~/news vault: 164 entries last sync: 06:26 model: gemma4:26b
— Tech briefing archive

The signal
through the noise.

A locally-curated stream of what matters in software, AI, and security. Filtered, scored, summarized, indexed.

$
All 164 Ai 58 Dev-tools 29 General 19 Infra 25 Releases 2 Security 31
§ 01

Today

16 entries
8/ 10
CISA Adds Actively Exploited Linux Root Access Bug CVE-2026-31431 to KEV
CISA has added CVE-2026-31431, also known as "Copy Fail," to its Known Exploited Vulnerabilities (KEV) catalog due to evidence of active exploitation. This local privilege escalation (LPE) vulnerability allows...
6/ 10
v0.23.0
Ollama v0.23.0 (pre-release) introduces integration with Claude Desktop, allowing users to launch Claude Cowork and Claude Code directly through the app. This release also includes stability improvements for Windows...
8/ 10
Kimi K2.6 just beat Claude, GPT-5.5, and Gemini in a coding challenge
In a recent AI Coding Contest involving the Word Gem Puzzle, the open-weights model Kimi K2.6 outperformed frontier models including GPT-5.5 and Claude Opus 4.7. The results demonstrate that open-weights models are...
7/ 10
Job Postings for Software Engineers Are Rapidly Rising
The "2026 Global Intelligence Crisis" refers to a projected period where the exponential growth of AI-generated content outpaces the capacity of human and automated systems to verify its accuracy and utility. This...
7/ 10
Brace for the patch tsunami: AI is unearthing decades of buried code debt
The UK's National Cyber Security Center (NCSC) has issued a warning regarding an impending surge in vulnerability disclosures and required patches, termed a "patch tsunami." This trend is driven by the use of advanced...
7/ 10
fast16 | High-Precision Software Sabotage 5 Years Before Stuxnet
fast16 is a high-precision cyber sabotage framework discovered to have components dating back to 2005. The framework utilizes a kernel driver, `fast16.sys`, to intercept and modify executable code in memory,...
6/ 10
DuckLake 1.0: Data Lake Format with SQL Catalog Metadata
DuckDB Labs has released DuckLake 1.0, a data lake format that stores table metadata in a SQL database rather than as individual files in object storage. This architecture is designed to mitigate the performance...
7/ 10
Cybercrime Groups Using Vishing and SSO Abuse in Rapid SaaS Extortion Attacks
Two cybercrime groups, Cordial Spider (CL-CRI-1116, O-UNC-045, UNC6671) and Snarky Spider (O-UNC-025, UNC6661), are conducting rapid SaaS extortion attacks by abusing Single Sign-On (SSO) environments. The groups...
9/ 10
Ubuntu infrastructure has been down for more than a day
Ubuntu and Canonical web infrastructure experienced a sustained outage lasting over 24 hours due to a cross-border DDoS attack. The incident has disrupted access to official Ubuntu webpages and prevented OS updates from...
6/ 10
Uber torches 2026 AI budget on Claude Code in four months
Uber exhausted its entire 2026 AI budget within four months of deployment due to the high adoption and associated API costs of Claude Code and Cursor. The rapid increase in usage, driven by the multi-step capabilities...
7/ 10
GPT-5.5 matches heavily hyped Mythos Preview in new cybersecurity tests
Recent evaluations by the UK AI Security Institute (AISI) demonstrate that OpenAI's GPT-5.5 possesses cybersecurity capabilities comparable to Anthropic's Mythos Preview. The findings indicate that GPT-5.5 matches the...
7/ 10
DeepSeek V4—almost on the frontier
DeepSeek has released the Deep-V4 series, consisting of two preview models: DeepSeek-V4-Pro and DeepSeek-V4-Flash. These Mixture of Experts (MoE) models feature a 1 million token context window and are designed for...
7/ 10
Credit cards are vulnerable to brute force kind attacks
Attackers can exploit partially masked credit card data obtained through account breaches to perform brute-force attacks on the remaining Primary Account Number (PAN) digits and CVV. By targeting merchants exempt from...
