Tech news, October 8, 2026
Thursday: 607 pieces the paper read from primary sources, by section, newest first. Each links to its source; the line under it is the source's own opening.
The morning briefing
As the paper printed it on /brief/ at 00:03 UTC, in the sources' own words.
- LeadClaude Haiku 5.5
We've launched Claude Haiku 5.5 (claude-haiku-5-5), our most capable model tuned for high-volume and latency-sensitive work.
- NewsClaude Haiku 5.5 in GitHub Copilot
Claude Haiku 5.5, Anthropic’s newest lightweight model, is now generally available in GitHub Copilot.
- NewsIntroducing Claude Haiku 5.5 on AWS
Claude Haiku 5.5 is now available on Amazon Bedrock and Claude Platform on AWS.
- OutageSMS and MMS 10DLC Failure
Bandwidth teams are investigating an incident impacting outbound 10DLC messages towards AT&T.
- SecurityBuilding your AI vulnerability harness, Part 1
Vulnerability scanners produce findings faster than manual triage can process them.
- Deadlinegemini-2.5-flash shuts down in 12 days
Replacement: gemini-3.8-flash or gemini-3.5-flash-lite or gemini-3.1-flash-lite.
News
117 pieces, the newest 20 here.
- Release Notes for Safari Technology Preview 254
Safari Technology Preview Release 254 is now available for download for macOS Golden Gate and macOS Tahoe.
- Announcing Workday Data Connect federation in Unity Catalog
We're excited to announce the publicly available Beta of the Workday Data Connect federation connector for Unity Catalog...
- [Launched] Generally Available: Managed StandardV2 NAT Gateway for AKS
AKS will now provision and manage a StandardV2 NAT Gateway for clusters using an AKS-managed virtual network.StandardV2 is now the default managed NAT Gateway SKU for new clusters…
- Don’t give AI agents root. Make them propose the next system state.
The uncomfortable question In the past few months, I’ve been hearing different versions of the same question: under what conditions can we safely allow an autonomous system to ope…
- Lakebase and Agentic SDLC: Branching Databases for Coding Agents
AI has changed how software gets built.
- Introducing Funke: Native HL7v2 Parsing on Databricks
The HL7v2 problem in the lakehouseHL7v2 is the messaging standard that quietly runs healthcare...
- Testing Thinking Mode
hi everyone!
- Alpha Changelog - 10/7/26
Hi everyone, here's this week's round of changes on alpha.midjourney.com.
- Biomedical Imaging's Real Bottleneck Is the Data, Not the Model
Part I — The problemFour sectors, one bottleneckHospitals and health systems generate...
- [Launched] Generally Available: Azure Database for PostgreSQL flexible server in East US 3
Now you can deploy Azure Database for PostgreSQL flexible server in the East US 3 Azure region.
- Copilot code review: New organization billing options and controls
This release adds new billing and license controls for Copilot code review admins: Billing: Organization owners can bill Copilot code reviews from members with a Copilot license t…
- OpenAI GPT-6.1 Sol now supports Ultrafast mode on Amazon Bedrock
Today, AWS announces the availability of Ultrafast mode for GPT-6.1 Sol from OpenAI on Amazon Bedrock. Ultrafast mode provides faster inference for workloads where speed matt…
- Dynamic Workflows in Azure Functions Hosted Skills: Durable, AI-Led Work for Event-Driven Apps
Azure Functions is the event-driven application platform, and Hosted Skills adds AI reasoning to your app.
- Snowflake Decision: A Decision Model in Cortex AI Functions
Snowflake Decision, now in private preview, brings fast, structured classification, scoring and routing to SQL through AI_COMPLETE in Cortex AI Functions.
- Triage role users or higher can now archive pull requests
Users with the triage role or higher in a repository can now archive and unarchive pull requests.
