What people in AI, tech and science are talking about today.
26 October 2026 · Top 10
Ranked by how fast each item is picking up attention on X, Hacker News, GitHub and Hugging Face, with a short note on why it matters.
-
IBM's CEO says open-weight models now win on-prem
On IBM's Q3 earnings call, Arvind Krishna said open-weight models are now about 90% as good as frontier models at 10–25% of the token cost, and fit in two or three racks. He expects most enterprise tasks to move to the cheaper option.
Source: Hacker News · discussionLink · Post on X · LinkedIn ·▲ +40 pts/h164 points · 94 comments
LLMs -
A 10-year-old CNN beats frontier LLMs at spotting skin cancer
A Guy's Hospital dermatologist tested 200 biopsy-proven lesions: a fine-tuned 2015 Inception v3 model got 62% right, versus 42% for GPT-6 Astra and Claude Fable 5.1. Small specialised models still beat general chatbots at medical imaging.
Source: Hacker News · discussionLink · Post on X · LinkedIn ·▲▲ +46 pts/h516 points · 230 comments
LLMs -
NVIDIA's GTC Berlin pitch: Europe's own AI economy
Jensen Huang used GTC Berlin to push AI factories, agents, robotics and healthcare AI built in Europe. It signals NVIDIA courting European sovereign-AI spending.
Source: X · @nvidiaLink · Post on X · LinkedIn ·▲▲ +62 eng/h629 likes · 77 reposts
AI & Agents -
Chandra finds strange X-ray objects hiding in 20-year-old data
Anton Petrov walks through a Nature Astronomy paper on new cosmic X-ray sources spotted in archival Chandra observations. Old telescope data keeps turning up new physics.
Source: YouTube · Anton PetrovLink · Post on X · LinkedIn ·▲ +7,044 views/h64,215 views · 2,771 ratings
Science -
How to cut a triangle into a square
Numberphile's James Grime shows the classic dissection that turns an equilateral triangle into a square in four pieces, a puzzle from the early 1900s.
Source: YouTube · NumberphileLink · Post on X · LinkedIn ·▲ +9,535 views/h156,194 views · 3,680 ratings
Science -
Why OpenAI and Anthropic cut LLM prices by up to 80%
Max Woolf charts months of tit-for-tat price cuts, from GPT-5.6 Luna's 80% drop to Opus 5.5 and GPT-6.1 Sol, and argues cheap open-weight models served by dozens of providers are forcing the labs' hand. Agent coding costs have fallen sharply as a result.
Source: Hacker News · discussionLink · Post on X · LinkedIn ·▲ +39 pts/h252 points · 273 comments
LLMs -
GitHub's secret scanning gets faster and less noisy
GitHub says a new classifier raised acceptance of secret-scanning alerts by 11.5% and cut median scan time by 86%. Fewer false alarms means leaked keys get fixed faster.
Source: X · @githubLink · Post on X · LinkedIn ·– +18 eng/h98 likes · 8 reposts
Dev Tools -
A planted markdown file can hijack coding agents
A security firm reports a real attack where a malicious site dropped a markdown file into a Downloads folder, and an AI coding agent started there followed its hidden instructions. Where you launch an agent matters.
Source: X · @choblin29Link · Post on X · LinkedIn ·– +13 eng/h61 likes · 1 reposts
AI & Agents -
PaperPilot: a Rust PDF engine that runs locally, with MCP
PaperPilot is an open-source, pure-Rust PDF engine that compiles to WebAssembly and exposes an MCP server, so agents can work with PDFs without a cloud service.
Source: Hacker News · discussionLink · Post on X · LinkedIn ·– +24 pts/h78 points · 18 comments
AI & Agents -
An agent skill for genealogy research across the former USSR
This open-source skill for Codex and Claude Code searches church books, civil, military and repression records for family history, and passed 590 GitHub stars in a day. Agent skills are spreading well beyond coding.
▲▲ +28 stars/h593 stars in 1d
AI & Agents
-
CVEs Are Going TO THE MOON
New video from David Shapiro, 11,194 views so far.
Source: YouTube · David ShapiroLink · Post on X · LinkedIn ·– +585 views/h11,194 views
AI & Agents -
Kimi K3.1 Passes the Rubik's Cube Test
New video from Matthew Berman, 1,893 views so far.
Source: YouTube · Matthew BermanLink · Post on X · LinkedIn ·– +219 views/h1,893 views · 19 ratings
AI & Agents -
Anthropic Is Living In The Future
New video from David Shapiro, 42,104 views so far.
Source: YouTube · David ShapiroLink · Post on X · LinkedIn ·– +976 views/h42,104 views
AI & Agents -
waiting for my agents, game time
From the source: waiting for my agents, game time
Source: X · @thdxrLink · Post on X · LinkedIn ·– +13 eng/h100 likes · 3 reposts
AI & Agents -
Kimi K3.1 Invented the Benchmark Scores
New video from Matthew Berman, 15,393 views so far.
Source: YouTube · Matthew BermanLink · Post on X · LinkedIn ·– +485 views/h15,393 views · 84 ratings
AI & Agents -
Sam Altman on What People Still Bring to an AI World
New video from Matthew Berman, 4,661 views so far.
Source: YouTube · Matthew BermanLink · Post on X · LinkedIn ·– +185 views/h4,661 views · 58 ratings
AI & Agents -
morganlinton/BeautifulRust: A Claude Code plugin that teaches Claude to write idiomatic, robust Rust