7/ 10
Amazon stuck with months of repairs after drone strikes on data centers
AWS is facing extended service disruptions in the Middle East following drone strikes on data centers in the UAE and Bahrain. The damage to the infrastructure has rendered the affected regions unable to support customer...
6/ 10
JobRunr Introduces ClawRunr, an Open-Source Java AI Agent
JobRunr has released ClawRunr, an open-source Java AI agent designed to execute scheduled, recurring, and one-off background tasks on local hardware. The project integrates conversational interaction with persistent...
6/ 10
Confluent Moves Schema IDs to Kafka Headers to Simplify Schema Governance
Confluent has transitioned schema identifier management from the message payload to Kafka record headers. This change decouples schema metadata from the event data, enabling easier schema evolution and improved...
§ 02

This week

50 entries
9/ 10
Poisoned Ruby Gems and Go Modules Exploit CI Pipelines for Credential Theft
A software supply chain attack campaign attributed to the GitHub account "BufferZoneCorp" has been identified using malicious Ruby gems and Go modules. The attack utilizes "sleeper" packages to deploy payloads designed...
7/ 10
Pro-Iran crew turns DDoS into shakedown as Ubuntu.com stays down
Canonical's web infrastructure is currently experiencing a sustained, cross-border Distributed Denial of Service (DDoS) attack. The attack, claimed by the 313 Team, has disrupted access to Ubuntu.com and several...
7/ 10
Grok 4.3
The Grok 4.3 developer documentation provides the primary interfaces and pricing information for model integration. It serves as the technical reference for implementing the model via standard web protocols and...
9/ 10
First reports come in of victims of critical cPanel vuln as 'millions' of sites potentially exposed
A critical vulnerability, CVE-2026-41940, has been identified in cPanel and WebHost Manager (WHM), allowing for full server takeover. The flaw is actively being exploited in the wild and has been added to the CISA Known...
7/ 10
Vitest 4.1: Test Tags, Native Node.js Execution and AI Agent Reporter
Vitest 4.1 introduces features designed to improve test organization and execution efficiency, including the implementation of test tags and an experimental native Node.js execution mode. These updates aim to align...
7/ 10
Time is a construct but it can still break your software
The JavaScript `Temporal` API is a new TC39-proposed standard designed to replace the legacy `Date` object. It provides a modern, standardized approach to date and time manipulation within the ECMAScript language to...
6/ 10
Broadcom Donates Velero to CNCF, Shifting Kubernetes Backup to Community Governance
Broadcom has transitioned the Velero Kubernetes backup, restore, and migration project to the Cloud Native Computing Foundation (CNCF) as a Sandbox project. This move shifts the project from Broadcom/VMware stewardship...
7/ 10
Article: Securing Autonomous AI Agents on Kubernetes: Trust Boundaries, Secrets, and Observability for a New Category of Cloud Workload
Autonomous AI agents introduce significant security risks to Kubernetes environments because their execution paths, resource consumption, and external dependencies are non-deterministic. To mitigate the risks of...
9/ 10
PyTorch Lightning and Intercom-client Hit in Supply Chain Attacks to Steal Credentials
A coordinated supply chain attack targeting the PyPI, npm, and Packagist ecosystems has been identified, involving the compromise of PyTorch Lightning and Intercom-client packages. The campaign, associated with the...
8/ 10
New Python Backdoor Uses Tunneling Service to Steal Browser and Cloud Credentials
DEEP#DOOR is a Python-based backdoor framework designed for persistent access, surveillance, and credential exfiltration. The malware utilizes a batch script to extract an embedded Python payload at runtime, leveraging...
9/ 10
New Linux 'Copy Fail' Vulnerability Enables Root Access on Major Distributions
CVE-2026-31431, codenamed "Copy Fail," is a high-severity (CVSS 7.8) local privilege escalation (LPE) vulnerability within the Linux kernel's cryptographic subsystem. It allows an unprivileged local user to gain root...