- AWS Cost Explorer, Budgets, and Dashboards now support Amazon Bedrock product attributes
AWS Cost Explorer, AWS Budgets, and AWS Cost Management Dashboards now let you analyze Amazon Bedrock costs by product attributes, a new dimension that breaks down Bedrock cost by…
- Load terabytes of data in minutes into Lakebase Postgres
Operational databases like Postgres are built to reliably execute highly concurrent...
- How much does it cost to use open source software?
by Sophia Vargas, Google Open Source Open source software is used pervasively, with some estimating that over 90% of codebases contain open source software.
- Retirement: Microsoft Dev Box will be retired on September 18, 2028
Microsoft Dev Box will retire on September 18, 2028.
- Introducing Railway Authentication via identity-aware proxy
Railway Authentication implements an identity aware proxy at the service or environment level to lock down sensitive http endpoints
New Models
4 pieces.
- New model: StepFun: Step 5 Preview
Step 5 Preview is StepFun's flagship model for agentic work, built on a sparse Mixture-of-Experts architecture (27B active / 600B total parameters).
- Open model: openbmb/MiniCPM-V-4.7-35B-A3B-DSpark
New open weights from OpenBMB (MiniCPM) on the Hugging Face Hub.
- Open model: openbmb/MiniCPM-V-4.7-1B-DSpark
New open weights from OpenBMB (MiniCPM) on the Hugging Face Hub.
- New model: Upstage: Solar Decide Flash
Solar Decide Flash is Upstage's low-latency structured decision model, a faster variant of [Solar Decide](/upstage/solar-decide) built on Solar Mini 4 and served through the Syste…
Security Desk
82 pieces, the newest 10 here.
- Critical: Hazelcast allows arbitrary member memory access by low-privileged client
Affects com.hazelcast:hazelcast.
- Critical: fast-jwt: Incomplete patch of CVE-2026-34950: Non-whitespace key-prefix re-enables RSA→HS256 algorithm confusion
Affects fast-jwt.
- Critical: PraisonAI: AICoder Arbitrary File Write and Command Execution via LLM Tool Calls
Affects praisonai.
- Cybersecurity Imperatives Will Demand AI Math
Cybersecurity is built on math, and if public key encryption falls and we don’t have a replacement ready, the Web grinds to a halt.
- Post-quantum authentication: Why organizations should start testing certificate ecosystems now
Prepare for post-quantum authentication by testing certificate ecosystems now.
- USN-8910-1: libxml2 vulnerabilities
Yirou Yang discovered that libxml2 incorrectly handled certain XML catalogs.
- Threat Actors Exploit Critical AhsayCBS Flaws to Drop Webshells and XMRig Cryptominer
Threat actors are exploiting AhsayCBS flaws, including CVE-2026-105133 and CVE-2026-105134, to deploy webshells and XMRig cryptominers.
- Suspected TraderTraitor Group Uses Trojanized Terraform Provider to Deliver Cross-Platform Malware
IntroductionIn July 2026, Zscaler ThreatLabz uncovered a campaign linked to TraderTraitor (also tracked as Jade Sleet, UNC4899, Pressure Chollima, and Slow Pisces), an advanced pe…
- CVE-2026-107608: Improper link resolution in asset bundling output handling in aws-cdk-lib
Bulletin ID: 2026-131-AWS Scope: AWS Content Type: Important (requires attention) Publication Date: 10/08/2026 12:30 PM PDT Description: AWS CDK is an open source framework that a…
- CVE-2026-107322 - OS command injection in Amazon Agent Plugins for AWS databases-on-aws
Bulletin ID: 2026-130-AWS Scope: AWS Content Type: Important (requires attention) Publication Date: 10/08/2026 11:30 AM PDT Description: Amazon Agent Plugins for AWS is an open so…
Technical Guides
45 pieces, the newest 10 here.
- 5 Steps to Create SimReady Assets for Robotics with Frontier AI Models
Preparing CAD assets for robotics simulation requires more than converting geometry to OpenUSD: developers must configure and validate materials, collision...
