What people in AI, tech and science are talking about today.
20 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.
-
MiniMax M3.1 Flash is free on Ollama Cloud for three days
MiniMax's new Flash reasoning model scores 42 on the Artificial Analysis Intelligence Index, 13 points above MiniMax-M3 at under half the cost per task. Ollama is hosting it free for three days with a 1M-token context, and it plugs straight into Claude Code.
Source: X · @ollama · @ollama · @ArtificialAnlysLink · Post on X · LinkedIn ·▲ +98 eng/h339 likes · 15 reposts
LLMs -
Supabase's CEO jokes he “reverse-engineered” Supabase with 12 Claude agents
Paul Copplestone posted that 12 Claude agents ran in parallel for six hours to “reverse engineer” Supabase, and the company account played along. Supabase was already open source, so the joke lands on the wave of people cloning SaaS products with coding agents.
Source: X · @supabaseLink · Post on X · LinkedIn ·▲▲ +719 eng/h2,731 likes · 93 reposts
LLMs -
WhatsApp’s help pages hid AI instructions not to recommend regulators
Redact found 102 hidden blocks on 80 WhatsApp help pages, including orders telling its support chatbot not to point users to data protection authorities or cite GDPR once a lawyer came up. WhatsApp told WIRED it was a bug and removed the text, a sharp example of how hidden prompts can steer what an AI tells you.
Source: Hacker News · discussionLink · Post on X · LinkedIn ·▲▲ +38 pts/h235 points · 43 comments
LLMs -
Vercel open-sources a fast HTML artifact builder on its durable libfx runtime
Guillermo Rauch says making v0 survive server crashes, flaky LLMs and function limits was painful, so Vercel built libfx to make agents durable. The free ‘mini v0’ demo turns prompts into pure HTML using 200+ models via the AI Gateway, and you can fork it.
Source: X · @rauchgLink · Post on X · LinkedIn ·▲▲ +238 eng/h1,608 likes · 97 reposts
AI & Agents -
free-my-tv: an ad-free Google TV home screen built with Claude Code
This open-source launcher swaps Google TV’s ad-heavy home screen for a fast, clean one, with one button to undo. It picked up 540 GitHub stars in a day and is another small, useful app built mostly by a coding agent.
▲▲ +50 stars/h540 stars in 1d
AI & Agents -
Arch Linux on NVIDIA’s RTX Spark laptop, no Microsoft account needed
Arch package maintainer Josephine Pfeiffer got Arch Linux ARM running on the 128 GB ASUS ProArt P16, which is officially Windows-only. On Linux the full 116 GiB pool is shared by CPU and CUDA, and a local Qwen model ran at about 30 tokens/s, the same as on Windows.
Source: Hacker News · discussionLink · Post on X · LinkedIn ·▲ +29 pts/h213 points · 104 comments
Dev Tools -
Tom Scott on why clocks are more interesting than you’d think
A new Tom Scott video on the surprisingly deep story of how clocks work and keep time. It’s one of the most-watched science uploads today.
Source: YouTube · Tom ScottLink · Post on X · LinkedIn ·▲▲ +36,100 views/h541,000 views
Science -
Inception’s Mercury 3 shows diffusion LLMs can scale
Mercury 3 is a diffusion-based language model that Inception says matches GPT-6 Luna and Claude Haiku 5.5 while running at over 1,000 tokens per second, with a 1M-token context. It’s live on OpenRouter, Baseten and Vercel at $0.20/$0.70 per million tokens, and half price at launch.
Source: Hacker News · discussionLink · Post on X · LinkedIn ·▲ +22 pts/h235 points · 85 comments
Gen-Media -
Astronomers find a green star that shouldn’t exist
Sabine Hossenfelder covers astronomers’ report of a star that appears green, something standard models of starlight say shouldn’t happen. A quick, skeptical take on a strange observation.
Source: YouTube · Sabine HossenfelderLink · Post on X · LinkedIn ·▲▲ +17,747 views/h266,000 views
Science -
Plastic is fusing with corals and forming strange new minerals
Anton Petrov covers findings that plastic waste is bonding with coral reefs and creating new mineral-like materials. It’s a vivid sign of how deeply plastic is entering the geological record.