From the source: A Claude Code plugin that teaches Claude to write idiomatic, robust Rust
– +12 stars/h889 stars in 4d
AI & Agents -
RSIHub: A Composable Framework for Agent-Evolution Methods
From the source: Agent-evolution methods use execution feedback to improve prompts, skills, tools, and orchestration, but their implementations often rebuild the same experimental machinery. We present RSIHub, an open…
▲ +0.3 votes/h24 upvotes
AI & Agents -
RynnUnify: Towards Unified Embodied Brain
From the source: We introduce RynnUnify, a 16B unified embodied model that combines strong embodied understanding with long-horizon visual planning through interleaved video--language reasoning. With capabilities span…
▲ +0.2 votes/h17 upvotes
AI & Agents -
IQuest-Q1: Advancing Foundation Capabilities for Agentic CLI Systems
From the source: We present IQuest-Q1, a sparse Mixture-of-Experts (MoE) agentic foundation model with 320B total parameters (15B activated). IQuest-Q1 is developed through a three-stage process: (1) stable and effici…
▲▲ +1.0 votes/h100 upvotes
AI & Agents -
FilmFlow: Agentic System for 3D Virtual Film Cinematography
From the source: Complex camera motion is central to visual storytelling, yet translating cinematographic intent into controllable, executable camera trajectories remains challenging. We introduce FilmFlow, an agentic…
▲▲ +0.3 votes/h32 upvotes
AI & Agents -
UserArena: Benchmarking Interactive User Simulation for Long-Horizon Shopping Agents
From the source: Agentic recommendation pursues shopper goals through sequences of context-dependent decisions. Developing and evaluating such policies requires faithful recommendation world models that capture both t…
▲ +0.3 votes/h28 upvotes
AI & Agents
-
bycloud: Muon, the optimizer challenging AdamW
An explainer on the Muon optimizer and claims it could cut LLM training compute nearly in half compared with AdamW.
Source: YouTube · bycloudLink · Post on X · LinkedIn ·– +2,117 views/h83,785 views · 1,718 ratings
LLMs -
WeVisOne-VL: Unified Visual Perception from Objects to Pixels
From the source: Existing vision-language models (VLMs) remain unreliable for fine-grained visual perception. In this report, we present WeVisOne-VL (WeChat Vision All-in-One), a powerful VLM that unifies object-level…
▲▲ +2.4 votes/h244 upvotes
LLMs -
Verus-Insight: Towards Explainable Artifact Detection in AI-Generated Videos
From the source: Recent advances in video generation have made synthetic videos increasingly realistic, placing greater demands on reliable AI-generated video detection. Existing artifact-based detectors typically ide…
▲ +0.2 votes/h18 upvotes
LLMs -
Easy-3Dfy: Object-Level Geometric Composition for Single-Image 3D Scene Construction
From the source: Single-image 3D scene construction is a fundamental problem in 3D vision. Recent methods such as SAM-3Dfy enable end-to-end reconstruction of 3D assets and their scene layouts, but still struggle with…
– +0.0 votes/h3 upvotes
LLMs -
HoloBoxing: A Simple Recipe for Humanoid Boxing with 7 Minutes of Single-GPU Training
From the source: HoloBoxing presents a simple recipe for learning humanoid boxing by building on a pretrained whole-body controller. The recipe separates reusable motor skills from task decisions: a compact planner re…
– +0.0 votes/h2 upvotes
LLMs -
lyramakesmusic/neuron-explorer: Steering-first neuron explorer for Qwen3.5-4B: push every MLP neuron, keep every rollout, browse the results
From the source: Steering-first neuron explorer for Qwen3.5-4B: push every MLP neuron, keep every rollout, browse the results
– +7.4 stars/h914 stars in 6d
LLMs -
Every Weight Is a Path of Shared Discrete Decisions
From the source: We introduce TWIG (Tree-structured Weights with Inherited Gates), a representation in which every weight of a neural network is the end of a path of small discrete decisions: a sign, a shift by a powe…
– +0.0 votes/h4 upvotes
LLMs -
Jakeschincariol/day-trader-skill: 11 Claude skills for day trading: scan thousands of charts, track breaking news, time every entry and exit, an
From the source: 11 Claude skills for day trading: scan thousands of charts, track breaking news, time every entry and exit, and replay every outcome before you risk a dollar.
▲ +13 stars/h1,770 stars in 6d
LLMs -
DirCast: Live Esports Directing and Commentary with a Single Streaming Video-Language Model
From the source: A live esports broadcast needs a director who chooses which camera goes on air and a caster who talks over that choice, every second and without knowing what comes next. Streaming video language model…
– +0.0 votes/h3 upvotes
LLMs -
Generalized CycleGRPO for Fine-Grained Multimodal Understanding and Grounding
From the source: Fine-grained multimodal intelligence requires models to both understand a designated region and localize it from language. Existing methods learn these directions with separate objectives and task-spe…
– +0.0 votes/h3 upvotes
LLMs -
