Subscribe

What people in AI, tech and science are talking about today.

18 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.

  1. DeepSeek ships V4.1 Flash 1017: faster, cheaper, shorter answers

    DeepSeek released V4.1 Flash 1017, claiming stronger benchmarks, faster task completion and shorter outputs than the previous version, live on its app and API. Ollama and Vercel's AI Gateway had it within hours, and the Hugging Face release was a top Hacker News thread with 600+ comments.

    Source: X · @deepseek_ai · @vercel_dev · @ollama · @ollama · theykkLink · Post on X · LinkedIn ·
    ▲▲ +2,289 eng/h32,516 likes · 3,645 reposts
    AI & Agents

  2. Higgsfield puts DeepSeek V4.1 Flash to work on 3D motion design

    Higgsfield posted a run of tests pitting DeepSeek V4.1 Flash against Claude Opus 5.5 and Gemini 4 Argon on 3D scenes, animation and UI transitions, calling it the new best model for motion design. The new model is quickly showing up in creative tools, not just code.

    Source: X · @higgsfield_ai · @higgsfield_ai · @higgsfield_ai · @higgsfield_ai · @higgsfield_aiLink · Post on X · LinkedIn ·
    ▲ +127 eng/h1,195 likes · 64 reposts
    LLMs

  3. Perplexity open-sources SPACE, the agent sandbox behind Computer

    Perplexity open-sourced SPACE, the sandbox platform its Computer product uses to control where agents run code and what they can reach. NVIDIA AI highlighted it as a step toward safer agents, and it gives teams a ready-made option for agent sandboxing.

    Source: X · @NVIDIAAILink · Post on X · LinkedIn ·
    – +44 eng/h416 likes · 36 reposts
    AI & Agents

  4. Huawei's MateBook Fold: an 18-inch screen that folds to magazine size

    DHH showed off Huawei's MateBook Fold, an 18-inch laptop that folds to the size of a magazine and runs a Linux-derived OS. He wants to get Omarchy, his Linux setup, running on it.

    Source: X · @dhhLink · Post on X · LinkedIn ·
    ▲▲ +356 eng/h745 likes · 17 reposts
    Dev Tools

  5. Claude cracks the decades-old mystery of the ECDSA curve seeds

    The NSA supplied the seeds behind NIST's standard elliptic curves (like P-256) in 2000 with no explanation, which fed years of suspicion. Anthropic's Frontier Red Team says Claude traced them back: “it's just timestamps”, so no action is needed. It was one of the top stories on Hacker News today.

    Source: Hacker News · discussionLink · Post on X · LinkedIn ·
    ▲ +35 pts/h1030 points · 380 comments
    LLMs

  6. Codex Desktop bug piles up crash dumps, over 5GB a day on macOS

    A GitHub issue says Codex Desktop on macOS keeps writing Crashpad crash dumps in the background while still working normally, reaching 4.9GB and 54,000 files in a day. It hit the Hacker News front page; check your Application Support folder if you use the app.

    Source: Hacker News · discussionLink · Post on X · LinkedIn ·
    – +11 pts/h115 points · 41 comments
    AI & Agents

  7. Llmspeed lets you feel a local LLM's speed before buying hardware

    Llmspeed is a Show HN site that simulates how fast a local model would stream on given hardware, so you can judge tokens per second before you spend money on a GPU or Mac.

    Source: Hacker News · discussionLink · Post on X · LinkedIn ·
    – +12 pts/h272 points · 78 comments
    LLMs

  8. A real-time WebGPU ocean in one 61 KB HTML file, built with Claude Code

    A from-scratch WebGPU ocean renderer with FFT waves, whitecaps and a physically based sunset sky, all in a single 61 KB HTML file. The author says it was built with Claude Code on Opus 5.5, and it gained 538 stars in two days.

    Source: GitHubLink · Post on X · LinkedIn ·
    ▲ +18 stars/h538 stars in 2d
    AI & Agents

  9. You could have invented transformers from memory

    A Hacker News favourite essay that rebuilds the transformer starting from associative memory, making attention feel like a natural idea rather than a trick.

    Source: Hacker News · discussionLink · Post on X · LinkedIn ·
    ▲ +17 pts/h535 points · 124 comments
    LLMs

  10. A Blender add-on puts an AI assistant and MCP server inside your scene

    blender-ai-mcp-addon adds a chat assistant and an MCP server to Blender, so an AI model or coding agent can see the scene and work on it with scene tools. It picked up 660 GitHub stars in three days.

    Source: GitHubLink · Post on X · LinkedIn ·
    – +12 stars/h661 stars in 3d
    AI & Agents

Latest from Noid

All videos

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: 23 from X, 7 from Hacker News, 30 from arXiv, 19 from GitHub, 50 from HF Papers. 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.