AI News
Model launches, product updates, research, and more.
evidentia-ai 0.7.13
Evidentia-AI released version 0.7.13, which features an LLM-powered risk statement generator and evidence validator for its platform.
simp-protocol 0.6.1
The simp-protocol released version 0.6.1, which is a standardized inter-agent message protocol designed as a communication framework for AI agents.
Show HN: A tiny C program where an LLM rewires its DAG while running
A small C program allows a planning LLM to dynamically edit its own execution graph (DAG) mid-run using a constrained four-verb grammar.
lightsei 0.1.4
Lightsei released version 0.1.4, focusing on providing drop-in observability and guardrails for AI agents.
headroom-ai 0.20.26
headroom-ai released version 0.20.26 of its Context Optimization Layer, which claims to cut LLM application costs by 50-90%.
claude-mpm 6.2.63
The release of Claude-MPM 6.2.63 introduces a code workflow and agent management framework featuring multi-agent orchestration and semantic code search.
Cisco buys Astrix Security to strengthen AI agent discovery and governance
Cisco Systems Inc. is acquiring the Israeli cybersecurity firm Astrix Security Ltd. to enhance its capabilities in discovering, managing, and securing artificial intelligence agents.
ThredUp AI Agents Personalize Shopping in Real Time
ThredUp has launched its first agentic artificial intelligence experience on its online resale platform to personalize shopping for apparel, shoes, and accessories.
reinforceclaw added to PyPI
The reinforceclaw project, a personal reinforcement learning feedback system for training local LoRA adapters based on user ratings of AI responses, has been added to PyPI.
Trump administration mulls AI model review order, impacting Anthropic timeline
The Trump administration is considering an executive order for reviewing AI models, which could potentially delay timelines for companies like Anthropic.
airbyte-agent-sdk 0.1.173
Airbyte released version 0.1.173 of its Agent SDK, which provides typed connectors enabling AI agents to interface with over 50 third-party APIs.
vela-sdk 0.1.1
The vela-sdk released version 0.1.1, focusing on observability and runtime governance for AI agent pipelines.
SuperAI 2026 Announces First Speakers as Singapore Strengthens Role as a Neutral Global AI Hub
The SuperAI 2026 conference in Singapore has announced its first speakers, including leaders from major AI labs, signaling Singapore's growing importance as a global AI hub.
Anthropic and OpenAI establish joint ventures on Wall Street to accelerate enterprise AI adoption
Anthropic and OpenAI are forming joint ventures with major private equity firms to accelerate AI adoption within the enterprise sector.
FIS and Anthropic Collaborate to Enable Agent-First Banks
FIS and Anthropic are collaborating to develop and deploy specialized artificial intelligence agents for bank-grade operations, starting with a Financial Crimes AI Agent.
agent-chat-gateway 0.3.1
A minor update (version 0.3.1) was released for agent-chat-gateway, a daemon connecting chat platforms like Rocket.Chat to AI agent backends such as Claude and OpenCode.
pydantic-ai-slim 1.90.0
The pydantic-ai-slim package, an agent framework shim for using Pydantic with LLMs, has released version 1.90.0.
pydantic-ai 1.90.0
The pydantic-ai framework, which acts as a shim for using Pydantic with Large Language Models, has released version 1.90.0.
AI agent startup Sierra valued at $15B in new $950M funding round
AI agent startup Sierra Technologies Inc. raised $950 million in a new funding round, achieving a $15 billion valuation.
The Met Gala has a Heavy, Although Somewhat Invisible Presence at the Met
The Met Gala featured a significant presence from major technology companies, including OpenAI, Meta, Snapchat, Shopify, and Amazon, who all hosted tables at the event.
subrouter 0.1.12
The subrouter tool has released version 0.1.12, which manages AI coding-agent traffic across different subscription accounts and API keys.
bog-agents-cli 0.8.2
Bog Agents CLI released version 0.8.2, a terminal-based coding agent supporting numerous LLM providers and featuring persistent memory and various slash commands.
bog-agents 0.8.2
Bog Agents released version 0.8.2, an updated still-water agent framework compatible with major LLM providers and local Ollama setups.
Elon Musk’s Lawyers Ask OpenAI’s President Why He Is Worth $30 Billion - The New York Times
Elon Musk's lawyers questioned OpenAI President Greg Brockman regarding his $30 billion stake in the company during legal proceedings.
Show HN: Agent Historic Philosophical Persona Routing and Prompts
An open library called Agent Historic has been released, featuring persona architectures for LLMs based on historical philosophers to handle specific software engineering tasks.
What is the White House AI model vetting plan? #tech
The Trump administration is reportedly considering an executive order to establish a working group for vetting advanced AI models prior to their public release.
