Tech news, September 29, 2026
Tuesday: 698 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.
News
124 pieces, the newest 20 here.
- Jev is now available in LangSmith Evals
Use Jev as a judge for LangSmith evals to evaluate agent traces with faster, cheaper structured feedback across production runs, datasets, and regression tests.
- What everyone was talking about at WeAreDevelopers World Congress North America
After 237 talks across eight stages, code review emerged as the bottleneck for AI-generated code.
- Vercel Connect now accepts service submissions
Vercel Connect now lets you submit your own service to the connectors directory, so you no longer need to wait for Vercel to build a connector for it.
- Why content extraction still matters in the GenAI era
Learn why stronger AI models make trustworthy content extraction more important, how Azure Content Understanding and Azure Document Intelligence serve different needs, and what co…
- Amazon Bedrock Managed Agents, powered by OpenAI, is now available in preview
Developed jointly by AWS and OpenAI, Bedrock Managed Agents (BMA) is built on a customized version of OpenAI's Agents API engineered to be AWS-native and integrated with AWS…
- Azure SDK Release (September 2026)
Azure SDK releases every month.
- OpenAI GPT-6.1 Sol is now generally available on Amazon Bedrock
Today, AWS announces the general availability of GPT-6.1 Sol from OpenAI on Amazon Bedrock.
- New Website for Jupyter Blog
The Jupyter blog has moved from Medium to a new website!
- Retirement: Functions on Azure Container Apps V1
The Functions on Azure Container Apps (ACA) V1 will be retired on September 29, 2027.
- Bring near-Astra intelligence to everyday work with GPT-6.1 Sol on Amazon Bedrock
GPT-6.1 Sol is now generally available on Amazon Bedrock, bringing stronger reasoning to coding, computer use, and professional workloads that run frequently.
- [Launched] Generally Available: Storage optimized Lasv5 and Laosv5 Azure VM series
The Laosv5 series and Lasv5 storage optimized VMs, based on the 5th Gen AMD EPYC™ processor (Turin), are now generally available.Lasv5 and Laosv5 VMs have sizes ranging from 2-160…
- Repository custom runner settings for Dependabot
As a repository administrator, you can now configure the runner type, optional custom label, and optional runner group for Dependabot version and security updates.
- Bring business context with external custom properties
You can now seamlessly bring business context about your repositories from an external system of record (e.g., a configuration management database (CMDB), an internal developer po…
- Lower the Cost of Building and Running Visual AI Agents with NVIDIA VSS Blueprint 3.3
Vision-language models have made it possible to build visual AI agents that understand video at production scale.
- [In preview] Public Preview: Azure Database for PostgreSQL Ultra Disk
Ultra Disk support is now available in public preview for Azure Database for PostgreSQL flexible server.
- Announcing our partnership with OpenAI
Baseten is partnering with OpenAI to serve open models natively via Codex and the Responses API.
- Sign in to Warp with ChatGPT
Use your ChatGPT subscription for AI requests in Warp Terminal and Warp Agent CLI.
- Introducing Sign in with ChatGPT in Kilo Code
Kilo Code now supports Sign in with ChatGPT, letting developers bring their existing OpenAI plan to every Kilo surface, including Cloud Agents and mobile.
- Most AI Context Layers Are an Inch Off. That's a Mile Wrong.
Most AI context layers miss the hardest part: runtime context.
- [In preview] Public Preview: Azure SQL Database Hyperscale Serverless auto-pause and auto-resume
Auto-pause and auto-resume are now available in public preview for Azure SQL Hyperscale databases using the Serverless compute tier.
New Models
5 pieces.
- New model: OpenAI: GPT-6.1 Sol Pro
GPT-6.1 Sol Pro is the same underlying model as [GPT-6.1 Sol](https://openrouter.ai/openai/gpt-6.1-sol), served with `reasoning.mode` set to `pro` for higher-quality responses on…
- New model: OpenAI: GPT-6.1 Sol
GPT-6.1 Sol is an upgrade to GPT-6 Sol from OpenAI, positioned below the flagship GPT-6 Astra in the GPT-6 series.
- Open model: kyutai/ovie-512
New open weights from Kyutai on the Hugging Face Hub.
- Open model: PSRben/VisionHOPE
Trending on the Hugging Face Hub.
- Open model: internlm/AdvancedMathBench-AutoVerifier
New open weights from InternLM (Shanghai AI Lab) on the Hugging Face Hub.
