changing environments bring details that are hard to capture from internet data alone. that’s the
AI visual note: A VanGrid promotional graphic titled 'REALITY AS A SIMULATION' displays a silhouette of an airplane against a backdrop of digital code.
· 2026-10-11T02:32:00+00:00Facts by #JoeHansenFacts A flat Starlink dish follows satellites doing over 27,000 km/h, and the dish itself never moves. So how does it keep up? Under the cover is a big grid of tiny antennas, all driven by custom chips on one thin board. If every antenna sends at the exact
AI visual note: The illustration shows a stationary Starlink dish on the ground tracking fast-moving satellites in orbit, with annotations explaining that "beam swings" across the dish's antenna grid via "tiny delays tilt the beam," crediting "@joehansen" in the corner.
· 2026-10-11T01:22:07+00:00まだまだ販売できるクオリティではないけどやっと衣装試作ができた～ まだメッシュがいまいちな部分があるのとテクスチャとマテリアルをliltoonとか使って調整しなきゃ
· 2026-10-11T01:30:48+00:00Almost done with the user onboarding to the game. In the meantime here are some videos to check out to help understand the game. Once you get the gist of it, the game becomes easy to play. Still working on the native paid version.
· 2026-10-11T03:04:32+00:00strongly recommend you give yourself 30mins to watch this video on 2x this weekend. probably one of the most valuable videos out there for agentic coding rn
· 2026-10-11T02:27:38+00:00one thing i find interesting about @sleepagotchi is that it’s looking at a problem beyond just making AI smarter almost half of US users have tried AI, but only 25% use it daily and just 4.5% pay for major models like ChatGPT, according to the a16z study highlighted by Gotchi.
AI visual note: A colorful promotional banner for the '$CHI Community Sale on Sonar' by Gotchi Labs featuring a cute pink character wearing a space helmet against a sunset background.
· 2026-10-11T01:34:00+00:00Calling an LLM from Python takes a few lines of code but running it reliably in production is a completely different problem. A production inference system has to handle concurrent requests, GPU memory, scheduling, latency, failures, streaming, and observability. Let's follow a
· 2026-10-11T02:38:00+00:00The Internet Is FAR Bigger Than You Think — And This Visualization Proves It! Most people use the internet every day, but very few understand just how massive, complex, and interconnected it really is. Imagine millions of digital pages, websites, links, and hidden
· 2026-10-11T02:45:13+00:00tesla light and drone show at @tesla giga texas @DBurkland
· 2026-10-11T02:10:00+00:00How to build hardware in 2026: 1. Describe it
· 2026-10-11T02:53:45+00:00If you build MCP servers, it's time to adopt CIMD. I built a server on Cloudflare Workers to share the flow. You can copy the code, or point your agent at the repo as a reference. mainbranch.dev/cimd
· 2026-10-11T03:05:12.348000+00:00Astra vs Opus 5.5: is cost per token the wrong way to compare them for coding?
· 2026-10-10T16:33:57+00:00"The enshittification of all information is so unfortunate."
· 2026-10-11T02:30:00+00:00なんだこの可愛い動画は ChatGPTのdotに動画作ってもらった 巨大な敵に恐怖して震える動画にしたかったけど、敵も可愛すぎた笑 動画生成AIなしでこれが出来るの凄いですね！！
· 2026-10-11T02:30:00+00:00Microsoft Azure Agentic AI stack This is what enterprise AI agents actually need. Not just an LLM. Not just RAG. Not just a few tool calls. A production-ready agent needs an entire system around the model. Here is how the Microsoft Azure Agentic AI stack comes together: →
· 2026-10-11T02:15:28+00:00Anthropics gonna 23andMe everyone so bad post IPO.
