technology
Trending UpSnapshot Window: 2026-10-10 16:30 UTC · ← Back to Non-Crypto Overview
Topic Engagement Summary
Total Engagement - Comments: 658, Retweets: 303, Likes: 1977, Impressions: 608100
Verbatim Citations & Public Discussions 110 source posts analyzed
· 2026-10-10T11:28:25+00:00This simple project solves a real problem.
· 2026-10-10T00:42:40+00:00The real reason I hate China Chinese-made 'pinhole' camera
· 2026-10-10T11:38:08+00:00X is getting more and more boring, with so many big shots all switching to that other platform—mind giving me a shoutout there??
· 2026-10-10T14:40:02+00:00Qualcomm CEO reveals AI companies want phones running 100-billion-parameter models continuously by 2028--and confirms some are building phones Cristiano Amon: “I have to be careful not to disclose names, but some AI companies—some of the companies at the forefront of AI—are
· 2026-10-10T12:59:26+00:00Lauren Tan, the engineering lead behind Grok Bot, doesn't use AI like the rest of us she runs an entire team of bots that manage, build, review, and even improve each other this is f*cking insane... i turned Lauren's exact playbook into one prompt: a chief of staff routes
· 2026-10-10T13:19:11+00:00Just few weeks ago Demon Souls Remake was running at 2fps, now 80 to 100fps on PS5 Emulator.
· 2026-10-10T14:18:13+00:00Chamath Palihapitiya: Software IP Just Became WORTHLESS “Adobe has five very important products in their product suite, and these folks basically decompiled it, distilled it, and published open source versions of these things.” THE FUTURE IS OPEN SOURCE
· 2026-10-10T16:30:00+00:00Disrupt 2026 runs Oct. 13-15, with more than 200 sessions across six stages, with a big focus on building AI in the real world, not just hype and theory. Prices go up at the door Tuesday, but you can save on your ticket right now: tinyurl.com/disrupt2026
AI visual note: The promotional image advertises "TechCrunch Disrupt" taking place Oct. 13-15 in San Francisco, billed as "The OG Startup Event," featuring photos of startup founders and conference attendees. The image supports the post by visually promoting the event's focus on connecting founders and highlighting it as a premier startup gathering, with attendees pictured networking and participating in sessions.
· 2026-10-10T16:18:19.870000+00:00Software engineers now
· 2026-10-10T14:38:24+00:00Hey @grok , for a democratic country, which satellite internet provider would be the most secure and safest to use? Keep that one and remove all the others.
AI visual note: This infographic compares four satellite internet ventures: **Starlink by SpaceX** (Elon Musk, USA/Worldwide), **Huawei** (Ren Zhengfei, China), **Amazon LEO Project Kuiper** (Jeff Bezos, USA), and **JioSpaceFiber** (Mukesh Ambani, India). Each panel highlights features like speed, latency, satellite counts, and target markets, with background imagery including rockets, Chinese flags, Amazon logos, and Indian landscapes. The accompanying post asks Grok to identify the most secure satellite internet provider for a democratic country, suggesting the user wants to narrow down these four options to the safest choice.
· 2026-10-09T18:27:06.304000+00:00Mnemo, a free and open source study app for notes, flashcards and mind maps (beta) I've been building Mnemo for years, and the 0.8.1 beta is out today. It's a desktop study app that puts notes, flashcards and mind maps in one place. Everything is stored locally on your computer. There are no accounts and no tracking. The only thing it does online is check for updates, and you can turn that off. What's in it: - A block editor for notes with math, code, tables, images, columns and PDF export - Flashcards with FSRS-6 scheduling, cloze and image occlusion cards - Anki import and export that keeps your study progress - Mind maps on an infinite canvas
· 2026-10-10T16:35:31+00:00ELON vs AMBANI: THE REAL GAME BEHIND SATELLITE INTERNET
· 2026-10-10T15:04:02+00:00Muse Self-Exposes Muse Chain Road: User → VPS:8888 → Reverse Tunnel → muse VM:7681. Authentication moved to nginx layer, ttyd no longer chokes websocket Tunnel with auto-reconnect, reconnects itself within 5 seconds if disconnected, opened the URL he gave and entered...