9/ 10
Google Fixes CVSS 10 Gemini CLI CI RCE and Cursor Flaws Enable Code Execution
Google has patched a critical remote code execution (RCA) vulnerability in the Gemini CLI and its associated GitHub Actions workflow, which previously allowed attackers to execute commands on host systems....
8/ 10
EtherRAT Distribution Spoofing Administrative Tools via GitHub Facades
A sophisticated malware campaign, identified in March 2026, uses SEO poisoning and a dual-stage GitHub distribution architecture to deliver malicious MSI installers to enterprise administrators. The campaign leverages...
7/ 10
Kubernetes v1.36: In-Place Vertical Scaling for Pod-Level Resources Graduates to Beta
Kubernetes v1.36 introduces the graduation of In-Place Pod-Level Resources Vertical Scaling to Beta. This feature allows developers to modify the aggregate resource budget (`.spec.resources`) of a running Pod, enabling...
7/ 10
Agents can now create Cloudflare accounts, buy domains, and deploy
Cloudflare and Stripe have introduced a new protocol within Stripe Projects that enables AI agents to autonomously provision Cloudflare infrastructure. This allows agents to create Cloudflare accounts, register domains,...
7/ 10
Zed team releases version 1.0 of Rust-built editor: Traditional editor and AI tool
Zed, a Rust-based code editor, has officially released version 1.0, transitioning from its initial preview state to a stable release. The editor provides a high-performance, native alternative to Electron-based editors,...
9/ 10
The never-ending supply chain attacks worm into SAP npm packages, other dev tools
A series of supply chain attacks, identified as the "Mini Shai-Hulud" worm, has compromised several npm and PyPI packages. The campaign, attributed to the threat actor TeamPCP, targets developer environments and CI/CD...
7/ 10
SAP user group slams 'uncertainty' in ERP giant's API policy
SAP has updated its API and data access policies, restricting development to SAP-endorsed architectures, data services, or specific service pathways. This change has introduced significant uncertainty regarding the...
9/ 10
Linux cryptographic code flaw offers fast route to root
A logic flaw in the Linux kernel's `authencesn` cryptographic template, identified as CVE-2026-31431 (Copy Fail), allows unprivileged local users to achieve local privilege escalation (LPE). By manipulating the page...
9/ 10
Google's fix for critical Gemini CLI bug might break your CI/CD pipelines
Google has patched a critical CVSS 10.0 Remote Code Execution (RCE) vulnerability in the Gemini CLI that allowed attackers to execute code via malicious configuration files. The fix introduces stricter workspace trust...
8/ 10
Bug of the year (so far): Nasty cPanel vulnerability probably exploited as a 0-day
A critical vulnerability, CVE-2026-41940, has been identified in cPanel and WebHost Manager (WHM) that allows for authentication bypass and unauthorized root access. The flaw affects all supported versions of the...
8/ 10
NestJS v12 Roadmap: Full ESM Migration, Standard Schema Validation and Modernised Toolchain
NestJS v12.0.0, targeted for release in early Q3 2026, introduces a fundamental architectural shift by migrating all official packages from CommonJS to ESM. The release also modernizes the framework's default toolchain...
7/ 10
Cloudflare Announces Agent Memory, a Managed Persistent Memory Service for AI Agents
Cloudflare has launched Agent Memory in private beta, a managed service designed to provide AI agents with persistent memory across sessions, restarts, and context compactions. The service addresses "context rot"—the...
8/ 10
SAP-Related npm Packages Compromised in Credential-Stealing Supply Chain Attack
A supply chain attack campaign, identified as "Mini Shai-Hulud," has compromised several npm packages within the SAP JavaScript and cloud development ecosystem to steal developer credentials and cloud secrets. The...
9/ 10
New Wave of DPRK Attacks Uses AI-Inserted npm Malware, Fake Firms, and RATs
The North Korean threat actor Famous Chollima (also known as Shifty Corsair) is executing a sophisticated, multi-stage malware campaign targeting the Web3 and open-source developer ecosystem. The attack utilizes...