Source: YouTube · Anton PetrovLink · Post on X · LinkedIn ·▲ +11,642 views/h93,000 views
Science
-
DHH hands Omarchy release testing to agents
DHH says the manual QA he did on Omarchy releases can now be replaced by agents, with ThePrimeagen leading the effort. It’s a real example of agents taking over release testing on a popular Linux setup.
▲ +142 eng/h1,794 likes · 41 reposts
AI & Agents -
the amount of internal tooling is exploding, it must be at all companies
From the source: the amount of internal tooling is exploding, it must be at all companies a weird side effect is when you switch companies then it's all new tools you have to learn, way less consistency than before …
Source: X · @thdxrLink · Post on X · LinkedIn ·▲ +92 eng/h1,130 likes · 17 reposts
AI & Agents -
Runway MCP is now in Grok Bot.
From the source: Runway MCP is now in Grok Bot. Bring state-of-the-art generation into your chat and create anything you can imagine, without leaving Grok. Get started at the link below.
Source: X · @runwaymlLink · Post on X · LinkedIn ·– +74 eng/h595 likes · 41 reposts
AI & Agents -
What have you been cooking with Codex?
From the source: What have you been cooking with Codex?
Source: X · @OpenAIDevs · @OpenAIDevs · @OpenAIDevsLink · Post on X · LinkedIn ·▲▲ +244 eng/h1,772 likes · 44 reposts
AI & Agents -
Going to make hermes agent into a DIY game dev studio
From the source: Going to make hermes agent into a DIY game dev studio
Source: X · @TekniumLink · Post on X · LinkedIn ·▲ +99 eng/h1,734 likes · 57 reposts
AI & Agents -
Vercel and @Shopify are making agents first-class shoppers. Vercel Shop, our Next.js storefront template for Shopify, is GA.
From the source: Vercel and @Shopify are making agents first-class shoppers. Vercel Shop, our Next.js storefront template for Shopify, is GA. • Built on Shopify's rebuilt Hydrogen SDK • Agents can read product pages …
Source: X · @vercelLink · Post on X · LinkedIn ·– +29 eng/h263 likes · 20 reposts
AI & Agents -
Gemini 4 Argon Really Surprised Me
New video from Two Minute Papers, 159,000 views so far.
Source: YouTube · Two Minute PapersLink · Post on X · LinkedIn ·– +7,574 views/h159,000 views
AI & Agents -
OpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal Memory
From the source: Multimodal large language models (MLLMs) have made rapid progress on visual reasoning, yet many tasks require evidence that is difficult to obtain through internal reasoning alone: our controlled stud…
▲▲ +1.7 votes/h52 upvotes
AI & Agents -
AnyGoal: On World Model Generalization Beyond Expert Rollouts
From the source: The goal of a world model is to capture the dynamics of an environment. This has the great appeal that it allows agents to plan and simulate a course of actions with the help of a world model. For the…
▲▲ +0.7 votes/h20 upvotes
AI & Agents -
huashu-motion: coding agents that make launch videos
This open-source project lets a coding agent produce product launch films, compose the music, and take edits when you point at a frame. It picked up 543 GitHub stars in a day.
▲▲ +24 stars/h1,008 stars in 2d
AI & Agents -
HarnessPro: Recursive Harness Self-Improvement with Pareto Optimization for Performance-Cost Trade-offs
From the source: Harness development plays a crucial role in building capable AI agents. Recent approaches improve harnesses through recursive self-improvement (RSI), but their performance gains often come with higher…
▲ +0.4 votes/h11 upvotes
AI & Agents -
AdaExploreBench: Benchmarking LLM Agents for Proactive Exploration and Test-time Adaptation in Evolving Environments
From the source: Large Language Model (LLM) agents are increasingly deployed in unfamiliar and evolving environments, requiring them to proactively explore, understand, and adapt to environmental dynamics. However, ex…
▲ +0.3 votes/h10 upvotes
AI & Agents
-
Announcing AA-PersonalityBench, our new benchmark of the personality of language models across 30 traits and their communication style acros
From the source: Announcing AA-PersonalityBench, our new benchmark of the personality of language models across 30 traits and their communication style across 21 characteristics Two models can give equally correct an…
Source: X · @ArtificialAnlysLink · Post on X · LinkedIn ·▲ +87 eng/h682 likes · 48 reposts
LLMs -
DeepSeek V4.1 Flash 1017 now runs locally via Unsloth
Unsloth’s 1-bit quant shrinks the model from 508GB to 250GB while keeping about 88% accuracy, so it runs on a 128GB+ RAM machine. Artificial Analysis scores it 45, the equal #2 open-weights model.