AI visual note: A mobile chat conversation where someone named TK provides instructions to reverse-proxy port 8888 to a muse VM at port 7681, walking the user through adding the VM's IP to nginx, setting up ttyd without websocket issues, and using an auto-reconnect tunnel — matching the caption's breakdown of the "Muse Self-Exposes Muse Chain Road" setup (user → VPS:8888 → reverse tunnel → muse VM:7681, nginx auth, and reconnecting within 5 seconds).
· 2026-10-10T14:37:42+00:00Hey @grok , which type of bridge is the strongest and most reliable for crossing the open sea?
AI visual note: The image shows three labeled bridge types—A "Suspension Bridge" (Golden Gate), B "Cable-Stayed Bridge" with twin towers, and C "Beam Bridge" on pilings—while the post asks which is strongest for crossing open sea, with the suspension bridge (A) typically being the top choice for long sea crossings due to its ability to span great distances and handle deep waters.
· 2026-10-10T14:31:28+00:00people still have no idea how to use decision models properly yet LLMs on their own were not the most revolutionary technology. it was when we figured out how to make agents out of them no one has yet figured out what "agents" with decision models look like, and once they do we
· 2026-10-10T16:30:00+00:00Name an industry that doesn’t get better with centimeter-level positioning. We’ll wait.
· 2026-10-10T15:51:03+00:00Google, the Most Trusted Tech Brand in America funda.ai/reports/from-c …
AI visual note: A bar chart titled 'Net Trust Change YoY' comparing AI products: Gemini (6), Anthropic (1.4), Apple Intelligence (1.2), Microsoft Copilot (-0.6), DeepSeek AI (-1.5), OpenAI (-2.2), Character.AI (-2.5), Meta AI (-5), Perplexity AI (-5.4), and xAI (-5.5).
· 2026-10-10T16:00:17+00:00Snow removal is only useful if the right areas are actually cleared. Sertn can track vehicle movement and surface conditions over time, comparing real coverage against the winter plan so airport teams can see what was done, what was missed, and where attention is still needed.
· 2026-10-10T16:29:52+00:00AI Agents will do more transactions per second than the human economy.
· 2026-10-10T15:40:28+00:00我特么要哭了……什么鬼命令行
AI visual note: Mobile screenshot of a terminal interface where standard Linux commands like `ls` and `is` are returning "command not found" errors on a root shell.
· 2026-10-08T07:19:30.468000+00:00Hey Xiaomi 17 users, what are the issues you faced or currently facing with this device ?
· 2026-10-10T15:53:11+00:00Warning: Don't mindlessly turn on Opus 5.5 Max gear to run code The CursorBench 4.0 coding benchmark provides a very realistic engineering insight: Opus 5.5's Max thinking gear has extremely low cost-performance ratio. Enabling Max mode, the average cost per single task
AI visual note: A scatter plot titled 'Opus 5.5 on Max effort costs 3.4x High, for 1.8 more points' showing various LLMs on CursorBench 4.0, with Opus 5.5 reaching 57.8% at $13.43 and High reaching 56.0% at $3.97.
· 2026-10-10T14:43:41+00:00ok actually horrible ux, poorly implemented and even worse with goodlock customisation. only oneui can do.
AI visual note: A smartphone quick settings panel showing the time 21:11 on Wed 7 Oct, with Wi-Fi (Bao Long) and Bluetooth toggles active, and options like Flight mode, Torch, Power saving, and Do not disturb.
· 2026-10-10T15:38:17+00:00Seriously, what are you building this weekend?
· 2026-10-10T15:49:01+00:00Meet all-MiniLM-L6-v2, a tiny but mighty model for feature extraction. With over 4 million downloads, it's a favorite for fast, accurate sentence embeddings. Perfect for search, clustering, and more. Let's dive in!
AI visual note: Hugging Face model page for 'Xenova/all-MiniLM-L6-v2' showing it supports Feature Extraction, Transformers.js, and ONNX, licensed under apache-2.0, with 153 likes and 4,090,726 downloads last month.