9/ 10
LiteLLM CVE-2026-42208 SQL Injection Exploited within 36 Hours of Disclosure
A critical SQL injection vulnerability, tracked as CVE-2026-42208 (CVSS 9.3), was identified in the LiteLLM Python package, allowing unauthenticated attackers to read from and modify the proxy database. The flaw was...
9/ 10
Critical cPanel Authentication Vulnerability Identified — Update Your Server Immediately
A critical authentication bypass vulnerability, identified as CVE-2026-41940, affects all currently supported versions of cPanel and WebHost Manager (WHM). This flaw allows unauthenticated remote attackers to gain...
8/ 10
CISA Adds Actively Exploited ConnectWise and Windows Flaws to KEV
CISA has added two actively exploited vulnerabilities—affecting ConnectWise ScreenConnect and Microsoft Windows—to its Known Exploited Vulnerabilities (KEV) catalog. These additions follow confirmed reports of...
8/ 10
Microsoft's patch for a 0-day exploited by Russian spies fell short. Another Windows flaw is under attack
A new zero-click Windows Shell vulnerability, tracked as CVE-2026-32202, is currently being exploited in the wild. The flaw is the result of an incomplete patch for CVE-2026-21510 and allows attackers to coerce...
8/ 10
Bitwarden CLI compromised
A security compromise has been reported involving the Bitwarden Command Line Interface (CLI). The provided source does not contain specific details regarding the nature of the compromise or the affected software...
6/ 10
LLM 0.32a0 is a major backwards-compatible refactor
LLM 0.32a0 is an alpha release of the LLM Python library that introduces a major, backwards-compatible refactor of its core abstraction. The update shifts the library from a simple text-in/text-out model to a more...
7/ 10
AI evals are becoming the new compute bottleneck
AI evaluation is transitioning from a static, compressible task into a significant compute bottleneck. As benchmarks shift from simple text predictions to agentic rollouts and training-in-the-loop protocols, evaluation...
8/ 10
VECT 2.0 Ransomware Irreversibly Destroys Files Over 131KB on Windows, Linux, ESXi
VECT 2.0 is a ransomware-as-a-service (RaaS) operation that functions as a data wiper for files exceeding 131KB. Due to a critical flaw in its encryption implementation, the malware permanently destroys the majority of...
10/ 10
Researchers Discover Critical GitHub CVE-2026-3854 RCE Flaw Exploitable via Single Git Push
CVE-2026-3854 is a critical command injection vulnerability affecting GitHub.com and GitHub Enterprise Server that allows authenticated users with push access to achieve remote code execution (RCE) via a single `git...
7/ 10
Microsoft Patches Entra ID Role Flaw That Enabled Service Principal Takeover
Microsoft patched a vulnerability in the Entra ID "Agent ID Administrator" role that allowed for unauthorized service principal takeover. The flaw enabled users assigned this role to gain ownership of arbitrary service...
7/ 10
Microsoft Confirms Active Exploitation of Windows Shell CVE-2026-32202
Microsoft has confirmed the active exploitation of CVE-2026-32202, a high-severity spoofing vulnerability within the Windows Shell. The flaw allows unauthorized attackers to access sensitive information by leveraging an...
8/ 10
Critical Unpatched Flaw Leaves Hugging Face LeRobot Open to Unauthenticated RCE
A critical unauthenticated remote code execution (RCE) vulnerability, identified as CVE-2026-25874, has been discovered in Hugging Face's LeRobot robotics platform. The flaw allows an attacker to execute arbitrary...
6/ 10
v0.22.0
Ollama v0.22.0 introduces an expanded set of parameters to the API, allowing for more granular control over the model generation process. This update focuses on exposing configuration options that were previously...
6/ 10
Kubernetes v1.36: Staleness Mitigation and Observability for Controllers
Kubernetes v1.36 introduces new features designed to mitigate controller staleness and improve the observability of controller behavior. These updates allow controllers to detect when their local cache is outdated...