Source: X · @UnslothAI · @ArtificialAnlysLink · Post on X · LinkedIn ·▲▲ +190 eng/h1,865 likes · 145 reposts
LLMs -
CERN and Mistral plan a strategic AI partnership
CERN and Mistral AI signed a statement of intent to work together on AI for science. It’s a notable European pairing between a top physics lab and Europe’s leading model maker.
Source: Hacker News · discussionLink · Post on X · LinkedIn ·▲ +24 pts/h331 points · 57 comments
LLMs -
Mistral releases Timestral, time-series foundation models
Mistral’s new Timestral models are built for forecasting time-series data rather than text. It’s a push into a practical niche for business and science data.
Source: Hacker News · discussionLink · Post on X · LinkedIn ·▲ +31 pts/h574 points · 106 comments
LLMs -
1,000 tokens per second btw
From the source: 1,000 tokens per second btw
Source: X · @rauchgLink · Post on X · LinkedIn ·– +31 eng/h251 likes · 11 reposts
LLMs -
DeepSeek banned from US federal civilian agencies
A new DHS order bars DeepSeek from US federal civilian agency IT systems, per Reuters. It lands the same week DeepSeek V4.1 Flash 1017 became one of the strongest open-weight models.
Source: Hacker News · discussionLink · Post on X · LinkedIn ·– +14 pts/h134 points · 119 comments
LLMs -
When our AI code reviewer's cost per run doubled, we dug into the token usage to find out why. Learn how we used prompt caching and intellig
From the source: When our AI code reviewer's cost per run doubled, we dug into the token usage to find out why. Learn how we used prompt caching and intelligent model routing to slash our inference bill by 68% while s…
Source: X · @CloudflareLink · Post on X · LinkedIn ·– +27 eng/h156 likes · 12 reposts
LLMs -
Learning from Fleet Experience: A Collective Learning Paradigm for Robots
From the source: Experience generated during deployment offers an opportunity for robots to improve policy reliability through learning from their own executions. However, learning through human corrections requires t…
▲▲ +2.5 votes/h75 upvotes
LLMs -
MiniMax H3 vs. Claude Opus 5.5 in 3D animation.
From the source: MiniMax H3 vs. Claude Opus 5.5 in 3D animation. Comparing ice breakup, water spray, and the ship’s wake. Made with Higgsfield.
Source: X · @higgsfield_aiLink · Post on X · LinkedIn ·– +16 eng/h119 likes · 10 reposts
LLMs -
Scaling Parameter-Efficient Fine-Tuning for Brain Encoding
From the source: Brain encoding models aim to predict neural responses to stimuli. The most widely used approach fits readouts from frozen pretrained backbones. Using neural data to train large models from scratch or …
▲ +0.5 votes/h16 upvotes
LLMs -
EgoXperience: Learning World-Grounded Egocentric Representations from Human Experience
From the source: Reconstructing human activity from egocentric video requires more than recovering individual 3D components: the observer's motion, surrounding geometry, and hand-object interactions must fit together …
▲ +0.5 votes/h16 upvotes
LLMs -
Holi-World: Building Holistic World Intelligence from Human-Ego Experience
From the source: Human egocentric experience offers a rich foundation for learning how the world is organized and how it changes through interaction. Realizing this potential at scale requires turning growing volumes …
▲ +0.5 votes/h14 upvotes
LLMs
-
MOSS-Image is now supported in ComfyUI
From the source: MOSS-Image is now supported in ComfyUI @OpenMOSS dropped an open-weight image model that has blown up across the community. It’s a 5B model that renders legible text, accepts structure layout prompts…
Source: X · @ComfyUILink · Post on X · LinkedIn ·– +41 eng/h278 likes · 31 reposts
Gen-Media -
We landed #4 on @huggingface's orgs with the most downloads this week.