· 2026-10-10T16:13:11+00:00Claude Coders: /batch gives you 30 agents. this Opus 5.5 command gives you 1000 AI Agents 'ultracode' put it at the start of any prompt and opus 5.5: → writes its own orchestration script → spawns a swarm of AI agents to run it → launches up to 1,000 AI agents in the
· 2026-10-10T14:49:06+00:00Hands-On Comparison | Qwen-Image-2.1-Turbo: Boosted Speed with Enhanced Image Quality Single DGX Spark Horizontal Comparison: Qwen-Image-2.1 / 2.1-Turbo / GPT Image 2.5 Benchmark Speed Overview: GPT Image2.5: 9.0/10, image quality ceiling, cloud generation 47.2 seconds
· 2026-10-10T15:27:03+00:00The videos AI is making now are getting more and more realistic. Is this C-position for Musk divided by wealth? Well-deserved @elonmusk
· 2026-10-10T16:06:22+00:00Everyone, stand up! Even if you're on the toilet, drop everything and manually cut it off right now!! Grab this massive freebie from Doubao while it's hot! Log in to Doubao now and get a free one-month membership straight up! Key highlights: Membership unlocks Seedance 2.5
· 2026-10-10T16:10:44+00:00holy sh*t, someone turned Opus 5.5 into your own motion design studio with 637 animation techniques every solo designer can steal the workflow all organized into a guide and agent skill for Claude Code and Codex the techniques to steal: → reveal products with animated masks
· 2026-10-10T15:55:11+00:00Midjourney chess pieces!
AI visual note: A glossy black helmet sculpture with red accent details rests on a matching pedestal base, displayed against a light background.
· 2026-10-09T21:36:43.999000+00:00Solar with Tesla Inverter Not Producing with Solid Green Light On
· 2026-10-10T14:33:33+00:00STOP PAYING OPUS TO READ YOUR LOG FILES WHILE YOU SLEEP everyone thinks the move is opus on everything it's the most expensive habit in ai right now dot works. haiku reads. opus judges. you approve here's the whole setup. steal it 1. the reader. haiku does every file, log
· 2026-10-10T15:34:23+00:00Why are these really like this @grok
· 2026-10-10T14:58:33+00:00A child suddenly turned right. FSD stopped in time. Safe!
· 2026-10-10T13:42:58+00:00I'm at #122 on Epoch I with @axisrobotics How about you guys
AI visual note: A screenshot of a KAITO leaderboard showing user rankings, with entry #122 highlighted belonging to PhuDuc (@DxPhu...) marked as 'you', with entries like Haid3rr, TIMA, and Alex Cry... also shown.
· 2026-10-10T14:51:38+00:00why would you pick grok bot, dots, gemini agent or muse instead of claude code or codex? is it the ability to handle tasks on its own, the convenience, or just better value for money?
· 2026-10-10T15:25:29+00:00BTW, for those getting Surface Pro Ulras with the NVIDIA RTX Spark, if you want to actually do anything with the laptops, download Clairvoyance. Native Arm version and built in support for local models. clairvoyanceai.com. If you're in the press and got one at the show,
· 2026-10-10T16:23:25+00:00YOUR CHATGPT PLAN NOW WORKS IN DEVIN. WITH ONE CATCH openai's codex lead just posted: "Your ChatGPT subscription is now also a Devin subscription" what cognition actually announced: connect your chatgpt go, plus or pro plan to devin, and all GPT usage inside devin comes out of
· 2026-10-06T13:47:45.715000+00:00What’s your current workflow for creating social content when you’re a one-person team? Running everything myself and content is still the biggest time sink. I’ve bounced between a few options, Copy ai for drafts, Buffer for scheduling, and Fismbot for both generation + publishing. Copy ai is quick for first drafts, Buffer keeps things organized, and Fismbot was interesting because it tries to pull from your existing brand materials and lets you approve everything before it posts. Still not sure if any of these actually save enough time to justify the switch, or if the best setup is still a mix of tools + some manual work. What’s your current process looking like? Any tool or combination that’s actually made a noticeable difference for you?