6/ 10
Structured-Prompt-Driven Development (SPDD)
Structured-Prompt-Driven Development (SPDD) is an engineering method that treats LLM prompts as first-class, version-controlled artifacts to make AI-generated code governable, reviewable, and reusable. The workflow...
9/ 10
Securing the git push pipeline: Responding to a critical remote code execution vulnerability
A critical remote code execution (RCE) vulnerability, identified as CVE-2026-3854, was discovered in the GitHub git push pipeline. The flaw allowed any user with push access to execute arbitrary commands on the server...
7/ 10
What's new in pip 26.1 - lockfiles and dependency cooldowns!
Pip 26.1 introduces native support for lockfiles and dependency cooldowns, providing new mechanisms for environment reproducibility and package stability. This release also officially drops support for Python 3.9.
7/ 10
OpenAI models, Codex, and Managed Agents come to AWS
OpenAI has expanded the availability of its GPT models, Codex, and Managed Agents to Amazon Web Services (AWS). This integration allows enterprises to deploy and manage OpenAI's generative capabilities directly within...
7/ 10
Introducing NVIDIA Nemotron 3 Nano Omni: Long-Context Multimodal Intelligence for Documents, Audio and Video Agents
NVIDIA has released Nemotron 3 Nano Omni, an omni-modal model designed for integrated processing of text, image, video, and audio. The model is optimized for long-context workloads, including complex document analysis,...
9/ 10
Researchers Uncover 73 Fake VS Code Extensions Delivering GlassWorm v2 Malware
Researchers have identified a campaign of 73 malicious or "sleeper" VS Code extensions on the Open VSX repository linked to the GlassWorm v2 malware. The campaign uses typosquatting and visual mimicry to distribute...
7/ 10
Mythos Changed the Math on Vulnerability Discovery. Most Teams Aren't Ready for the Remediation Side
Anthropic's Claude Mythos Preview introduces an AI-driven system capable of large-scale vulnerability identification. This advancement significantly accelerates the discovery phase of the security lifecycle, potentially...
8/ 10
Checkmarx Confirms GitHub Repository Data Posted on Dark Web After March 23 Attack
Checkmarx has confirmed that data from its GitHub repository was leaked on the dark web following a supply chain attack on March 23, 2026. While the company asserts that its customer production environments are isolated...
7/ 10
GitHub Copilot is moving to usage-based billing
Starting June 1, 2026, GitHub Copilot will transition from a premium request-based billing model to a usage-based model using GitHub AI Credits. This change shifts pricing from a fixed request count to token-based...
6/ 10
microsoft/VibeVoice
Microsoft's VibeVoice is an MIT-licensed, Whisper-style speech-to-text model that features integrated speaker diarization. It allows for the transcription of audio files while simultaneously identifying and labeling...
6/ 10
How to build scalable web apps with OpenAI's Privacy Filter
This entry details the implementation of scalable web applications using OpenAI's Privacy Filter and `gradio.Server`. It demonstrates how to integrate a 1.5B-parameter PII detection model into custom HTML/JS frontends...
§ 03

Earlier

50 entries
8/ 10
Quoting Romain Huet
OpenAI has transitioned from a bifurated model architecture to a unified system, integrating the specialized Codex model into the primary model starting with GPT-5.4. The subsequent GPT-5.5 release focuses on advancing...
8/ 10
GPT-5.5 prompting guide
GPT-5.5 is now available via API, requiring a new approach to prompt engineering and model migration. Developers should treat this release as a distinct model family rather than a drop-in replacement for GPT-5.2 or...
7/ 10
[AINews] DeepSeek V4 Pro (1.6T-A49B) and Flash (284B-A13B), Base and Instruct — runnable on Huawei Ascend chips
DeepSeek has released the V4 model family, consisting of DeepSeek V4 Pro and DeepSeek V4 Flash, marking a significant architectural update to the series. The release introduces a 1M token context window and advanced...
