Chronicle 41 items · updated 2026-11-22 13:20 UTC

Chronicle AI Brief, November 22, 2026

The latest in AI, clustered and ranked. Repeated hype gets pushed down so the actual signal stays up top.

Edition highlights

Data Processing Incidents on Inference Providers

OpenRouter reports data processing incidents on inference providers, detailing specific privacy and data handling issues.

Hacker News (AI-filtered)·2026-11-22 02:14 UTC·company announcement·model summary

Current Stack

A personal list of software tools and services currently in use.

Daniel Miessler·2026-11-21 23:40 UTC·tool·model summary
Viewing 2026-11-22
Last 3 hours(5)
  1. Security researchers show how cheap AI inference providers steal keys and snoop on users

    Security research reveals vulnerabilities and data privacy risks in third-party AI inference providers.

    The Decoder·2026-11-22 12:07 UTC·news·Story link0.85(n 0.83 · t 0.74)
    why surfaced · high
    • specific engineering subject or instructional task in source metadata
    • high novelty against the 30-day history
    • classified as concrete builder or research signal
    • fresh within the current refresh window
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for Security researchers show how cheap AI inference providers steal keys and snoop on users
  2. Banks Have Had It Up to Here With Your AI Agent

    US banks are banning third-party AI agents from accessing accounts and disclaiming liability for bot errors.

    WIRED AI·2026-11-22 11:00 UTC·news·Story link0.85(n 0.81 · t 0.76)
    why surfaced · high
    • specific engineering subject or instructional task in source metadata
    • high novelty against the 30-day history
    • classified as concrete builder or research signal
    • fresh within the current refresh window
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for Banks Have Had It Up to Here With Your AI Agent
  3. Claude Code Has Been Leaking Secrets Through Its "Don't Ask Again" Feature

    Security vulnerability identified in Claude Code where local settings files inadvertently expose secrets.

    InfoQ AI/ML/Data·2026-11-22 11:00 UTC·news·Story link0.72(n 0.78 · t 0.78)
    why surfaced · high
    • high novelty against the 30-day history
    • classified as concrete builder or research signal
    • fresh within the current refresh window
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for Claude Code Has Been Leaking Secrets Through Its "Don't Ask Again" Feature
  4. AI is getting better faster than experts can reliably test whether new systems are safe, METR warns

    METR report highlights challenges in evaluating safety and capabilities of rapidly advancing AI models.

    The Decoder·2026-11-22 13:00 UTC·news·Story link0.67(n 0.83 · t 0.74)
    why surfaced · high
    • high novelty against the 30-day history
    • classified as concrete builder or research signal
    • fresh within the current refresh window
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for AI is getting better faster than experts can reliably test whether new systems are safe, METR warns
  5. Lean and Mathlib community officially confirms Levent Alpöge's proof of the Riemann hypothesis is correct

    Formal verification of the Riemann Hypothesis proof using the Lean and Mathlib theorem provers.

    The Decoder·2026-11-22 11:08 UTC·news·Story link0.66(n 0.77 · t 0.74)
    why surfaced · high
    • high novelty against the 30-day history
    • classified as concrete builder or research signal
    • fresh within the current refresh window
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for Lean and Mathlib community officially confirms Levent Alpöge's proof of the Riemann hypothesis is correct
Earlier today(17)
  1. AWS Lambda Now Runs GPU Inference with 48 Concurrent Invocations Sharing One Model per GPU

    AWS Lambda now supports GPU inference with shared model instances and vLLM/SGLang integration.

    InfoQ AI/ML/Data·2026-11-22 06:27 UTC·company announcement·Story link0.85(n 0.84 · t 0.78)
    why surfaced · high
    • specific engineering subject or instructional task in source metadata
    • high novelty against the 30-day history
    • classified as concrete builder or research signal
    • fresh within the current refresh window
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for AWS Lambda Now Runs GPU Inference with 48 Concurrent Invocations Sharing One Model per GPU
  2. Google Announces Agent Compute Engine for Secure AI Agent Execution

    Google introduces Agent Compute Engine for isolated, secure execution of AI agents in microVMs.