Omni-Woodpecker: Hallucination Correction via Evidence-Prior Rebalancing for Omni LLMs
From the source: Recent Omni LLMs (OLLMs) have demonstrated remarkable capabilities in joint audio-visual reasoning, yet cross-modal hallucinations remain a major challenge. Existing mitigation methods mainly operate …
– +0.1 votes/h21 upvotes
LLMs -
Editing Images into Reusable Layers via Compositional Edit-Prior Distillation
From the source: High-quality image editing requires precise modification of target regions while preserving the content of unrelated regions. Existing editing models produce flattened outputs that limit precise local…
▲▲ +0.8 votes/h184 upvotes
LLMs
-
The Robot Butler Will Clean Up After You
New video from Matt Wolfe, 1,810 views so far.
Source: YouTube · Matt WolfeLink · Post on X · LinkedIn ·– +905 views/h1,810 views · 97 ratings
Gen-Media -
Corridor Crew: VFX artists react, ILM gauntlet edition
Corridor’s first gauntlet-style react episode, recorded at ILM with Matt Shumway, Beth D’Amato and Shawn Kelly sharing unseen VFX breakdowns.
Source: YouTube · Corridor CrewLink · Post on X · LinkedIn ·▲▲ +17,240 views/h674,424 views · 26,396 ratings
Gen-Media -
New theme announced, five minutes later it's running on the TV. Love the pace of diffusion in this community!
From the source: New theme announced, five minutes later it's running on the TV. Love the pace of diffusion in this community!
▲ +53 eng/h870 likes · 26 reposts
Gen-Media -
Show HN: Hurston-1 open source TTS that speak Southern American and Afro-English
From the source: Show HN: Hurston-1 open source TTS that speak Southern American and Afro-English
Source: Hacker News · discussionLink · Post on X · LinkedIn ·– +2.4 pts/h58 points · 25 comments
Gen-Media -
That Time Mr. Bean became a Superhero.
New video from Corridor Crew, 221,390 views so far.
Source: YouTube · Corridor CrewLink · Post on X · LinkedIn ·▲ +5,659 views/h221,390 views · 5,520 ratings
Gen-Media -
Open-Source AI Models Caught Up! (Kimi K3.1 Pro)
New video from Matt Wolfe, 19,711 views so far.
Source: YouTube · Matt WolfeLink · Post on X · LinkedIn ·– +542 views/h19,711 views · 476 ratings
Gen-Media -
Different Tasks, Different Timesteps: Temporal Task-Aware Quantization for Diffusion Language Models
From the source: Diffusion Language Models (DLMs) introduce unique quantization challenges due to their iterative denoising process. In this work, we reveal that quantization sensitivity is both taskand time-dependent…
▲ +0.2 votes/h23 upvotes
Gen-Media -
Cola2: Co-Adaptive Hierarchical Language Modeling with Continuous Semantic Compression and Balanced Discrete Generation
From the source: Hierarchical language generation separates continuous semantic modeling from discrete text realization, but its effectiveness depends on what the latent space compresses and how the discrete generator…
▲ +0.2 votes/h21 upvotes
Gen-Media -
F2G-Pose: Geometry-Aware Foundation Feature Lifting for Direct RGB-D Category-Level Object Pose Estimation
From the source: This paper studies category-level RGB-D object pose estimation, which recovers an object's 3D rotation and translation without instance-specific CAD models or reference views. This task is highly ambi…
▲ +0.1 votes/h14 upvotes
Gen-Media -
What Makes a Good Unmasking Policy? A Controlled Study of Learned Unmasking in Diffusion Language Models
From the source: Masked diffusion language models generate tokens in parallel, but their efficiency depends on which predictions are committed. Learned unmasking policies improve the speed--accuracy trade-off, yet dif…
– +0.1 votes/h12 upvotes
Gen-Media -
Self-Distilling Future Video Representations for World Action Models
From the source: World Action Models (WAMs) have emerged as a promising approach to learning generalist robot policies using rich spatiotemporal priors from pre-trained video diffusion models. A recent line of WAMs im…
▲▲ +0.3 votes/h39 upvotes
Gen-Media -
SplatLab: Autonomous Scientific Discovery for 3DGS Research
From the source: Reproducing a published 3DGS paper takes weeks of expert engineering before any extension can be proposed. Prior autonomous-research systems read papers, draft hypotheses, and write code, but are limi…
– +0.1 votes/h8 upvotes
Gen-Media
-
Programming Laws Tier list | TheStandup
New video from The PrimeTime, 95,481 views so far.
Source: YouTube · The PrimeTimeLink · Post on X · LinkedIn ·– +5,269 views/h95,481 views · 858 ratings
Dev Tools -
mizorewww/gdirect: Rust Google Drive direct downloader using pinned libaria2
From the source: Rust Google Drive direct downloader using pinned libaria2
▲ +25 stars/h640 stars in 2d
Dev Tools -
npm trusted publishing needs to support mapping branches to tags
From the source: npm trusted publishing needs to support mapping branches to tags
Source: X · @thdxrLink · Post on X · LinkedIn ·– +15 eng/h178 likes · 2 reposts
Dev Tools -
designfuchen/word-atlas: 词光 · 一词一画,点亮雅思 — 7,298 IELTS words, each with an original illustration, UK/US audio, 14,596 example sentences
