AI news that matters

1. Users share how they build bot teams with Grok Bot

   Posts describe Grok Bot, in beta since August 11, handling tasks like CRM updates and ticket triage, with daily active users reported above 600,000. Developers share playbooks for teams of specialized bots, while noting it needs setup, clear rules and a $30 monthly subscription.

   23K posts · View on X (opens in a new tab) · Read our story

2. Anthropic's S-1 filing puts revenue, losses and compute deals in focus

   Anthropic's S-1 targets a Nasdaq listing under the ticker ANT in the week of November 16 and reports $11.6 billion in Q2 revenue, according to posts citing the filing. Discussion centers on an $80.7 billion net loss driven by non-cash charges and compute spending shifting toward SpaceX.

   3K posts · View on X (opens in a new tab) · Read our story

3. Gemini 4 Carbon tops benchmarks but splits developers on real-world coding

   Google's Gemini 4 Carbon leads the Artificial Analysis Intelligence Index v5.0 with a score of 59, reports say. Some developers praise it, while others, including t3.gg's Theo, say it hallucinates more in real codebases.

   1.3K posts · View on X (opens in a new tab) · Read our story

4. Pika H3 Flex and Glif's synthetic models fuel AI video debate

   Pika says its H3 Flex video model handles up to 12 references and exports separate audio stems, running 1.8 times faster and 80% cheaper than earlier versions. Glif's new AI modeling agency for generated ads has drawn criticism over risks to modeling and production jobs.

   1.2K posts · View on X (opens in a new tab)

5. DeepSeek publishes super-agent prototype for autonomous GitHub coding

   DeepSeek released dsh-superagent-prototype on GitHub under an MIT license, using a persistent router and up to three short-lived workers that open draft pull requests. It is labeled research code, and users voice caution about giving agents repository autonomy.

   1.2K posts · View on X (opens in a new tab) · DeepSeek news

6. Gemini 4 Argon builds web apps, reports and edits Docs in chat

   Google's Gemini 4 Argon can break down complex tasks such as generating web apps, 3D files and reports, and edit PDFs and Docs directly in chat. Users praise its explanations, though some report occasional inaccuracies on architecture details.

   1K posts · View on X (opens in a new tab) · Gemini news

7. GPT-6.1 Terra benchmarks spark cost-versus-performance debate

   OpenAI's GPT-6.1 Terra reportedly reaches a 90.00 combined VulcanBench score at max effort, passing all 23 tasks for about $0.36 each. Users weigh its $1/$5 per million token pricing for high-volume work against GPT-6.1 Sol for coding.

   811 posts · View on X (opens in a new tab) · Read our story

8. Leaks suggest Zhipu AI's GLM-5.5 is nearing completion

   Leaked details claim Zhipu AI's GLM-5.5 is in post-training, merging 32B, 122B and 256B parameter versions into a new base model. Z.ai has not confirmed the model, and users are hoping for a local 32B variant that runs on a single GPU.

   596 posts · View on X (opens in a new tab)

9. Vercel Labs releases open-source agent-desktop for AI computer control

   Vercel Labs' agent-desktop lets AI coding agents control macOS, Windows and Linux apps through the OS accessibility tree, with parallel agents each using their own cursor. One demo cut token use from 198,000 to 31,000 over 52 steps, according to posts.

   501 posts · View on X (opens in a new tab)

10. Anthropic rewrites its Messages API in Rust with help from Claude

    Anthropic says it moved its Messages API from a Python monolith to Rust, with Claude writing most of the code and shadow testing on real traffic before a September rollout. The company reports 10 to 20% faster responses.

    289 posts · View on X (opens in a new tab) · Anthropic news

Based on public posts on X.