    InfoQ AI/ML/Data·2026-11-22 06:30 UTC·company announcement·Story link0.83(n 0.75 · t 0.78)
    why surfaced · high
    • specific engineering subject or instructional task in source metadata
    • high novelty against the 30-day history
    • classified as concrete builder or research signal
    • fresh within the current refresh window
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for Google Announces Agent Compute Engine for Secure AI Agent Execution
  3. Data Processing Incidents on Inference Providers

    OpenRouter reports data processing incidents on inference providers, detailing specific privacy and data handling issues.

    Hacker News (AI-filtered)·2026-11-22 02:14 UTC·company announcement·Story link0.74(n 0.80 · t 0.65)
    why surfaced · high
    • high novelty against the 30-day history
    • classified as concrete builder or research signal
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  4. 87% of the Anthropic power will come from gas and 11% from coal in 2028

    Analysis projects Anthropic's 2028 power mix will be 87% gas and 11% coal, highlighting energy sourcing issues.

    Hacker News (AI-filtered)·2026-11-22 08:57 UTC·news·Story link0.66(n 0.77 · t 0.65)
    why surfaced · high
    • high novelty against the 30-day history
    • classified as concrete builder or research signal
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  5. Google’s Elephant-Goldfish Model: Design First, Code Last

    This video is a hands-on tutorial on Elephant-Goldfish model which is Open-source plugin for designing and reviewing software with agents. 📬Weekly AI Newsletter: https://fahdmirza.substack.com/ 🔥 Ge

    Fahd Mirza YouTube·2026-11-21 19:15 UTC·video·Story link0.64(n 0.00 · t 0.66)
    why surfaced · familiar
    • specific engineering subject or instructional task in source metadata
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • Assessment: deterministic. Metadata evidence rules; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for Google’s Elephant-Goldfish Model: Design First, Code Last
  6. Pandex – Official European Law for AI Agents, over MCP

    Pandex proposes a European legal framework for AI agents built on top of the Model Context Protocol (MCP).

    Hacker News (AI-filtered)·2026-11-21 20:47 UTC·tool·Story link0.63(n 0.82 · t 0.65)
    why surfaced · high
    • high novelty against the 30-day history
    • classified as useful but lower-confidence signal
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  7. GPT-6.2 Astra and Sol get to the point faster but score lower on transparency than Claude

    Comparison of GPT-6.2 Astra and Sol versus Claude regarding response brevity and transparency.

    The Decoder·2026-11-22 09:41 UTC·news·Story link0.55(n 0.82 · t 0.74)
    why surfaced · high
    • high novelty against the 30-day history
    • classified as useful but lower-confidence signal
    • fresh within the current refresh window
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for GPT-6.2 Astra and Sol get to the point faster but score lower on transparency than Claude
  8. Current Stack

    A personal list of software tools and services currently in use.

    Daniel Miessler·2026-11-21 23:40 UTC·tool·Story link0.54(n 0.81 · t 0.78)
    why surfaced · high
    • high novelty against the 30-day history
    • classified as useful but lower-confidence signal
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  9. Nvidia researchers argue huge AI context windows are a dead end

    Nvidia research suggests periodic summarization and keyword search can replace massive context windows for efficiency.

    The Decoder·2026-11-21 16:23 UTC·paper·Story link0.53(n 0.00 · t 0.74)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for Nvidia researchers argue huge AI context windows are a dead end
  10. Qwen Sparse Attention: How Qwen4-Max Reads a Million Tokens

    This video explains QSA with animations. 📬Weekly AI Newsletter: https://fahdmirza.substack.com/ 🔥 Buy Me a Coffee to support the channel: https://ko-fi.com/fahdmirza #qsa PLEASE FOLLOW ME: ▶ LinkedI

    Fahd Mirza YouTube·2026-11-22 06:00 UTC·video·Story link0.53(n 0.83 · t 0.66)
    why surfaced · high
    • high novelty against the 30-day history
    • classified as useful but lower-confidence signal
    • Assessment: deterministic. Metadata evidence rules; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for Qwen Sparse Attention: How Qwen4-Max Reads a Million Tokens
  11. Meta ads is now available in AWS Quick

    AWS Quick adds a Meta ads connector allowing natural language management of Meta ad campaigns and data integration.

    AWS Machine Learning Blog·2026-11-22 01:59 UTC·company announcement·Story link0.52(n 0.68 · t 0.80)
    why surfaced · medium
    • meaningfully different from recent coverage
    • classified as useful but lower-confidence signal
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  12. Researcher plants 2m canaries into the AI supply chain. A privacy disaster surfaced

    Researcher embedded 2 million canaries in the AI supply chain, revealing widespread exposure of secret user data.