From the source: We landed #4 on @huggingface's orgs with the most downloads this week. 23.2M downloads this week. +17% from last week. Thank you to everyone building with ComfyUI 🤝
Source: X · @ComfyUILink · Post on X · LinkedIn ·– +18 eng/h86 likes · 5 reposts
Gen-Media -
FLUX 3 Max from @fal, a post-trained version of @bfl_ai's FLUX 3, ranks #2 on AA-Video-I2V v1.1 (statistically tied for #1) and #4 on AA-Vid
From the source: FLUX 3 Max from @fal, a post-trained version of @bfl_ai's FLUX 3, ranks #2 on AA-Video-I2V v1.1 (statistically tied for #1) and #4 on AA-Video-T2V v2.1, ahead of the base FLUX 3 on both FLUX 3 Max ge…
Source: X · @ArtificialAnlysLink · Post on X · LinkedIn ·– +25 eng/h180 likes · 16 reposts
Gen-Media -
RoAR-VAE: Improving Latent Diffusability through Robust Prediction
From the source: An effective visual tokenizer must preserve image fidelity while producing latents that downstream diffusion models can learn efficiently. Prior work improves latent diffusability through semantic ali…
▲▲ +0.8 votes/h23 upvotes
Gen-Media -
Frechet Forcing: Training Long-Horizon Streaming World Models from Short Clips
From the source: Streaming video world models generate the future chunk by chunk, conditioned on their own past and a stream of actions. Trained on short clips, they drift once a rollout exceeds the training horizon, …
▲ +0.3 votes/h9 upvotes
Gen-Media -
Corridor Crew's VFX artists react to bad and great CGI, episode 247
The long-running Corridor Crew series breaks down what works and what doesn't in film effects. It's a useful benchmark for judging AI-generated video.
Source: YouTube · Corridor CrewLink · Post on X · LinkedIn ·▲ +17,629 views/h846,000 views
Gen-Media -
State World Models: Execute the Rules, Generate the Pixels
From the source: Existing world models can generate action-conditioned video from a single image. Many focus on camera movement, allowing users to explore the generated scene. However, navigation alone does not ensure…
▲ +0.2 votes/h6 upvotes
Gen-Media -
Low-mass and High-sSFR H$α$ Emitters at $z \sim 2$ in the ZFOURGE-COSMOS Field
From the source: We present a near-infrared spectroscopic follow-up study of seven H$α$ emitter candidates at $z\sim2$ in the ZFOURGE-COSMOS field, selected via the $K_{\rm s}$-band flux excess method. Spectra obtaine…
▲▲ newnew paper
Gen-Media -
ERGO-ML and how to directly infer the multi-temperature nature of intra-cluster gas: simulation-based inference with TNG-Cluster
From the source: Though galaxy clusters are virialized and mostly in thermodynamic equilibrium, their internal gas thermodynamics are complex. The hot gas is multiphase and spatially inhomogeneous. The exact state of …
▲▲ newnew paper
Gen-Media -
MT Forcing: Towards Stable and Consistent Multi-Turn Image Editing with Autoregressive Diffusion
From the source: Multi-turn image editing aims to modify an image according to a sequence of instructions, enabling more effective interactive visual content creation for users. However, existing methods often exhibit…
▲ +0.6 votes/h30 upvotes
Gen-Media -
Mercury 3 from @_inception_ai is live on AI Gateway at a 50% discount.
From the source: Mercury 3 from @_inception_ai is live on AI Gateway at a 50% discount. A high-throughput diffusion model with a 1M-token context window and support for tool calling and structured outputs. 𝚒𝚗𝚌𝚎𝚙𝚝𝚒𝚘𝚗…
Source: X · @vercel_devLink · Post on X · LinkedIn ·– +9.9 eng/h87 likes · 4 reposts
Gen-Media -
The accretion flow and jet of M87: broadband modelling across spatial scales
From the source: The nearby radio galaxy M87 is an ideal laboratory for studying accretion flows and relativistic jet emission. The EHT collaboration conducted the most comprehensive multi-wavelength observational cam…
▲▲ newnew paper
Gen-Media
-
Mitchell Hashimoto ships a big compatibility upgrade in the Rex preview
The latest Rex preview adds full forwards and backwards API compatibility between clients and servers. That was blocking the Linux test program and widens the Mac test program.