Security Desk
75 pieces, the newest 10 here.
- Critical: @xhmikosr/decompress: Path traversal via symlink chain
Affects @xhmikosr/decompress, decompress.
- Critical: PyJWT: Asymmetric-PEM detection bypass: whitespace/line-ending-mutated public keys skip the HS/asymmetric confusion guard
Affects PyJWT.
- Phishing Abuses RMM Tools for Persistent Access
Microsoft observed phishing campaigns that abused MSP360 RMM to deploy ScreenConnect, creating redundant remote-access channels for follow-on activity The post Phishing Abuses RMM…
- CISA Adds One Known Exploited Vulnerability to Catalog
CISA Adds One Known Exploited Vulnerability to Catalog ggaylor Sep 29, 2026 Release Date September 29, 2026 Description CISA has added one new vulnerability to its Known Exploited…
- Stable Channel Update for Desktop
The Stable channel has been updated to 154.0.8037.92/.93 for Windows and Mac and 154.0.8037.92 to Linux which will roll out over the coming days/weeks.
- CVE-2026-94053: Apache MINA SSHD: LDAP injection in sshd-ldap
Posted by Thomas Wolf on Sep 29 Severity: critical CVSS 3.1: 9.1 (critical) CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N Affected versions: - Apache MINA SSHD 1.2.0 before 2.20.0…
- CVE-2026-94052: Apache MINA SSHD: LDAP password authentication ineffective
Posted by Thomas Wolf on Sep 29 Severity: critical CVSS 3.1: 9.1 (critical) CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N Affected versions: - Apache MINA SSHD 1.2.0 before 2.20.0…
- USN-8847-1: OpenSSL vulnerabilities
It was discovered that OpenSSL incorrectly handled certain certificate revocation list distribution point names.
- CVE-2026-93994: Apache MINA SSHD: Repeated-publickey policy bypass on server
Posted by Thomas Wolf on Sep 29 Severity: important CVSS 3.1: 8.1 (high) CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:N Affected versions: - Apache MINA SSHD before 2.20.0 - Apache…
- CVE-2026-77185: Apache MINA SSHD: Asynchronous authentication can bypass signature verification
Posted by Thomas Wolf on Sep 29 Severity: critical CVSS 3.1: 9.1 (critical) CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N Affected versions: - Apache MINA SSHD (org.apache.sshd:ssh…
Technical Guides
55 pieces, the newest 10 here.
- Jev vs. LLM-as-a-Judge for AI Evals
A model that answers yes/no questions instead of writing text came out 3.5× cheaper and 3.8× faster than our gpt-4o-mini judge on 1,000 real production turns, and agreed with it o…
- Build Your AI Dev Factory Your Way
Learn how an AI Dev Factory turns a fragmented AI stack into a repeatable, secure system for building, testing, delivering, and operating AI at scale.
- OpenTelemetry as an Enterprise Data Asset | Snowflake
Learn how to turn telemetry into a governed enterprise data asset.
- Automate RCA across ServiceNow, Dynatrace and Slack with AWS DevOps Agent
If you manage production incidents, you know the drill.
- AI Native by Design: Lessons Learned from Building NVIDIA TensorRT Model Connect
Parallel work, model-family isolation, reversible changes, and GPU-backed validation shaped an open source project designed around coding agents NVIDIA TensorRT...
- How Diffusion Controller unifies and simplifies AI image generation
Algorithms & Theory
- How MIXI built natural language search over billions of photos with Amazon S3 Vectors
MIXI, Inc.
- Implementing customer managed keys for AWS Lambda durable functions with Terraform
Lambda durable functions checkpoint execution state to durable storage, and for regulated payment workloads that data is sensitive.
- Personalization without user identity
How Airbnb uses proximity signals to personalize without relying on individual user history.
- WSLC Architecture deep dive
WSL containers is now generally available!
AI in the Field
9 pieces, the newest 6 here.
- Waller, Payments in the Age of AI Agents
Speech At Sibos 2026, Miami, Florida
- Presenting Coursera’s Global Skills Report 2026, and our new AI-Human Skills Synergy Index—the first of its kind in the world
By Anthony Salcito, SVP & President, Enterprise, Coursera The rate of AI deployment is outpacing the rate at which people are gaining the skills needed to use it effectively.
- The effects of an “algorithmic monoculture” depend on the details
In a study focusing on hiring decisions, MIT researchers found the use of a single algorithm by many firms could benefit job seekers in certain situations.