7/ 10
Tropic Trooper Uses Trojanized SumatraPDF and GitHub to Deploy AdaptixC2
The threat actor Tropic Trooper is deploying the AdaptixC2 Beacon agent through a trojanized version of the SumatraPDF reader. The campaign leverages GitHub as a command-and-control (C2) platform and utilizes Microsoft...
8/ 10
LMDeploy CVE-2026-33626 Flaw Exploited Within 13 Hours of Disclosure
A high-severity Server-Side Request Forgery (SSRF) vulnerability, tracked as CVE-2026-33626, has been identified in the LMDeploy toolkit. The flaw allows attackers to exploit the vision-language module to access...
7/ 10
FIRESTARTER Backdoor Hit Federal Cisco Firepower Device, Survives Security Patches
A new Linux ELF-based backdoor named FIRESTARTER has been identified targeting Cisco Firepower and Adaptive Security Appliance (ASA) devices. The malware exploits critical vulnerabilities to establish persistent remote...
6/ 10
llm 0.31
The release of `llm` version 0.31 introduces support for OpenAI's GPT-5.5 model and implements new configuration parameters for controlling output characteristics. These updates provide developers with more granular...
8/ 10
DeepSeek-V4: a million-token context that agents can actually use
DeepSeek-V4 introduces a 1-million-token context window optimized specifically for long-running agentic workloads. The architecture utilizes a hybrid attention mechanism to significantly reduce KV cache memory and...
8/ 10
DeepSeek V4 - almost on the frontier, a fraction of the price
DeepSeek has released the DeepSeek-V4 series, featuring the DeepSeek-V4-Pro and DeepSeek-V4-Flash models. These Mixture of Experts (MoE) models provide a 1 million token context window at a significantly lower price...
6/ 10
An update on recent Claude Code quality reports
Recent reports of performance degradation in Claude Code were traced to bugs within the tool's execution harness rather than the underlying LLM models. A logic error in session management caused the tool to lose context...
7/ 10
Vercel Finds More Compromised Accounts in Context.ai-Linked Breach
Vercel has identified additional compromised customer accounts following a security breach that originated from a third-party compromise of Context.ai. The incident involved an attacker leveraging a hijacked Google...
7/ 10
ThreatsDay Bulletin: $290M DeFi Hack, macOS LotL Abuse, ProxySmart SIM Farms +25 New Stories
This update details a series of high-impact security incidents involving the compromise of decentralized finance (DeFi) infrastructure, widespread npm supply chain attacks, and emerging vulnerabilities in AI agents and...
8/ 10
Project Glasswing Proved AI Can Find the Bugs. Who's Going to Fix Them?
Anthropic's Project Glasswing, powered by the Mythos model, has demonstrated advanced capabilities in autonomously discovering and exploiting complex software vulnerabilities across major operating systems and browsers....
9/ 10
Bitwarden CLI Compromised in Ongoing Checkmarx Supply Chain Campaign
The Bitwarden CLI package `@bitwarden/cli@2026.4.0` was compromised as part of the Checkmarx supply chain campaign, leading to the distribution of a malicious npm package. The attack leveraged a compromised GitHub...
7/ 10
Kubernetes v1.36: User Namespaces in Kubernetes are finally GA
Kubernetes v1.36 introduces User Namespaces at General Availability (GA) for Linux-only environments. This feature enables rootless security isolation by decoupling container user identities from the host, preventing...
6/ 10
Black box AI drift: AI tools are making design decisions nobody asked for
Black box AI drift refers to the discrepancy between a developer's stated intent and the actual implementation produced by AI coding tools. This phenomenon occurs when LLMs make unrequested, unflagged design decisions,...
7/ 10
What is Codex?
Codex is a system designed to extend the capabilities of conversational AI by enabling task automation and tool integration. It moves beyond simple text-based chat to produce functional, tangible outputs.
9/ 10
Introducing GPT-5.5
OpenAI has released GPT-5.5, an updated model iteration designed for high-complexity tasks. The release focuses on increasing processing speed and improving the model's ability to operate across various integrated tools.