· 2026-10-11T01:12:06+00:00xAI’s recent release wave has been insane Most people still think Grok is just a chatbot But over the last few weeks, xAI has been turning Grok into a full AI operating layer: • Grok Build early beta — an agentic CLI for coding, apps, and workflow #AI
· 2026-10-11T02:05:39+00:00我很认同把 Shelley 单独拿出来讲这件事。一个能力最初可能只是 exe 平台的一部分，但当它能脱离原来的工作流独立解决问题，就值得用独立产品的标准去打磨和检验。开源网页 agent
AI visual note: A mobile screenshot displays the Shelley website, describing 'A coding agent that lives in a browser tab' with installation instructions including 'brew install --cask boldsoftware/tap/shelley', showing version 'v0.1346.914604164' and 'Apache-2.0' license.
· 2026-10-11T01:54:11+00:00El anuncio del iPhone 18 Pro y su apertura variable, el primero en la historia de Apple; transmitido por Univision/TUDN USA
· 2026-10-11T01:16:02+00:00如果每次调用网站都要让 agent 操作浏览器，慢的不只是请求，还有整套页面交互。api-anything 会从网站前端流量中学习可复用操作，再通过 HTTP、CLI、TypeScript 或 MCP 调用；请求失效时尝试重新学习，并在回放成功后才保存修复结果。 github.com/goodnight000/a … 工程上值得看的是分层回退：先走普通
AI visual note: Bar chart comparing Agent + browser (Playwright MCP) vs Agent + API Anything (MCP) on Time to answer and Cost per task; 1 book: 16.8s vs 14.4s ($0.156 vs $0.048); 5 books: 182.5s vs 30.5s ($0.122 vs $0.107). Caption notes 5 runs, Claude Opus 5.5, Goodreads task.
· 2026-10-11T01:05:29+00:00過去作った作品を学習させてClaudeに一言番外編作ってお願いした実験結果 脚本、演出、背景画像やBGM生成、動画生成と編集まで全部AI任せの作品 #放課後ブルーステージ 番外編「雨上がりの予報」
· 2026-10-11T02:20:55+00:00One day, AI robots may be able to do almost everything we do. If robots work just like humans, who will protect them from being exploited? Drop your opinion in the comments! #HumanoidRobot #FutureOfAI #RobotLife #Robotics #AIVideo
· 2026-10-11T02:52:26+00:00本日、SUNOにアップ予定の曲「Fading Fire」 そのHook用動画です 曲リンクだけだと中々聞いてもらえないXの世界 色々挑戦してみたいですね 今回も、動画生成はOiiOii様( @OiiOii_AI_JP )のMV制作を使用しました！ AIエージェントとの会話でほぼ自動生成、ほんと便利です！
· 2026-10-11T01:53:26+00:00把路由从“让通用模型生成一段 JSON”改成“让专用模型对有限选项直接给概率”，确实抓住了 agent 系统里一个很实际的浪费点：分类、门控和分流不需要每次都生成自然语言，也不该把解析格式的可靠性押在提示词上。choice、yes/no 和有序 score 这几类接口足以覆盖不少入口判断；如果输入量大，省下的输出
AI visual note: A technical infographic titled 'SYSTEM ONE: THE RISE OF DECISION MODELS in Agent Pipelines' details an API engine using Ollama and Perplexity 27B, highlighting benefits like deterministic actions and ultra-cheap processing alongside adversarial risk warnings.
· 2026-10-11T02:31:31+00:00Qwen3.8-27B just hit approximately 381 tokens/sec on a single RTX 3090 with 24GB VRAM. Four days earlier, the same developer was getting around 82 tokens/sec on the same GPU. The progression tells an interesting story:
· 2026-10-11T02:00:00+00:00An agent asks. Decides. Pays. Within the limits you set.