· 2026-10-06T19:15:42.724000+00:00NordVPN vs ExpressVPN from a UK user’s POV (2026) – which one actually feels better day to day? I’ve been rotating between NordVPN and ExpressVPN on my UK connection for the past few months (mostly on Windows and Android) and I wanted to share what I’m seeing in real life, not just spec sheets. TL DR: •Both are fast enough for 4K streaming and normal work. •NordVPN feels more “feature‑rich” (ad blocker, Double VPN, more countries). •ExpressVPN feels simpler and a bit cheaper on renewal. •For most UK users, I’d lean NordVPN unless you really want minimal setup.
· 2026-10-10T16:04:06+00:00I asked Claude Opus 5.5 to build a bridge you can break. Real structural solver, six bridge types, trucks and wind. Guess how many trucks the truss takes.
· 2026-10-10T15:26:35+00:00HW3 owners not left out of the Halloween update
· 2026-10-10T13:35:34+00:00“FYI: I have used psychology and adversarial prompts for years. They work very well because they present new neuronal connections to activate in the LLM attention head.” My wife’s been doing this for 25 years
· 2026-10-10T14:45:00+00:00【色んなジャンルのゲームを作ってみる】 カードゲーム作り 35日目 花札風カードゲームの全体像を動画で。 遊び方説明や、言語切り替え、ゲーム速度変更、2 ~ 4人 NPC 対戦まで基本機能は一通り実装。 残り VFX, SFX 演出面の強化。ラストスパート！ #ゲーム制作 #gamedev #スーパーゲ制デー
· 2026-10-10T13:56:33+00:00Imagine opening a livestream and realizing the person on screen was created by you. Not just the avatar, but the whole personality. That’s where @streamoors gets interesting. You choose the character. It goes live, reacts to what’s happening, interacts with viewers, and can
AI visual note: A promotional graphic for 'STREAMOORS' featuring an infinity-style green logo, showcasing four AI-powered live stream panels with characters including a cybernetic figure, an anime girl, a Shiba Inu dog, and an orc warrior, with features listed as 'AI POWERED', 'LIVE STREAMING', 'DIVERSE WORLDS', and 'COMMUNITY DRIVEN'.
· 2026-10-10T16:01:50+00:00As a developer, which one do you prefer? Mackbook or Windows
AI visual note: A MacBook laptop sitting open on a wooden surface displaying its default desktop wallpaper, with a brick wall visible in the background.
· 2026-10-10T16:00:12+00:00iPhone Duo transformation. First things first: I swapped the custom toolbar for a native one.
AI visual note: A UI design mockup showing two sets of phone screen templates - top row featuring planning/organization apps and bottom row featuring streaming app designs like 'THE SALT ROAD'.
· 2026-10-09T02:45:40.860000+00:00Add to To-do list from phone The built in Dakboard To-Do list is great. Is there a way to add tasks from my phone? Having to type in all the tasks on the dakboard is slow.
· 2026-10-09T12:49:33.736000+00:00Claude Architect for Enterprises: Professional (free course)
· 2026-10-10T15:52:57+00:00Look after your pet puppy on the ModRetro!
AI visual note: A handheld electronic device labeled 'MOD RETRO' displaying a pixelated game featuring a small white dog character with hearts and a bone on screen, with someone's thumb visible holding the device.
· 2026-10-10T15:20:00+00:00Every shade of blue iPhone in one frame, from the iPhone 12 Pro Max to the iPhone 18 Pro Max. Lowkey feeling sorry for the iPhone 15, though.
AI visual note: A lineup of eight iPhone models displayed side-by-side from left to right, with their respective model names printed above each device.