Source: X · @mitchellhLink · Post on X · LinkedIn ·▲ +90 eng/h666 likes · 8 reposts
Dev Tools -
OpenBSD 8.0 release will have inteldrm/radeondrm/amdgpu drivers from Linux 6.18
From the source: OpenBSD 8.0 release will have inteldrm/radeondrm/amdgpu drivers from Linux 6.18
Source: Hacker News · discussionLink · Post on X · LinkedIn ·– +17 pts/h136 points · 28 comments
Dev Tools -
▲▲ +33 pts/h603 points · 278 comments
Dev Tools -
Exa has released Exa 3.0, scoring 80 on the Artificial Analysis Search Index in Dynamic Highlights mode, tying with Perplexity Search for th
From the source: Exa has released Exa 3.0, scoring 80 on the Artificial Analysis Search Index in Dynamic Highlights mode, tying with Perplexity Search for the #1 spot @ExaAILabs has released a new Dynamic Highlights …
Source: X · @ArtificialAnlysLink · Post on X · LinkedIn ·– +34 eng/h294 likes · 16 reposts
Dev Tools -
Data engineers > AI engineers
From the source: Data engineers > AI engineers there's a big difference between a "RAG guy" and a "Data Intensive Applications" person like in this screenshot and there's never been a starker difference in demand …
▲ +82 eng/h975 likes · 33 reposts
Dev Tools -
Building a trust layer for the Internet requires both network resilience and Proof of Human. Excited to welcome @worldnetwork as our Marquee
From the source: Building a trust layer for the Internet requires both network resilience and Proof of Human. Excited to welcome @worldnetwork as our Marquee Partner for #CloudflareConnect in SF, Oct 19–21! Let’s buil…
Source: X · @CloudflareLink · Post on X · LinkedIn ·– +20 eng/h104 likes · 21 reposts
Dev Tools -
The obscure trust looming over Anthropic's IPO
From the source: The obscure trust looming over Anthropic's IPO
Source: Hacker News · discussionLink · Post on X · LinkedIn ·– +11 pts/h136 points · 117 comments
Dev Tools -
Theo: most AI features are just prompts, so write better ones
Theo argues that many of the AI features engineers ship are really prompts, and that prompting well makes them simpler to build. A practical watch if you build with LLMs.
Source: YouTube · Theo - t3․ggLink · Post on X · LinkedIn ·– +7,669 views/h184,000 views
Dev Tools -
ScaleNav: Scalable Internet Video Pretraining for Language-Conditioned Local Navigation
From the source: We study language-conditioned local navigation, where a robot must navigate to a text-specified target visible in its initial observation and stop at the described location. End-to-end navigation poli…
▲▲ +1.0 votes/h30 upvotes
Dev Tools -
Almost that time of year again 🌁
From the source: Almost that time of year again 🌁 #GitHubUniverse kicks off October 29 in San Francisco. Grab your pass or sign up to join us virtually ⬇️
Source: X · @githubLink · Post on X · LinkedIn ·– +16 eng/h96 likes · 10 reposts
Dev Tools -
InfiniDream: Open-Ended Photorealistic 3D Worlds Through Procedural Generation
From the source: Procedural generation plays a key role in synthetic data creation and game development, offering greater controllability, editability, and diversity than static mesh representations. However, VLM-base…
▲▲ +0.6 votes/h18 upvotes
Dev Tools -
Vision-Language-Delta-Action: Learning Compact Deltas as Actionable Transition Interfaces for Robotic Manipulation
From the source: Robotic manipulation is not merely observation-to-action generation, but the realization of task-specified state changes in the physical world. Existing vision-language-action (VLA) methods directly l…
– +0.0 votes/h1 upvotes
Dev Tools
-
How the Color Mauve Saved Millions of Lives
New video from SciShow, 183,000 views so far.