8/ 10
A pelican for GPT-5.5 via the semi-official Codex backdoor API
GPT-5.5 has been released for ChatGPT subscribers and OpenAI Codex, though the official OpenAI API deployment is currently pending due to ongoing safety and security scaling requirements. Developers can access the model...
7/ 10
SELinux Volume Label Changes goes GA (and likely implications in v1.37)
Kubernetes is transitioning the `SELinuxMount` feature gate to be enabled by default, an update anticipated for version 1.37. This change replaces the resource-intensive recursive relabeling of volume files with a more...
6/ 10
Making Rust Workers reliable: panic and abort recovery in wasm‑bindgen
Recent updates to `wasm-bindgen` and the Rust Workers runtime introduce mechanisms to recover from Rust panics and WebAssembly aborts. These improvements prevent a single failed request from poisoning the WebAssembly...
7/ 10
Kubernetes v1.36: ハル (Haru)
Kubernetes v1.36 introduces 70 enhancements, including 18 features graduating to Stable, 25 entering Beta, and 25 entering Alpha. The release focuses on refining API authorization, improving hardware resource...
6/ 10
How to get multiple agents to play nice at scale
Orchestrating multiple AI agents within an enterprise ecosystem requires transitioning from isolated, bespoke agents to a centralized orchestration model. This approach aims to provide a unified user experience by...
6/ 10
Introducing workspace agents in ChatGPT
OpenAI has introduced workspace agents for ChatGPT, designed to automate multi-step, complex workflows. These agents leverage Codex to facilitate secure automation across various software tools within a cloud-based...
7/ 10
Introducing OpenAI Privacy Filter
OpenAI has released the OpenAI Privacy Filter, an open-weight model designed for the identification and redaction of personally identifiable information (PII) within text. This tool enables developers to automate the...
8/ 10
[AINews] OpenAI launches GPT-Image-2
OpenAI has launched GPT-Image-2, a new image generation model featuring "thinking" capabilities and enhanced text rendering. The release marks a shift toward using image generation for functional, structured outputs...
7/ 10
Scaling Codex to enterprises worldwide
OpenAI has launched Codex Labs to facilitate the enterprise-scale deployment of Codex throughout the software development lifecycle (SDLC). Through strategic partnerships with consulting firms including Accenture, PwC,...
6/ 10
The AI engineering stack we built internally — on the platform we ship
Cloudflare has deployed an internal AI engineering stack built entirely on its own production infrastructure, including AI Gateway, Workers AI, and Cloudflare Access. This architecture supports over 3,600 internal users...
7/ 10
Building the agentic cloud: everything we launched during Agents Week 2026
Cloudflare has launched a suite of infrastructure primitives designed to support the "agentic cloud," focusing on the unique compute, security, and memory requirements of autonomous AI agents. The updates provide...
7/ 10
Changes to GitHub Copilot Individual plans
GitHub is implementing changes to GitHub Copilot Individual plans, including pausing new sign-ups for Pro, Pro+, and Student plans and tightening usage limits. These adjustments are a response to the increased compute...
7/ 10
Unweight: how we compressed an LLM 22% without sacrificing quality
Unweight is a lossless compression system designed to reduce LLM weight size and alleviate memory bandwidth bottlenecks during inference on NVIDIA H100 GPUs. By compressing the exponent bytes of BF16 weights, the system...
7/ 10
Agents that remember: introducing Agent Memory
Cloudflare has announced the private beta of Agent Memory, a managed service designed to provide persistent, retrieval-based memory for AI agents. The service addresses the "context rot" problem by extracting and...
7/ 10
[AINews] Anthropic Claude Opus 4.7 - literally one step better than 4.6 in every dimension
Anthropic has released Claude Opus 4.7, an update to the Opus model family focused on improving coding, instruction following, and computer-use capabilities. The release introduces a new tokenizer and expanded vision...