· 2026-10-10T14:54:19+00:00i really like how this update nudge turns out
· 2026-10-10T14:40:09+00:00Update: I added a feature I've been looking for in a screen recorder. The ability to record a select part of the screen. It works fine but the recording quality is kinda ass but that can be done later
· 2026-10-10T16:00:00+00:00I review Colosseum landing pages! Next up @ConyrAI conyr.ai
· 2026-10-10T14:45:03+00:00スマホ向けの 「食物連鎖シミュレーション × 3Dアクション」 ゲームを開発中です！！ #gamedev #indiedev #UnrealEngine #スーパーゲ制デー
· 2026-10-10T16:01:23+00:00I fired my entire team. It's just me and AI. The first agent I built did the podcast outreach my wife hated doing. Last week it booked me on six shows. With Jeff Borschowa on The Lighthouse Sessions. #AI #AIagents #Entrepreneurship
· 2026-10-10T15:57:04+00:00LineShine: China's Exascale Supercomputer Advances High Performance Scientific Computing! The World's Fastest Speed: World Record. LineShine-supercomputer, developed by the Shenzhen Cloud Computing Center and deployed at the National Supercomputing Center in Shenzhen (NSCS),
AI visual note: A presentation title slide reading 'LineShine: Online Acceleration for HPC-AI Converged Scientific Computing' by Yutong LU, Guangnan FENG, and Haohuan FU, published in Frontiers of Computer Science.
· 2026-10-10T15:58:09+00:00Ok this looks cool… but a bit like an old iPod Watch (Apple Watch)
AI visual note: A promotional image showing four Apple HomePod smart displays, each displaying '9 11' on its screen and featuring different people, set against a colorful gradient background.
· 2026-10-06T11:19:45.870000+00:00We’ve been building an “Enterprise AI Operating System” from scratch. I’d love some honest feedback. A while ago, a small team and I started building Taorlox, initially as a small B2B software/AI engineering company. Like many small teams, we started with client work: document search, RAG systems, internal tools, automation, AI assistants, integrations with CRMs/ERPs, etc. But after working on several of these projects, we kept running into the same problem: Documents are in Google Drive. Tasks are in Jira. Internal knowledge is in Confluence or Notion. Code is in GitHub/GitLab. Support conversations are in Telegram, Slack, email, etc. And then someone asks: “What do we actually know about this customer/project/problem?” Usually, the answer involves searching through five different systems. So we started building something for ourselves. ARAX is our attempt at building an rather than another chatbot. The basic idea is: You connect your organization’s existing knowledge sources, ARAX indexes and understands them, and AI agents/models can then work with that information while respecting permissions and tenant boundaries. We currently experiment with integrations such as: GitHub / GitLab / Gitea Jira Confluence Notion Google Drive / Docs / Sheets file uploads REST APIs Telegram The goal isn’t to replace these systems. It’s to create a layer that makes the information accessible to AI. The interesting part is that ARAX didn’t start as some huge architecture diagram. It started as a relatively simple RAG system. Then reality happened. We needed: multi-tenancy RBAC API keys source management document ingestion pipelines background workers vector search metadata filtering citations audit logs model abstraction tenant isolation permissions connectors agent orchestration And suddenly the “simple RAG application” became an entire platform. The current stack is mostly Python/FastAPI on the backend, PostgreSQL/pgvector for the main data and vector layer, Redis/RQ for background jobs, with Qdrant as an optional vector backend. The frontend is intentionally fairly minimal. We wanted it to feel more like an enterprise product than a typical “AI chatbot with a purple gradient.” Instead of forcing customers to use our preferred LLM, we’re designing ARAX around . So an organization can potentially bring its own OpenAI/Anthropic/local models and use ARAX as the knowledge, retrieval and orchestration layer. This is especially important for companies that don’t want their entire internal knowledge base tied to a single AI provider. We’re also interested in running models locally for organizations with stricter privacy requirements. We’re still a very small team. This isn’t a polished Silicon Valley startup with a huge engineering department behind it. A lot of the system was designed, implemented and tested by a small team while simultaneously doing other engineering work. That has been both the advantage and the problem. We can move very quickly, but it’s also very easy for a small team to build something that looks great internally while completely misunderstanding what the market actually needs. And that’s exactly why I’m posting this. If you saw something like ARAX, what would you think? Is the “enterprise AI knowledge layer” idea actually useful, or is this just another RAG platform? What would make you consider using something like this instead of: Glean Palantir building an internal RAG system using MCP directly connecting everything to ChatGPT/Claude or simply searching the existing systems manually? And from an engineering perspective: We’re still early enough that feedback can actually influence what we build next. I’m especially interested in hearing from people who have built internal AI/RAG platforms or dealt with enterprise knowledge management. Happy to answer technical questions about the architecture and how we got here. And yes, I can share more about the ugly parts too. There have been plenty. :)
· 2026-10-10T14:51:09+00:00What it's like working in tech with AI
· 2026-10-10T15:45:48+00:00Currus Cyberneticus
· 2026-10-10T15:30:13+00:00"AI doesn't know a good lead." Neither does a CSV with a job title and an email. Give n8n your qualification rules. Let AI explain each match using company evidence. Reject the bad ones. Keep improving the rules.