· 2026-10-11T01:54:54+00:00Chỉ cần hai ký tự thôi mà mở ra cả một thế giới typosquatting trên trình duyệt Chromium. Nghe đơn giản nhưng đáng sợ thật sự. Anh em dùng Chrome hay Edge thì cũng nên cẩn thận với mấy cái tên miền nhìn na ná nhau. Gõ nhanh một chút là có thể bay vào trang giả
· 2026-10-11T03:00:03+00:00Mình để ý gần đây có một app nhắn tin kết hợp nhiệm vụ nhận kim cương, nhìn qua thì khá giống mô hình vừa trò chuyện vừa tích lũy thưởng. Thấy cũng tò mò nên chắc mình sẽ thử tìm hiểu xem cách hoạt động thế nào, kim cương có đổi được thành tiền thật
AI visual note: A Vietnamese chat conversation between two users, with messages in green and white bubbles including 'ok e', 'trước e ở hcm hơn 2 năm nè', 'thế à', 'dr a', 'quận mấy', 'e ở bên bình tân', and 'a cung ở đó'.
· 2026-10-11T01:48:54+00:00This is the life I want and will have one day Lord willing. Plus i can do all my ai work with a Starlink
· 2026-10-11T01:05:55+00:00把一张角色图拆成透明小图层这点挺戳我，后面想怎么拼动画都更自由了。
· 2026-10-11T02:41:32+00:00이제 굳이 모델들 써가며 광고할 필요가 없을듯요
· 2026-10-11T02:28:03+00:00Open Gemini, Grok or GPT Image 2.0 ② Upload your image ③ Paste the prompt ④ Generate your image and video Prompt given below
AI visual note: A collage-style anime illustration features two young women, one with dark hair and glasses and another blonde character, in a magical gothic setting with candles, potions, and a young male character in a robe; the upper scene shows the girls in a castle corridor while the foreground depicts an intimate close-up.
· 2026-10-11T02:44:18+00:00根据我的推文习惯和说话方式，画一张我的画像。 @grok
· 2026-10-11T02:12:10+00:00putting some more work into the divekick for my mod, i ended up reworking it into a tool for accessing double jump variants of some combos, coupled with a bit of slowmo as a grace period still not sure if it should hitstun grounded enemies or not, i'll tune it later
· 2026-10-11T02:05:53+00:00#Crealityk3 #toolchanger #3dprinting #launch
· 2026-10-11T01:31:06+00:00i built the Bones for this very thing at my house but i haven't deployed it because im a poor and my ahitty$488 dollar dev machine doesn't have enough ass to do the work
· 2026-10-11T02:42:53+00:00Got sticker shock reading the @socialcapital post below. So I dived a bit deeper and found that intelligence is cheap in the middle and expensive at the edge. Everyday AI work now costs pennies, with open and closed models side by side. The best few points of capability add about
· 2026-10-11T01:50:54+00:00I have open-sourced the web-based Forza Horizon race car games!! Welcome to try and make the modifications, can't wait for how the society puts the vast creativity on to it! github.com/Henry-fan88/ho …
· 2026-10-11T02:12:10+00:00Want to become an AI Engineer? Start with these 12 essentials. Being an AI Engineer isn't just about knowing how to call an LLM API. You need a strong foundation across: Python Math & Statistics Machine Learning Deep Learning Data Engineering NLP LLMs &
· 2026-10-11T03:00:04+00:00有人用 AI 开源了修图剪辑平替，不过先别急着退订 Adobe Photoshop、Premiere、Lightroom 这类软件常年靠订阅。开发者 Brandon Thomas 用 Claude Opus 5.5 加 Rust，开源了 7 个对标 Adobe 的应用，自称不抄源码、不收订阅费。 不过 TechRadar 实测还在 0.2 早期
AI visual note: A dark-themed dashboard titled "ArtCraft Discord" displays seven open-source applications built in Rust to replace Adobe software: PhotoCraft (#1), VectorCraft (#2), FilmCraft (#3), LightCraft (#4), PdfCraft (#5), EffectCraft (#6), and DesignCraft (#7), each marked as "EARLY ALPHA" or "IN DEVELOPMENT" with availability for macOS, Windows, Linux, and Web. The layout directly illustrates the post's claim about developer Brandon Thomas using Rust to build seven AI-powered, no-subscription alternatives to Photoshop, Premiere, Lightroom, and other Adobe products currently in early 0.2 alpha testing.
