Tech news, October 23, 2026

Beginning today, all uploads from user accounts with 2FA enabled will be required to use an API Token or Trusted Publisher configuration in place of their password.

What's changing?

Incident Report: User Account Takeover What happened?

Twilio customers may be experiencing SMS delivery delays and failures from Twilio to Telecall network subscribers in Brazil.

The Beta channel is being updated to OS version 16852.10.0 (Browser version 157.0.8103.11 ) for most ChromeOS devices.

Replacement: deep-research-preview-04-2026.

The Copilot usage metrics API now reports model usage from more of the tools your developers use with Copilot, including tools like OpenCode that weren’t previously captured in yo…

Now that code layers is in open beta, you can bring code to the canvas.

Amazon OpenSearch intelligent ingestion is a fully managed, serverless capability that turns documents and structured data in Amazon S3 into RAG-ready, searchable indexes with no…

You can now view and manage your team's Spend Management budget from the Vercel CLI.

The new GitHub app in Teams brings notifications and Copilot together, helping teams turn issues into fixes and ship code faster without leaving the conversation.

In an AI software factory, agents build and engineers decide what ships.

Follow through on your meeting tasks with meeting recaps in Google Chat To make it easier for teams to move from planning to action, we’re introducing meeting recaps in Google Cha…

Web Search on Amazon Bedrock now supports two new controls that give you more precision over the results that ground a model's response: domain filtering and user location.

Vercel MCP now includes submit_feedback, a tool for sending bug reports, friction reports, and feature requests directly to the Vercel team from your agent or harness.

An agent working on production systems needs permission to act.

See how Snowflake's sales finance team used AI to cut its monthly forecast from more than a week to a few hours, and what it took to get there.

Gemini Notebook is now also available as a Connected App in the Gemini app.

Trusted Identity Propagation with AWS IAM Identity Center now works across Regions.

Azure Backup now supports Item-Level Recovery (ILR) for Azure virtual machines using Premium SSD v2 and Ultra disks with both 512 byte and 4KB logical sector sizes.

AI agents now write a growing share of production data pipelines (over 70% of SDP...

Why does real-time energy trading need more than faster data?Energy trading desks...

Deployment summaries now show the size of the build cache each deployment saved and how close it is to the limit for its build machine.

You ship an agent.

Between June 1 and June 13, 2026, some Enterprise Managed User accounts that were deprovisioned through SCIM were incorrectly deleted instead of being suspended and obfuscated as…

Generative AI

Trending on the Hugging Face Hub.

Kimi K3.1 Pro is Moonshot AI's flagship open-weight multimodal reasoning model and a post-trained successor to Kimi K3, with a 1M-token context window.

New open weights from OpenBMB (MiniCPM) on the Hugging Face Hub.

New open weights from InternLM (Shanghai AI Lab) on the Hugging Face Hub.

Trending on the Hugging Face Hub.

Posted by Timothy Legge on Oct 23 ======================================================================== CVE-2026-115936 CPAN Security Group ====================================…

Posted by Tomasz Kaye on Oct 23 Follow-up: the fix for this issue has been merged upstream.

Posted by Tomasz Kaye on Oct 23 Follow-up: the fix for this issue has been merged upstream.

Posted by Tomasz Kaye on Oct 23 Follow-up: the fix for this issue has been merged upstream.

Posted by Robert Rothenberg on Oct 23 ======================================================================== CVE-2026-113877 CPAN Security Group ================================…

Affects wasmtime.

Affects pillow.

Affects laravel/framework.

Last week, there were disclosed in and that have been added to the Wordfence Intelligence Vulnerability Database, and there were that contributed to WordPress Security last week.

Socket uncovered secret-stealing workflows that ran in LiteLLM’s GitHub repository 11 days before the supply chain attack on PyPI and Docker.

Multi-Turn Reinforcement Learning (MTRL) on Amazon SageMaker AI trains models on the outcome of a whole multi-step task rather than labeled steps.

The Global Performance Tuner is a feature for automatic performance tuning released in NVIDIA TensorRT 11.1.

Deploying a trained neural network involves more than translating each operator into a GPU kernel.

Updating a SageMaker AI endpoint that has Application Auto Scaling registered makes AWS CloudFormation and AWS CDK deployments fail.

Amazon Quick Sight layer maps now support Point and LineString geometry from the same GeoJSON workflow, so you can plot store locations and the routes between them.

Bayer Crop Science built the Agentic Decision Engine (ADE), a multi-agent platform on Amazon Bedrock AgentCore that lets breeders, analysts, and researchers ask natural language q…

Teaching a robot to navigate an obstacle course requires more than a walking policy.

Python f-strings give you a readable, concise way to interpolate and format strings, and they beat.format() and the % operator on speed.

Build a Django example project step by step.

Use Python's property() to add managed attributes to your classes, with validation, computed values, and caching, without breaking your API.

In Compton Lecture, the distinguished legal scholar warns of potential “loss of the Enlightenment” while suggesting ways innovations can meet our civic and cultural needs.

During the School of Humanities, Arts, and Social Sciences’ 75th Anniversary Symposium, President Sally Kornbluth said “MIT has never needed your particular insight and guidance m…

Today, supply chain teams are challenged by how long it takes to resolve urgent issues, and disruptions are estimated to cost the industry billions in losses each year.

AI Now's journalist in residence Brian Merchant reflects on the People's AI Assembly, held in NYC on October 19th.

Assistant Professor Benjamin Lindquist studies how engineers deliberately injected randomness into early AI, giving rise to deep-learning models that sometimes hallucinate.

Notice is hereby given, pursuant to the provisions of the rules and regulations of the U.S.

Last week at a small conference, there was talk about a future event referred to as The Hackening.

This article in Nature argues that it’s due to an overfishing of predators and warming sea temperatures.

Trainium Best for Inference?

The Concept of Ruliobiology Biology is (as Charles Darwin put it a century and a half ago) full of “endless forms most beautiful and most wonderful”.

The best Stratechery content from the week of October 19, 2026, including the best iPad but worst phone, SS Sarandos in a storm, and You Can See Everything.

This is the second half of my HPTS 2026 notes.

This article argues that instruction-fine-tuned language models should be understood as implementing a genre, what is proposed as the instructional-conversational genre.

Generative recommenders retrieve items by writing their semantic IDs (SIDs) code by code under trie-constrained beam search, an approach now deployed on large web platforms.

Training generalizable embodied agents requires realistic, varied, and simulation-ready indoor environments.

Actions that condition world models are highly heterogeneous, and thus existing action-conditioned world models are typically built for individual action types, fragmenting alread…

To estimate correspondences across images, existing approaches match either sparse interest points, which is highly efficient but limited by repeatability, or dense pixel grids, w…

Robust reinforcement learning (RL) is critical for reliable real-world deployment, where policies may fail under diverse shifts and disruptions.

Collecting high-quality demonstrations remains a central bottleneck for learning dexterous manipulation.

Semantic misinformation can arise when an authentic image is paired with a caption that changes or distorts its narrative.

Learning humanoid control from monocular human videos offers a scalable way to acquire diverse whole-body behaviors.

Few-shot industrial anomaly detection requires modeling category-specific normality from only a handful of defect-free images.