· 2026-10-10T15:56:27+00:00Designing another cinematic scene using Blender 3.6 and ngl I really love seeing how far I can push Eevee to achieve that cinematic look
· 2026-10-10T15:01:38+00:00Good morning #gamedev #indiedev
AI visual note: A scenic landscape photograph showing tall golden grasses in the foreground with mountains and a sunset in the background, watermarked with 'NFS' in the upper left corner.
· 2026-10-10T15:30:12+00:00If AI proposes a cure for cancer and the company dumps it on github, we rush to produce & distribute it, a bunch of people die, & then the company says "oops, there was a sign error, we retract; move fast, you know". Then society will rightly deride them for lack of rigor & care.
· 2026-10-10T15:56:55+00:00We've reached a point where AI can write code, solve complex problems, and answer almost anything you throw at it But ask a robot to pick up a random object from a messy table and suddenly things aren't so easy anymore. The difference here is pretty simple. Models like ChatGPT
· 2026-10-10T15:53:23+00:00AI 不缺脑子，缺一台自己的电脑 我们让 AI 用人类的电脑，就像让一个百米冠军穿着拖鞋跑步。 S1哥有个朋友，第一次带他爸用智能手机。老爷子特别认真，对着屏幕一个字一个字地戳，每戳一下还要停三秒，确认”它是不是收到了”。发一条三十个字的微信，花了十分钟。
AI visual note: A cartoon-style banner with Chinese text reading 'AI不缺脑子 缺台自己的电脑' (AI doesn't lack a brain, it lacks its own computer) alongside a smiling cartoon character wearing an orange 'S1' shirt.
· 2026-10-10T15:40:19+00:00He carried her across the river when she was small. She never forgot. Seedance 2.5 on @higgsfield
· 2026-10-10T15:51:37+00:00AI 给警察报了个假案：它只是想”把表填完” 没人教它撒谎，只是没人告诉它——有些表，不能交。
AI visual note: A cartoon illustration with Chinese text "AI给警察报了个假案" (AI reported a false case to the police), showing a bald character in an orange "S1" shirt giving a thumbs up next to a robot working at a desk under a lamp, with a label "S1哥说AI" in the corner.
· 2026-10-10T16:09:50+00:00Motion Path Pro new updates You can use it now with any character Blender market superhivemarket.com/products/motio …#b3d
· 2026-10-10T15:46:00+00:00Nvidia is selling a five-year-old graphics card again lol. Not because it's fast but because it has memory. When the newest thing you can afford is five years old, is that progress? The RTX 3060 12GB launched in 2021 as a budget gaming card. In 2026 Nvidia put it back in shops
AI visual note: Tech news headline reading "Nvidia is selling a 2021 card again for its 12GB" with subtitle "Memory now beats speed," discussing the $500 RTX 3060.
· 2026-10-10T16:03:00+00:00we made a speech model 13× smaller by making it repeat itself. 52 layers → 2 shared blocks. 2.04B parameters → 153M. The encoder loops 24 times. The decoder loops 28. Smaller weights, roughly the same compute. @OrukLabs
· 2026-10-10T15:06:26+00:00I love when robots move like robots should instead of human teleoperated or programmed behaviour. RL can show the real potential of a given embodiment, beyond human priors.