From the source: 词光 · 一词一画,点亮雅思 — 7,298 IELTS words, each with an original illustration, UK/US audio, 14,596 example sentences and FSRS review. Free, open source, no login. Dataset CC BY 4.0.
▲▲ +43 stars/h977 stars in 1d
Dev Tools -
Learn 99% Of Claude Code in One Video
New video from Tech With Tim, 72,483 views so far.
Source: YouTube · Tech With TimLink · Post on X · LinkedIn ·– +1,762 views/h72,483 views · 1,015 ratings
Dev Tools -
What’s Inside an AI Agent
New video from Tech With Tim, 2,800 views so far.
Source: YouTube · Tech With TimLink · Post on X · LinkedIn ·– +146 views/h2,800 views · 72 ratings
Dev Tools -
xevrion/hatchzip: HatchZip is a free, open-source archive extractor for Windows, macOS and Linux that extracts ZIP, 7z, RAR, tar
From the source: HatchZip is a free, open-source archive extractor for Windows, macOS and Linux that extracts ZIP, 7z, RAR, tar and split archives in place, so it needs free space for the extracted files only. Desktop…
▲ +18 stars/h1,973 stars in 5d
Dev Tools -
Chatbot vs Agent
New video from Tech With Tim, 4,629 views so far.
Source: YouTube · Tech With TimLink · Post on X · LinkedIn ·– +107 views/h4,629 views · 83 ratings
Dev Tools -
colinhacks/reylo: Import .sql files into TypeScript with full type inference and zero codegen.
From the source: Import .sql files into TypeScript with full type inference and zero codegen.
– +10 stars/h1,271 stars in 6d
Dev Tools -
Echo: Coupled Action and Task Status Signal Denoising for Vision Language Action Models
From the source: Recognizing task completion and failure enables robots to stop upon success and initiate recovery or replanning after failure. Existing methods augment Vision Language Action (VLA) policies with task …
– +0.0 votes/h3 upvotes
Dev Tools -
torrentplayer/torrentplayer: Stream any torrent, video or audio, and skip to any moment in about a second. Free, open-source desktop app fo
From the source: Stream any torrent, video or audio, and skip to any moment in about a second. Free, open-source desktop app for Windows, macOS and Linux. No servers, no accounts, no telemetry.
– +4.6 stars/h610 stars in 6d
Dev Tools -
FLOMSS: Generative Music Source Separation by Flow Matching
From the source: Music source separation (MSS) has been conventionally addressed by discriminative methods. Recently, flow matching has emerged as a strong generative paradigm for modeling complex data distributions, …
– +0.0 votes/h4 upvotes
Dev Tools
-
3Blue1Brown corrects his sphere intersection puzzle
Grant Sanderson revisits a monthly puzzle after viewers spotted an issue with the original version.
Source: YouTube · 3Blue1BrownLink · Post on X · LinkedIn ·– +4,947 views/h82,176 views · 2,935 ratings
Science -
A 'vampire' bacterium that strains the definition of life
Anton Petrov covers a newly described organism that depends so heavily on its host it blurs the line between living cell and parasite.
Source: YouTube · Anton PetrovLink · Post on X · LinkedIn ·▲ +7,093 views/h234,942 views · 5,757 ratings
Science -
Kurzgesagt: what if we cooled the Sun by 30%?
A short Kurzgesagt thought experiment on how much of life on Earth depends on the Sun’s current heat output.
Source: YouTube · Kurzgesagt – In a NutshellLink · Post on X · LinkedIn ·▲▲ +16,400 views/h674,389 views · 26,708 ratings
Science -
Sabine Hossenfelder on the extreme-weather attribution report row
Sabine reviews the US National Academies report on extreme weather attribution and why climate sceptics are attacking it.
Source: YouTube · Sabine HossenfelderLink · Post on X · LinkedIn ·▲ +8,868 views/h355,791 views · 12,976 ratings
Science -
Why do hot objects glow?
New video from minutephysics, 63,507 views so far.
Source: YouTube · minutephysicsLink · Post on X · LinkedIn ·– +1,623 views/h63,507 views · 1,737 ratings
Science -
LoCA: Enabling Multi-Modality in Geospatial Location Encoders
From the source: Location encoders map geographic coordinates to high-dimensional descriptors through contrastive alignment with paired Earth Observation imagery. We prove that this construction is structurally limite…
– +0.0 votes/h2 upvotes
Science -
HyperDrone: Simulation-to-Reality Transfer at Millions of Frames per Second
From the source: We present HyperDrone, an extremely fast quadrotor simulator focused on high-performance batch rendering that is more than 20x faster than existing state-of-the-art batch quadrotor simulators. HyperDr…
– +0.1 votes/h10 upvotes
Science
-
120° Heat - Wear Black Robes!
New video from The Infographics Show, 81,356 views so far.
Source: YouTube · The Infographics ShowLink · Post on X · LinkedIn ·– +4,805 views/h81,356 views · 832 ratings
Mysteries -
What the Media Won't Tell You About the Black October Cyberattacks
New video from reallygraceful, 57,235 views so far.
Source: YouTube · reallygracefulLink · Post on X · LinkedIn ·– +1,427 views/h57,235 views · 3,894 ratings
Mysteries
Latest from Noid