    Lobsters (AI tag)·2026-11-22 02:03 UTC·news·Story link0.48(n 0.00 · t 0.70)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  13. OpenAI shares new prompting tips for GPT-6.2 Astra and Sol to fix brief answers and unnecessary questions

    OpenAI provides prompt templates to mitigate verbosity and unnecessary follow-up questions in GPT-6.2 models.

    The Decoder·2026-11-21 13:32 UTC·tutorial·Story link0.37(n 0.00 · t 0.74)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as useful but lower-confidence signal
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for OpenAI shares new prompting tips for GPT-6.2 Astra and Sol to fix brief answers and unnecessary questions
  14. Signal restricts mobile screenshots as it confronts ‘existential threat’ of AI surveillance

    Signal restricts screenshots on mobile apps citing AI surveillance concerns.

    TechCrunch AI·2026-11-21 21:37 UTC·news·Story link0.32(n 0.00 · t 0.72)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as useful but lower-confidence signal
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  15. A tale of two Nvidias

    Analysis of Nvidia's recent financial performance and executive leadership sentiment.

    TechCrunch AI·2026-11-21 21:07 UTC·news·Story link0.32(n 0.00 · t 0.72)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as useful but lower-confidence signal
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  16. Is selling an AI version of your face to this microdrama startup the future of Hollywood?

    Discussion on the implications of actors licensing their likeness to AI microdrama startups.

    TechCrunch AI·2026-11-21 16:00 UTC·news·Story link0.31(n 0.00 · t 0.72)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as useful but lower-confidence signal
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
Yesterday & older(19)
  1. Lakonik/pi-H3 (9038 downloads, 133 likes)

    pi-H3 is a distilled flow-matching model for audio-video generation tasks.

    Hugging Face trending models·2026-11-12 03:39 UTC·listed 2026-11-22·model release·Story link0.52(n 0.68 · t 0.58)
    why surfaced · medium
    • meaningfully different from recent coverage
    • classified as useful but lower-confidence signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  2. SulphurAI/Sulphur-3 (78931 downloads, 739 likes)

    Sulphur-3 is a fine-tuned variant of the MiniMax-H3 base model.

    Hugging Face trending models·2026-10-31 20:31 UTC·listed 2026-11-22·model release·Story link0.51(n 0.61 · t 0.58)
    why surfaced · medium
    • meaningfully different from recent coverage
    • classified as useful but lower-confidence signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    • Read the source's license and usage conditions before adopting the release.
    How Chronicle selects stories · report a correction
  3. Qwen/Qwen3.8-27B (6297396 downloads, 20837 likes)

    Large-scale multimodal Qwen model for conversational and vision tasks.

    Hugging Face trending models·2026-08-05 08:22 UTC·listed 2026-11-22·model release·Story link0.50(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    • Read the source's license and usage conditions before adopting the release.
    How Chronicle selects stories · report a correction
  4. Lightricks/LTX-2.5 (2993088 downloads, 13004 likes)

    LTX-2.5 is a diffusion-based model for video generation tasks including image-to-video and video-to-video.

    Hugging Face trending models·2026-07-23 07:55 UTC·listed 2026-11-22·model release·Story link0.50(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  5. MiniMaxAI/MiniMax-H3-Image (952024 downloads, 1710 likes)

    MiniMax-H3-Image is a diffusion model supporting text-to-image, image-to-image, and image editing.

    Hugging Face trending models·2026-11-03 06:34 UTC·listed 2026-11-22·model release·Story link0.50(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  6. higurashi-ai/Haru-4B (8608 downloads, 858 likes)

    4B parameter model designed for full-duplex speech-to-speech interaction.

    Hugging Face trending models·2026-11-12 11:05 UTC·listed 2026-11-22·model release·Story link0.50(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  7. inworld/voice-isolator (0 downloads, 570 likes)

    Specialized model for speech enhancement and noise suppression.

    Hugging Face trending models·2026-11-14 02:14 UTC·listed 2026-11-22·model release·Story link0.50(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  8. apple/FastFrames-L (0 downloads, 510 likes)

    Apple-released fine-tune of Wan2.1-T2V-14B for video generation tasks.