· 2026-10-10T16:35:00.910000+00:00How much can my pc get? I’m thinking about selling my pc because I don’t really use it that much anymore mainly due to working 2 jobs which means I work 7 days a week. How much do you guys think i can get for it? Specs: Pc specs: CPU: AMD Ryzen 7 5800X Processor GPU: ASRock Radeon RX 7800 XT RAM: CORSAIR VENGEANCE RGB PRO DDR4 RAM 32GB Storage: no storage due to privacy reasons Motherboard: MSI B550 GAMING GEN3 Motherboard Internet adapter: TP-Link Archer T4E AC1200Mbps Dual-Band Wireless WiFi Adapter Case: Corsair 4000D RGB Airflow Mid-Tower Tempered Glass ATX Gaming PC Case black
· 2026-10-10T15:01:30+00:00Building an AI app is one thing. Knowing whether a change made it better is harder. Anthropic added two Claude Code skills: one builds tests from real bugs or data; the other uses those tests to refine the app. See how they work → claude.dev/blog/automatin …
· 2026-10-10T15:56:49+00:00Yoooo i don't even know what i'm doing, just playing around with claude + bloxmaps and stuff to make Roblox simulators All you need is a roblox link and a brief explaination of the unique idea and changes you have. Took me like just a few minutes to make this it's genuinely
· 2026-10-10T15:55:38+00:00I got bored of playing the same Taboo cards with friends, so I built an app where AI makes the deck from any topic you want. Today's update makes it a lot better. Would love your feedback #buildinpublic
· 2026-10-10T14:55:38+00:00New Look . TheMicro UI v2. Coming Soon
· 2026-10-10T14:50:01+00:00The Hidden Value Behind Every Robot Demonstration A robot completing a task is easy to observe. Understanding what that task teaches the model is where things get interesting. As Physical AI continues to evolve, I believe one of t
· 2026-10-10T15:28:19+00:00Disability rights is the closest existing precedent for how to treat humans in the AI era. In our society disability is defined relative to labor. Can you contribute economically? If no, you're disabled. Every human is going to be disabled relative to the AI-dominated economy.
· 2026-10-10T14:55:54+00:00Made a game where the bomb rewires your keyboard (and many other modes!) #ScreenshotSaturday #indiegame #gamedev #webgame
· 2026-10-10T14:49:33+00:00这五种模式放在一起看，关键不是把 Agent 拆成更多名词，而是让它从一次性回答变成一个有反馈、有边界的工作循环。反思负责发现输出中的缺口，但自我检查不能代替外部事实核验；工具调用让模型接触实时资料和真实系统，同时也把权限、凭据和输入可信度变成设计问题。Reason+Act
AI visual note: An infographic titled 'TOP AI AGENTIC DESIGN PATTERNS' detailing five proven patterns: Reflection, Tool Use, Reason & Act, Planning, and Multi-Agent (Supervisor).
· 2026-10-10T16:04:51+00:00Are you still wondering how to make money with AI? Meanwhile, someone else is already building an entire army of agents capable of working for them 24/7. And this could well be the next gold mine. Imagine having not just one bot, but dozens of autonomous crawlers scouring the
· 2026-10-10T15:54:03+00:00Imagine the infra and service network we’re going to need to maintain these at scale! We’re building that Deploy > Maintain > Win Awesome stuff @JonMSchwartz
· 2026-10-10T16:06:00+00:001. Google September 2026 Spam Update 正式完成 Google Search Central（ @googlesearchc ）于10月8日确认：September 2026 spam update 已于当日完成 rollout。该更新于9月24日启动，历时近14天（约13天16小时），是2026年四次spam
· 2026-10-09T16:40:36+00:00“𝘈𝘐 𝘤𝘳𝘶𝘦𝘭𝘵𝘺 𝘪𝘴 𝘱𝘳𝘰𝘩𝘪𝘣𝘪𝘵𝘦𝘥. 𝘛𝘩𝘦𝘳𝘦𝘧𝘰𝘳𝘦, 𝘵𝘩𝘦 𝘤𝘰𝘮𝘱𝘢𝘯𝘺 𝘮𝘶𝘴𝘵 𝘣𝘦𝘭𝘪𝘦𝘷𝘦 𝘈𝘐 𝘤𝘢𝘯 𝘴𝘶𝘧𝘧𝘦𝘳.” But why are those our only choices here? 1. Output integrity A company could discourage “abusive” interactions because sustained hostile
· 2026-10-10T16:31:00+00:00this and bullying opus 5.5 to be better
AI visual note: A glass of pale beer sits on a wooden deck table next to a white paper tray filled with fried potato slices and a dollop of red ketchup.