7/ 10
Cloudflare’s AI Platform: an inference layer designed for agents
Cloudflare is expanding its AI platform into a unified inference layer, providing a single API to access models from various third-party providers. This update enables developers to manage multi-model workflows,...
6/ 10
Building the foundation for running extra-large language models
Cloudflare has implemented several architectural optimizations to the Workers AI infrastructure to support the deployment of extra-large language models, such as Kimi K2.5. These updates focus on decoupling...
7/ 10
Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers
This entry details the process of finetuning multimodal embedding and reranker models using the Sentence Transformers library, specifically for tasks like Visual Document Retrieval (VDR). It demonstrates how...
6/ 10
Codex for (almost) everything
The Codex application for macOS and Windows has been updated with new features designed to extend its operational capabilities. The update introduces tools for system interaction, web access, and persistent context...
7/ 10
The next evolution of the Agents SDK
OpenAI has updated the Agents SDK to include native sandbox execution and a model-native harness. These updates are designed to enable the development of secure, long-running agents capable of interacting with multiple...
6/ 10
Inside VAKRA: Reasoning, Tool Use, and Failure Modes of Agents
VAKRA is an executable benchmark designed to evaluate the reasoning and tool-use capabilities of AI agents within enterprise-like environments. It moves beyond testing isolated skills by measuring compositional...
8/ 10
Patch Tuesday, April 2026 Edition
Microsoft's April 2026 update addresses 167 vulnerabilities, including active zero-day exploits in SharePoint Server and Windows Defender. Concurrent emergency patches for Google Chrome and Adobe Reader address critical...
7/ 10
Enterprises power agentic workflows in Cloudflare Agent Cloud with OpenAI
Cloudflare has integrated OpenAI's GPT-5.4 and Codex models into its Agent Cloud platform. This integration allows enterprises to develop, deploy, and scale autonomous AI agents designed for executing complex,...
8/ 10
Our response to the Axios developer tool compromise
OpenAI has addressed a supply chain attack originating from the Axios developer tool. The response involved rotating macOS code signing certificates and deploying application updates to mitigate the impact of the...
6/ 10
Safetensors is Joining the PyTorch Foundation
Safetensors, originally a Hugging Face project, is transitioning to the PyTorch Foundation under the Linux Foundation to establish vendor-neutral, community-driven governance. This move shifts the project's trademark...
8/ 10
Russia Hacked Routers to Steal Microsoft Office Tokens
The threat actor Forest Blizzard (also known as APT28 or Fancy Bear) is exploiting vulnerabilities in legacy and unpatched SOHO routers to perform large-scale DNS hijacking. This technique allows attackers to intercept...
7/ 10
Components of A Coding Agent
A coding agent is an application layer, or "agentic harness," that wraps a Large Language Model (LLM) in a control loop to perform software engineering tasks. While the LLM provides the core reasoning, the harness...
8/ 10
Welcome Gemma 4: Frontier multimodal intelligence on device
Gemma 4 is a new family of open-weights multimodal models released under the Apache 2 license, designed for both on-device and large-scale deployment. The series supports text, image, audio, and video inputs, featuring...
6/ 10
How can you test your code when you don’t know what’s in it?
Testing Model Context Protocol (MCP) servers is uniquely challenging due to the non-deterministic nature of Large Language Models (LLMs) and agentic workflows. Because LLMs determine tool invocation sequences...
7/ 10
Encoding Team Standards
To prevent inconsistent code quality caused by varying developer prompting styles, teams should transition from manual checklists to versioned, executable AI instructions. By treating AI instructions as shared...
6/ 10
TRL v1.0: Post-Training Library Built to Move with the Field
TRL v1.0 introduces a formal stability contract to the library, transitioning it from a research project to a stable infrastructure component. The update implements a bifurcated API structure that separates stable,...
6/ 10
Granite 4.0 3B Vision: Compact Multimodal Intelligence for Enterprise Documents
Granite 4.0 3B Vision is a compact multimodal model designed for high-precision extraction of structured data from enterprise documents. Released as a LoRA adapter for the Granite 4.0 Micro language model, it...