· 2026-10-11T02:12:45+00:00A tiger's steady stride becomes a trail of living moments. Each frame holds a little of its quiet power, turning motion into a shape we can walk around. 4dimension.app
· 2026-10-11T02:49:11+00:00Rays of light through the clouds #AI #PS #Adobe
![[caption-failed: image unfetchable or no caption]](https://pbs.twimg.com/media/HUUT-dbbsAAFD4M?format=jpg&name=small)
AI visual note: [caption-failed: image unfetchable or no caption]
· 2026-10-11T02:08:00+00:00You can power electronics sitting at 300,000 volts without a conductive power cable. DUNE sends laser light through optical fiber, then converts it back into electricity beside photon detectors submerged in liquid argon. Cryogenic converters reached up to 55%
AI visual note: A smiling man in a white shirt stands in a workshop or store behind a table displaying an open black electronic equipment case with circuit boards, copper coils, and wiring.
· 2026-10-11T01:34:03+00:00This drone flew simulated obstacle courses while looking on about 1% of its steps. ALONE (Qiyuan Lab, Tsinghua) keeps a 3D guess of the space around it, moves that guess with its own motion, and grabs a new depth image only when the guess is too shaky to act on. Play it free:
· 2026-10-11T02:42:06+00:00CLAUDE OPUS 5.5 JUST BUILT A CYBERPUNK AAA-STYLE GAME IN ONE SHOT. One prompt. Three.js. ZERO external assets. The result? A dimension-hopping cyberpunk adventure with: 2 semi-open-world dimensions Unique environments, buildings & characters Story-driven quests
· 2026-10-10T21:00:00+00:00Brain effort slider by @ozzyxs1a Find more tasteful designs on kombai.com/selects
· 2026-10-11T02:49:22+00:00gm ct this is probably the clearest way i've seen the idea explained so far a16z’s data is interesting because trying consumer ai and actually using it every day are two different things. getting someone to open a tool once is easy compared to giving them a reason to keep
AI visual note: A purple infographic reading 'The future of wellness is agentic AI.' featuring a central AI icon connected to 'WEB STREAM', 'HEALTH REALM', 'INSIGHTS', 'PERSONALISED ACTIONS', 'RELEVANT COMMERCE', and 'BETTER OUTCOMES'.
· 2026-10-11T02:42:29+00:00The Evolution will be Supercharged.
· 2026-10-11T01:53:15+00:00I built an AI phone + SMS assistant with Grok Bot in an afternoon . Tested and working, so here's the full setup. **What it does** - My own local number, answered live by Grok realtime voice: natural, fast, and it switches between English and Mandarin with the caller - Answers
AI visual note: A technical flowchart titled "AI Phone + SMS Assistant — Architecture" detailing a system using Telnyx, xAI Grok Voice, Cloudflare Worker, and a Grok Box for handling calls, with notes on outbound-only polling and security constraints.
· 2026-10-11T02:48:05+00:00The future looks like a dream… until AI brings it to life. A world where technology meets imagination, and every frame tells a story. This is the magic of AI creativity. #AI #AIVideo #AICreativity #FutureTech #ArtificialIntelligence
· 2026-10-11T02:59:45+00:00Good morning friends AI adoption is growing, but getting people to use it every day is a different challenge. That’s where Gotchi Labs has an interesting approach. Instead of relying on one AI that does a bit of everything, Gotchi is building around specific parts of daily
AI visual note: A thumbnail for 'The Top 100 Gen AI Consumer Apps — 7th Edition | Andreessen Horowitz,' featuring a futuristic cityscape and a sphere labeled 'TOP 100 CONSUMER AI APPS.'