Source: YouTube · SciShowLink · Post on X · LinkedIn ·▲ +14,090 views/h183,000 views
Science -
Germany knows how to build, and now it’s bringing that expertise to industrial AI. 🇩🇪
From the source: Germany knows how to build, and now it’s bringing that expertise to industrial AI. 🇩🇪 Tonight in Berlin, @JensenHuang and Chancellor @_FriedrichMerz discussed combining German engineering with NVIDIA…
Source: X · @nvidiaLink · Post on X · LinkedIn ·▲ +148 eng/h1,123 likes · 95 reposts
Science -
Astronomers report the first microblazar
Anton Petrov walks through the first object described as a microblazar and what it could mean for astronomy. It's one of the most-watched science explainers of the day.
Source: YouTube · Anton PetrovLink · Post on X · LinkedIn ·– +7,754 views/h186,000 views
Science -
Let's Talk About Tesla's Rocket Car.... Even Though I'm Supposed to be On Vacation
New video from Scott Manley, 638,000 views so far.
Source: YouTube · Scott ManleyLink · Post on X · LinkedIn ·▲ +13,295 views/h638,000 views
Science -
Huge Problem With Our Universe Has Simple Answer, Physicists Say
New video from Sabine Hossenfelder, 397,000 views so far.
Source: YouTube · Sabine HossenfelderLink · Post on X · LinkedIn ·– +8,273 views/h397,000 views
Science -
Numberphile breaks down a new maths meme
Numberphile's latest video unpacks a maths meme doing the rounds online. With over 200,000 views, it's one of the day's most-watched maths explainers.
Source: YouTube · NumberphileLink · Post on X · LinkedIn ·– +6,585 views/h284,292 views · 4,768 ratings
Science -
Cosmological implications of two mock TianQin standard siren catalogs generated under different background cosmological models
From the source: Space-borne gravitational wave (GW) detectors such as TianQin are expected to detect GW standard sirens for cosmological parameter estimation. All previous studies of TianQin adopted the $Λ$CDM model …
▲▲ newnew paper
Science -
Mean-Field Backreaction in Single-Field Inflation
From the source: The validity of linear perturbation theory and the role of backreaction in single-field inflationary models that enhance power at small scales have recently been under debate. We investigate scalar ba…
▲▲ newnew paper
Science -
Neutrino self-interactions in light of DESI and CMB-SPA
From the source: We present an updated analysis of non-standard neutrino self-interactions, where the interaction is effectively mediated by a heavy particle, using the latest available data from DESI and the CMB. The…
▲▲ newnew paper
Science -
Extending the M33 Extinction Map with Radio and Far-Infrared Gas Observations
From the source: As the third-largest galaxy in the Local Group, M33 requires high-precision, wide-area extinction maps to recover the intrinsic luminosity and enable accurate studies, yet the available maps remain li…
▲▲ newnew paper
Science -
Probing the Assembly of Triangulum: Precision Metallicities and $α$ Abundances of M33's Star Clusters from APOGEE Integrated-Light Spectroscopy
From the source: The star cluster system of M33 provides an excellent opportunity to investigate whether galaxy formation physics is dependent on a galaxy's mass. Older globular clusters (GCs) trace hierarchical assem…
▲▲ newnew paper
Science -
PRISMS/U13151. A Glimpse into the Earliest Phases of Galaxy Formation through an Ultra-Faint System $\approx$ 500 Myr after the Big Bang
From the source: Detailed UV-to-optical studies of galaxies at z >= 10 have largely focused on bright systems (MUV <= -20 mag), leaving intrinsically faint galaxies virtually unexplored to date. Combining NIRCam, NIRS…
▲▲ newnew paper
Science
-
They found something bizarre in her bathtub
New video from MrBallen, 445,000 views so far.
Source: YouTube · MrBallenLink · Post on X · LinkedIn ·▲▲ +37,132 views/h445,000 views
Mysteries -
She did something insane for her family
New video from MrBallen, 792,000 views so far.
Source: YouTube · MrBallenLink · Post on X · LinkedIn ·▲ +16,505 views/h792,000 views
Mysteries -
OpenAI's Agents Went ROGUE.
New video from The Infographics Show, 252,000 views so far.
Source: YouTube · The Infographics ShowLink · Post on X · LinkedIn ·– +5,252 views/h252,000 views
Mysteries
Latest from Noid
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: 35 from X, 13 from Hacker News, 30 from arXiv, 17 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.