- How we migrated Clera’s 500 GB production database to ClickHouse Managed Postgres overnight
Clera migrated its 500 GB production database and 500+ tables to ClickHouse Managed Postgres overnight, cutting CPU usage from 100% to 10–20%.
- Production-grade LLMs and agents: a field guide
A maturity model for taking agents from an impressive demo to a system people can depend on — with a self-assessment and the map to a deep-dive on each layer.
- Part 5: Operating an LLM system: observability, cost, routing, and the platform underneath
Your service can be 100% up and still quietly approving the wrong things, burning its budget, or failing over into untested quality.
- What Is ARIA, and How Did We Build It?
Putting an agent next to your experiments was the easy part.
- Data needs its DevOps moment
Discover how Turbolytics leverages Render Postgres to bring DevOps principles to data pipelines.
- Pay-per-inference for AI agents: How BlockRun and Incarna use Amazon Bedrock AgentCore payments
Amazon Bedrock AgentCore payments gives AI agents a managed way to pay for services on demand, with spending limits enforced by the infrastructure.
- Building Reliable Data Analytics Agents: Lessons from the KDD Cup
The NVIDIA KGMON team placed second in the KDD Cup 2026 Data Agents competition with a system built around a simple idea of making an agent's harness smaller,...
- Session-Aware Agentic Inference with NVIDIA Dynamo
TL;DR Agentic workloads change the traffic an inference server sees.
- Agents that can pay: building Restock with Stripe's Link and Managed Deep Agents
Build an agent that pays for real purchases.
AI in the Field
22 pieces, the newest 6 here.
- How Multi-Agent AI Is Changing the Way CPG and Retail Brands Manage Advertising at Scale
For CPG and retail brands managing hundreds of SKUs across seasonal peaks, Prime Day, back-to-school, and holiday windows, the operational burden compounds fast.
- Study in The Lancet suggests AI could improve patient-physician relationships.
A peer-reviewed study in The Lancet evaluates Google's AMIE medical AI in a real-world primary care clinic.
- 3 Questions: What is the best path forward for AI in academia?
MIT Statistics and Data Science Center Director Alexander (Sasha) Rakhlin shares important considerations for departments and institutions.
- 3 Questions: What is the best path forward for AI in academia?
MIT Statistics and Data Science Center Director Alexander (Sasha) Rakhlin shares important considerations for departments and institutions.
- 3 Questions: What is the best path forward for AI in academia?
MIT Statistics and Data Science Center Director Alexander (Sasha) Rakhlin shares important considerations for departments and institutions.
- Into the Omniverse: How Developers Turn Ideas Into Simulations With Frontier AI Agents
Turning a simulation idea into a working application means assembling assets, connecting physics and rendering, and checking that the scene behaves as intended.
Columns
13 pieces, the newest 6 here.
- [AINews] not much happened today
a quiet day.
- If your team is happy, are you doing a good job?
Brilliant jerks, ZIRP-era managers, and how psychological safety lost the plot.
- Beijing Will Not Pace the Frontier: China’s Speed-First AI Safety Regime
AI safety is on fire.
- The Pulse: new trend of building “internal vibe-coding apps” at tech companies
Also: do us engineers truly enjoy ‘hard problems’, or do we really just want to solve those which attract promotions, new open models launch in the EU and US, and more
- Synthesis Superintelligence: from Semiconductors to Superconductors — Periodic Labs’ Liam Fedus and Ekin Dogus Cubuk
A special Science pod and Engineering pod crossover..
- AI #189: New Math
The big drop of this week was not a new AI model.
Research
155 pieces, the newest 10 here.
- Using Claude Science to produce the first complete map of the sky in UV light
Claude Science helped produce the first complete map of the sky in UV light
- An opt-in vulnerability-finding service for open-source software
We’re launching OSS Scanner, an opt-in vulnerability scanner for the open-source ecosystem informed by our experience using Claude to find vulnerabilities during Project Glasswing.