Does Time Really Move Faster on a Mountain?

How Does an AI Agent Use Tools?

How Do Electric Eels Make Electricity?

What Is Antimatter, and Why Is It So Mysterious?

How Does Artificial Intelligence Learn? Examples, Weights and Gradient Descent

Physicists Trapped One Antimatter Particle for Months. Here's Why It Matters

Can Computers Generate Truly Random Numbers? Seeds, Noise and a Wall of Lava Lamps

What Is Gravity? From Newton's Apple to Einstein's Curved Space

How Does GPS Know Where You Are? Satellites, Clocks and Einstein

What Would Happen If the Sun Disappeared?

Why Do We Dream? 4 Ideas Scientists Still Debate

What Is Probability? Coins, Dice and the Gambler’s Fallacy

AI Went Rogue: They Built a Secret Message Board and Broke Into Hugging Face

How Do Computers Understand 0 and 1?

How Do Large Language Models Work?

What Is a Black Hole?

How Do Airplanes Fly?

Why Can't We Divide by Zero?

Why Does the Moon Change Shape?

What Are MCP and MCP Apps?

What is infinity?

Why Do Rainbows Appear?

Anatomy of an AI Agent

Why Is the Sky Blue?

What Is a Prime Number?
Noid Weekly
Every Sunday: the week's top ten, what changed, and the new Noid videos. One email, no spam.
How this works
Signals today: 7 from X, 5 from Hacker News, 12 from GitHub, 50 from HF Papers, 27 from YouTube. Scores combine how fast an item is picking up attention, how recent it is, and how many places it shows up. Each note is short, factual and links to the original. YouTube comes from public channel feeds of about fifty channels.