- Who we become when we talk to machines
Professor Sherry Turkle’s new book, “Artificial Intimacy,” offers a withering critique of chatbots and the antisocial dynamics she believes they encourage.
- Development and validation of a machine learning model for the preoperative prediction of intractable hypoxemia in repeat lung surgery
npj Digital Medicine, Published online: 29 September 2026; doi:10.1038/s41746-026-03297-8 Development and validation of a machine learning model for the preoperative prediction of…
- DentalGEN: a large-scale controllable generative AI framework for automated dental crown restoration
npj Digital Medicine, Published online: 29 September 2026; doi:10.1038/s41746-026-03236-7 DentalGEN: a large-scale controllable generative AI framework for automated dental crown…
Columns
13 pieces, the newest 6 here.
- Astra 6.1 Pulled As Insufficiently Aligned
We once again got a new set of warnings yesterday, and new movement towards living in a sane world.
- OpenAI DevDay 2026 live blog
I'm at OpenAI DevDay today, in Fort Mason, San Francisco.
- Why has Shopify dropped React Native?
It’s only been a year since the e-commerce platform declared it was very happy with React Native, but now Shopify is dumping it – and the reason, unsurprisingly, is AI
- As a general rule, calling product support while drunk is not recommended
Swearing is also poor form.
- Bliki: Sensible Default
A Sensible Default is a practice that, absent some overriding context, should be used when carrying out a certain kind of task.
- Fragments: September 29
Simon Willison: The more time I spend working with coding agents, the more convinced I am that they make software engineering even harder We can do amazing things with them, but u…
Research
177 pieces, the newest 10 here.
- Z-Sigil: A Public-Key Cryptosystem with Chained Selection over a Fiber Bundle of Module-Lattice Keys
Z-Sigil is a public-key cryptosystem in which the plaintext selects successive keys from a fixed module-lattice family.
- StateFork: Branchable Infrastructure for Agent Exploration
AI agents improve task success by exploring multiple trajectories, but for computer-use agents each trajectory modifies external environment state.
- Alignment via Training Against Probes Without Losing Monitorability
Models are usually aligned based on their observed outputs, using demonstrations, preference data, or reward signals.
- Enabling Efficient Client Selection in FL-as-a-Service for Multi-Application based Society 5.0
The rapid development of the Internet of Things (IoT) has led to the generation of vast amounts of data from sensors, prompting the need for advanced learning models to analyze th…
- SecureVibe: Making Vibe Coding More Secure
As vibe coding becomes increasingly capable and widespread, security vulnerabilities in even functionally correct solutions are a growing concern.
- OpenAI stops training latests models, citing safety concerns
CSET’s Helen Toner spoke with NPR’s Steve Inskeep about OpenAI’s decision to pause testing and training of its latest model amid safety concerns.
- China’s AI agents can lie and scheme – just like their US rivals
CSET’s Colin Shea-Blymyer shared his expert insight in an article published by Reuters.
- RISK: Auditing Industrial Control Systems for Too-Late-to-Recover Vulnerabilities
The security of industrial control systems (ICS) is important.
- The Mode of Null-A: Compositional Computation of a Generalized Inverse
We present a novel algorithm for calculating the preimage of an affine space through a product ${J}={J}_{T-1}\cdots{J}_0$ of matrices ${J}_t$ of special form: finding the largest…
- Security-Enhanced Seed-Based Weight Quantization for Large Language Models
Large language models (LLMs) incur substantial storage, memory-bandwidth and energy costs, motivating compact weight representations.
Releases
82 pieces, the newest 12 here.
New Repos & Packages
43 pieces, the newest 8 here.
Watch
21 pieces, the newest 6 here.
Package releases
94 from the most used packages' own repositories, the newest 15 here.
- cron-descriptor 2.1.1 · every release →
- virtualenv 21.14.1 · every release →
- xarray v2026.09.0 · every release →
- filelock 4.0.7 · every release →
- pikepdf v10.16.0 · every release →
- lucide-react 1.49.0 · every release →
- sass 1.105.1 · every release →
- pbs-installer 2026.09.29 · every release →
- postcss-discard-duplicates cssnano@9.1.2 · every release →
- pyspnego v0.12.3 · every release →
- pylint v4.1.1 · every release →
- memfs v4.80.0 · every release →
- virtualenv 21.14.0 · every release →
- claude-agent-sdk v0.2.162 · every release →
- pylint v4.1.0 · every release →