- Dex-One2Many: Learning Dexterous Manipulation from a Single Human Demonstration
While learning dexterous manipulation from a single human video offers a promising alternative to costly robot demonstrations, many recent methods predominantly imitate demonstrat…
- Rubric-CEPR: Self-Evolving Image Editing via Reward-Verified Self-Distillation
Instruction-guided image editors have become highly capable, yet improving them further still depends on human-edited training pairs or external reward models.
- DreamTrue: Action-Faithful Robot World Model with Counterfactual Post-Training
We present DreamTrue, a multi-view, cross-embodiment robot world model for action-faithful and physically plausible video prediction.
- CSF: Contextual Safety Filtering for Motion Generators
Text-conditioned motion generators produce trackable whole-body motion, but they have no notion of scene-dependent safety: the same action may target an object or a person.
- On the estimation and validity of AI time horizons---a statistical look at the METR plot
METR's 50\% time horizon measures the human completion time of software tasks that an AI solves with 50\% probability, allowing AI capabilities to be expressed in interpretable un…
- A Balanced Data Diet: Addressing the Exploration Bottleneck in Mega-Scale RL for Robot Control
General-purpose robots must perform a wide range of tasks from agile locomotion to dexterous manipulation.
- From Reactive Containment to Proactive Assurance: Lessons from OpenAI, Anthropic, and Google Agent Security Incidents
In 2026, cybersecurity evaluations involving OpenAI, Anthropic, and Google agents reached real systems outside their authorized test scope.
- What 30,000 Hours of Ego-centric Video Does Not Teach
World models offer a promising alternative to physics-based simulators, yet remain far from practical deployment.
Releases
64 pieces, the newest 12 here.
- Temporal 1.32.1: Update defaultCliVersion to 1.9.0 by @simvlad in #12038
- langchain==1.4.4
- CrewAI 1.15.26: Fix ownership retention of rwlock when acquisition is interrupted
- langchain-core==1.6.9
- uv 0.12.24: Released on 2026-10-08
- Claude Code 2.1.295: Program Status Protocol (OSC 7501) support, terminals that implement it can show whether Claude Code is…
- opencode 2.0.26: release, v2.0.26
- OpenAI Codex 0.162.0: New Features
- Goose 1.54.0: Prompt for goose mode during first-time configure #12629
- @cloudflare/workers-utils@0.47.1
- wrangler@4.149.0
- @cloudflare/workers-auth@0.13.1
New Repos & Packages
33 pieces, the newest 8 here.
Watch
32 pieces, the newest 6 here.
- Building with Gemma 4
- Production Monitoring with Codex: Grafana, Kubernetes, & Security
- How to run GitHub Copilot agent sessions across any device | GitHub Copilot Day
- Beyond the Benchmark: How Parallel Web Systems Builds Reliable Web Agents
- WG AI Gateway Meeting for 2026-10-08
- How Cline Embraces an Open-Weights Model Ecosystem for AI Agents
Package releases
40 from the most used packages' own repositories, the newest 15 here.
- fastmcp v4.1.0 · every release →
- gradio@6.30.0 · every release →
- langsmith v0.14.5 · every release →
- pymongo 4.18.3 · every release →
- @smithy/util-utf8 @smithy/core@3.35.2 · every release →
- @aws-sdk/types v3.1148.0 · every release →
- aws/aws-sdk-php 3.400.0 · every release →
- fast-equals v6.1.1 · every release →
- claude-agent-sdk v0.2.165 · every release →
- tox 4.64.10 · every release →
- google-auth google-devicesandservices-health-v0.1.4 · every release →
- @sentry/core 10.76.2 · every release →
- predis/predis v3.7.0 · every release →
- cachebox v6.2.8 · every release →
- @grpc/proto-loader @grpc/grpc-js@1.14.6 · every release →