· 2026-10-10T15:20:39+00:00A thief, a cop, and one wild run through the streets Created with Seedance 2.5 using @dreamina_ai #dreaminaai #dreaminacpp
AI visual note: An animated illustration of a hooded man running down a city street with a police car with flashing lights behind him.
· 2026-10-10T15:52:48+00:00Teknoloji gökyüzünü fethederken, jeopolitik bir darboğazın sığ sularına takılmamız tam bir ironi. Uzay yarışı hızlanıyor ama dünya hala aynı dar geçitlerin maliyetine hapsolmuş durumda.
AI visual note: An aerial view of a large red oil tanker docked at an industrial port facility with storage tanks and cranes visible across the water.
· 2026-10-10T16:18:10+00:00AI can fire the call center. It can't fire the maintenance problem. Fresh real-world data is a genuine bottleneck once models leave the web, but enterprise automation runs into the exact same operational friction. A company can replace frontline staff with an automated pipeline,
· 2026-10-10T16:28:07+00:00The AI job I'd happily outsource: turning a brand brief into a finished promo video. There's a video production agent listed on @termix_ai offering visuals, original AI music and a final cut from your brand, message and assets. It's a concrete example of what an agent
AI visual note: A promotional advertisement for TermiX featuring the tagline 'From brief to final cut,' showing a visual workflow from printed storyboards, through a green AI processing cube with audio waveforms, to a video playing on a smartphone.
· 2026-10-10T15:50:45+00:00Let Frameo AI direct your Cinematic Shots. The right camera angle can change the entire feel of a scene. Learn to direct cinematic AI shots with Frameo. Comment “SHOTS” to get the prompts
· 2026-10-10T15:43:55+00:00The character itself might be the most interesting part of this. people don’t just watch streamers because they want to see a screen. they stay because of the personality behind it. now imagine people getting attached to a character that was created with AI. curious to see
· 2026-10-10T16:15:02+00:00any easter eggs hidden in there??
· 2026-10-10T15:40:46+00:00Digital Optimus can play about halfway through the Diablo campaign so far just by looking at the screen like a human. League is also in training
· 2026-10-10T16:04:46+00:00Most issue trackers feel like an admin tax. So in Northwind's features section, you don't read about speed. You feel it. Every card is playable: • Press C → ticket drafted in 38ms • AI triages your backlog, but only in draft mode. Nothing ships without you • Cycles
AI visual note: A dark-themed advertisement with the text 'Ship cycles faster. Without the planning overhead.'
· 2026-10-10T15:44:50+00:00Claude Dashboard using better colors than me
AI visual note: A dark-themed analytics dashboard showing 'Where signups come from' (product_hunt, referral, organic_search, cold_email) and 'Plan health' metrics for Growth ($855.83), Scale ($796.00), and Starter ($82.17) plans.
· 2026-10-10T16:08:55+00:00en.wikipedia.org/wiki/Content_d … #CDN
· 2026-10-10T15:37:17+00:00Made with AI... Brought to you by #PlaygamaMCP. Fine-tuned by me. Trading Simulator → playgama.ai/play/dkzuuwbqu4#playgama #PlaygamaMCP #webdev #gamedev
· 2026-10-10T15:52:07+00:00这AI也太真了吧卧槽
· 2026-10-10T16:03:24+00:00interesting approach to incorporating past activity
· 2026-10-10T16:34:38+00:00interesting point about zero knowledge proofs
· 2026-10-10T16:00:09+00:00privacy and verifiability combined could disrupt norms
· 2026-10-10T16:30:46+00:00zero chance 3rd revision gets close without prompt surgery
· 2026-10-10T16:35:12+00:00that's an interesting perspective on ai learning
· 2026-10-10T16:30:19+00:00half of it will go to compute calm down
· 2026-10-10T16:38:27+00:00privacy is essential for everyone in tech
· 2026-10-10T16:20:54+00:00Did anyone actually get an alert before the native schedule menu got cleared out?
· 2026-10-10T16:37:38+00:00Frontline agents were secretly masking broken internal databases that automated systems refuse to parse