    Hugging Face trending models·2026-11-16 17:21 UTC·listed 2026-11-22·model release·Story link0.50(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    • Read the source's license and usage conditions before adopting the release.
    How Chronicle selects stories · report a correction
  9. n4ze3m/tiny-voca (10768 downloads, 387 likes)

    Tiny-voca is a lightweight text-to-speech model provided in ONNX and GGUF formats.

    Hugging Face trending models·2026-11-14 15:46 UTC·listed 2026-11-22·model release·Story link0.50(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    • Inspect the evaluation setup, data and baselines before applying the reported results.
    How Chronicle selects stories · report a correction
  10. apple/DepthPro2 (0 downloads, 231 likes)

    Apple's updated model for metric monocular depth estimation in images and video.

    Hugging Face trending models·2026-11-03 16:55 UTC·listed 2026-11-22·model release·Story link0.50(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  11. New open model Inkling 2 Mini thinks in loops to solve problems with fewer words and still runs on a laptop

    Inkling 2 Mini uses a looped transformer architecture to perform internal reasoning with fewer generated tokens.

    The Decoder·2026-11-21 10:33 UTC·model release·Story link0.47(n 0.00 · t 0.74)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as concrete builder or research signal
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
    Thumbnail for New open model Inkling 2 Mini thinks in loops to solve problems with fewer words and still runs on a laptop
  12. thinkingmachines/Inkling-2-Mini (23543 downloads, 3994 likes)

    Multimodal conversational model with image-text-to-text capabilities.

    Hugging Face trending models·2026-11-17 00:26 UTC·listed 2026-11-22·model release·Story link0.34(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as useful but lower-confidence signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  13. harness-hub/Qwen3.8-27B-pi-adapted (2831 downloads, 954 likes)

    Qwen-based model fine-tuned using GRPO for alignment.

    Hugging Face trending models·2026-11-16 06:52 UTC·listed 2026-11-22·model release·Story link0.34(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as useful but lower-confidence signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  14. unsloth/Inkling-2-Mini-GGUF (805993 downloads, 764 likes)

    GGUF quantization of the Inkling-2-Mini model for local inference.

    Hugging Face trending models·2026-11-20 13:09 UTC·listed 2026-11-22·model release·Story link0.34(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as useful but lower-confidence signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  15. chatrlabs/Qwen-Image-2.1-Uncensored-Omni-Quantized (381769 downloads, 767 likes)

    An uncensored, quantized version of the Qwen-Image-2.1 model for text-to-image generation.

    Hugging Face trending models·2026-11-03 16:13 UTC·listed 2026-11-22·model release·Story link0.34(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as useful but lower-confidence signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  16. tencent/Hy4-flash (6617 downloads, 578 likes)

    Flash-optimized version of the Tencent Hunyuan text generation model.

    Hugging Face trending models·2026-11-17 03:44 UTC·listed 2026-11-22·model release·Story link0.34(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as useful but lower-confidence signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  17. sensenova/SenseNova-Loop-9B (1863 downloads, 476 likes)

    SenseNova-Loop-9B is a multimodal text-generation model supporting image-text-to-text tasks.

    Hugging Face trending models·2026-11-17 10:16 UTC·listed 2026-11-22·model release·Story link0.34(n 0.00 · t 0.58)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as useful but lower-confidence signal
    • trending observed 2026-11-22; publication date preserved
    • fresh within the current refresh window
    • source-native discussion or engagement is unusually high
    • Assessment: llm. Model assessment from source metadata; not independently verified
    How Chronicle selects stories · report a correction
  18. When the fix makes the outage worse #microsoft #outage #tech #ai

    Microsoft needed nearly a day to restore email, and part of that came from the recovery itself. Traffic sent to healthy servers during the fix took those servers down too. Now engineers are defending

    AI News & Strategy Daily·2026-11-21 04:00 UTC·video·Story link0.27(n 0.00 · t 0.62)
    why surfaced · familiar
    • kept for context despite familiar coverage
    • classified as useful but lower-confidence signal
    • Assessment: deterministic. Metadata evidence rules; not independently verified
    How Chronicle selects stories · report a correction
You're caught upNext refresh follows the public schedule.

Previous editions

Same signal-first ranking, earlier dates.

Open archive