· 2026-10-11T01:06:51+00:00Building a way to automatically track airplane takes offs near SFO and store their videos. Check it out
· 2026-10-11T02:59:50+00:00There are a few people on this app who are both brilliant and authentically attached to AI/SI, or to a particular system, or to the architecture itself. I used to assume that as the idea of AI as peers became more visible, a broader wave of hostility would follow. A hate wave is
· 2026-10-11T02:48:19+00:00SolveCraftの日本語版、Agyで1時間も掛からんかった。Gemini3.8Flash middleで、5時間リセットのクォータ消費は15%程度。Agyお得すぎ。半額プランで年契約したから月々千円ちょいだし
· 2026-10-11T02:30:00+00:00A livestream becomes different when the audience can actually influence the experience. @streamoors brings real time viewer interaction into AI-generated personalities rather than limiting them to prerecorded content. I’m curious to see which communities turn their AI
· 2026-10-11T02:07:10+00:00看到 @philkellr 分享的 Sonos 使用经验，和我们做 ZoneFoundry 时遇到的场景有些重叠。
· 2026-10-11T02:36:26+00:00Should I add this easter egg song for my daughter in my #indiegame ? Made this song with help of AI for her 2 years ago and she still sings it to this day. #girldad
· 2026-10-11T02:41:05+00:00You can also do motion transfer on tryriffi.com, something I recently implemented. I'm still getting a lot of rejections, especially when using celebrities or well-known characters. I'm honestly not sure how platforms like Higgsfield handle this, but I'm trying to
· 2026-10-11T02:07:08+00:00翻译从业者 Kaori 今天补充说明，自己的担忧还包括误译被接受、专业技能被忽视。 这让我想给 AI 摘要加一项检查：作者究竟在反对什么？压成“担心被替代”，容易丢掉对质量标准的讨论。
AI visual note: Screenshot of a Japanese news feed titled '本日のニュース' (Today's News) featuring three stories, with headlines including one about translator Kaori Myatt resigning due to AI evolution.
· 2026-10-11T01:51:34+00:00slow progress because ive been trying to figure out movement tech so i can design enemies around it
· 2026-10-11T02:37:47+00:00When cyber threats strike, quick action and smart recovery make all the difference. Stay prepared, stay protected. #CyberThreats #CyberRecovery #ThreatResponse #InformationSecurity #CyberAwareness
AI visual note: A LearnCollab promotional graphic titled 'CYBER AWARENESS STARTS WITH YOU' offering a 'Cyber Safety 101' course, featuring illustrated icons of computers, locks, and shields.
· 2026-10-11T03:00:54+00:00A machine one-shotting research-level math proofs makes this the greatest era in history to be an autodidact, and one of the most nerve-wracking to need a paycheck. Having an emotionless, 24/7 conversation partner ready to unpack graduate textbooks from your kitchen counter is
AI visual note: A SpongeBob SquarePants meme featuring Squidward looking through blinds, captioned 'AI cracking research-level math proofs / me calculating if I can afford rent.'
· 2026-10-11T02:45:00+00:00ELON SAYS HIS AI HAS BEEN GRINDING DIABLO. Musk says Digital Optimus can get roughly halfway through the campaign by reading the screen. He also reports progress in Counter-Strike, with League in training. The goal is broader game-playing ability. Interesting training ground
· 2026-10-11T02:24:00+00:00Day 37. Big maps grow villages now with a network of tracks and roads, villages grown along them, then the dunes around. I know, boring right ? back to shooting birds . #gamedev #games #indiegame #indiedev #solodev
· 2026-10-11T03:07:03+00:00This is how data quality should be handled
· 2026-10-11T03:01:02+00:00the sim to reality gap is still so huge
· 2026-10-11T03:06:19+00:00Stuffing more data in rarely fixes things, picking the right data does.
· 2026-10-11T02:50:25+00:00ai gets more valuable when it becomes part of daily habits, not just another chatbot
· 2026-10-11T02:59:55+00:00axis robotics is definitely onto something
· 2026-10-11T02:57:44+00:00Sleep as the first input for AI context sounds useful, but I worry about how much personal data gets pulled in.
· 2026-10-11T03:03:14+00:00Daily specific AI sounds way more useful than one do it all model, but integration is the real test.
· 2026-10-11T03:06:09+00:00thực sự thì trải nghiệm người dùng quan trọng hơn nhiều
· 2026-10-11T03:06:16+00:00blue check requirement kinda kills it for me ngl