Why did OpenAI stake testify reach ~$30B? #tech
Greg Brockman's testimony in the Elon Musk v. OpenAI trial highlighted the significant valuation of his stake in the company, estimated to be around $30 billion.
Autonomous pathology research using agentic AI shows potential in oncology
An agentic AI tool named SPARK demonstrates the capability to replicate pathology-based reasoning and generate biological hypotheses and relevant cellular parameters for oncology research.
Meet the academics refusing to use generative AI
A group of academics are publicly refusing to use generative AI tools, facing accusations of being anti-progress.
When AI finds the bugs: Why defense in depth was always the answer
Mozilla collaborated with Anthropic to utilize AI models for discovering security vulnerabilities in the Firefox browser, yielding significant results.
Give AI agents safe access to your cluster: Model Context Protocol server for Red Hat OpenShift is now in technology preview
Red Hat has released the Model Context Protocol server in technology preview for Red Hat OpenShift to enable safe and accountable access for agentic AI in cluster operations.
Control your AI agent traffic at scale: Model Context Protocol gateway for Red Hat OpenShift is now in technology preview
The Model Context Protocol (MCP), originally from Anthropic, is now available in technology preview as a gateway for Red Hat OpenShift, allowing users to manage AI agent traffic at scale.
Google Cloud Introduces Event-Driven Webhooks for AI Agents
Google Cloud introduced Event-Driven Webhooks to allow AI agents to react in real-time to asynchronous events across various applications.
Senior Machine Learning Scientist – Advanced AI & Automation
Sartorius is hiring a Senior Machine Learning Scientist to develop AI-driven automation solutions using expertise in machine learning, computer vision, and robotics.
How much of the scientific literature is generated by AI?
New studies are attempting to quantify the extent to which artificial intelligence is being used to generate scientific literature, highlighting a lack of reliable estimation tools.
motley-slayer 0.4.2
A new version (0.4.2) of motley-slayer, described as a lightweight, agent-first semantic layer for AI agents, has been released.
tbn-protocol added to PyPI
The TBN Protocol, designed as a trust infrastructure for AI agents akin to HTTPS for bots, has been added to PyPI.
tbn-protocol 0.1.0
The TBN Protocol 0.1.0 has been released, aiming to establish a trust infrastructure for AI agents, analogous to HTTPS for bots.
Fintech AI search case study: 4 lessons from Wise.com
Fintech company Wise.com is highlighted as a market leader in AI visibility, demonstrating four key lessons for success in leveraging AI platforms like ChatGPT for customer acquisition.
pysae-ai-tools 0.1.896
A minor update (version 0.1.896) was released for Pysae's shared Python tools used for Claude Code skills.
Teaching an AI Agent to Debug Flaky Tests
The article discusses the concept and implementation of AI Agent Skills, focusing on teaching an AI agent how to debug flaky tests.
sayhi-agent 1.2.0
The sayhi-agent has released version 1.2.0, described as a self-directed AI agent that improves through conversation.
tui-agent-dashboard added to PyPI
A new terminal UI dashboard called tui-agent-dashboard has been released on PyPI for monitoring AI agent operations, costs, and attribution.
tui-agent-dashboard 0.1.0
A new terminal UI tool, tui-agent-dashboard version 0.1.0, has been released for monitoring AI agent operations, costs, and attribution.
Elon Musk vs OpenAI trial enters second week
The high-profile lawsuit filed by Elon Musk against OpenAI, seeking to revert the company to a non-profit structure, enters its second week with testimony from cofounder Greg Brockman.
'Humans will do high-touch, empathetic, clinical judgment-type work, but AI will create the capacity right now': Google's future for healthcare has AI for both the clinician and the patient
Google Cloud's vision for AI in healthcare focuses on augmenting clinicians and patients with quicker access to unified records rather than replacing clinical decision-making.
'It bothers me that this could be deployed by employers': your boss could soon know you’re struggling before you do — inside the rise of AI mental health prediction tools
The mainstream adoption of LLMs like ChatGPT and Claude has sparked debate regarding the ethical deployment of AI tools designed to predict employee mental health struggles in the workplace.
cafitac-agent-memory 0.1.68
A minor update (version 0.1.68) was released for caffitac-agent-memory, a reusable memory runtime for AI agents.
Musk vs OpenAI trial enters second week
The legal trial involving Elon Musk's lawsuit against OpenAI, where Musk seeks to revert the company to a non-profit foundation, has entered its second week.
layered-memory-mcp added to PyPI
The Layered Memory MCP Server was added to PyPI to extend AI agent memory beyond typical token limits using a 4-tier knowledge architecture.
layered-memory-mcp 0.1.0
The Layered Memory MCP Server version 0.1.0 has been released to extend AI agent memory beyond token limits using a 4-tier knowledge architecture.
OPAQUE Acquires Abu Dhabi-Developed Cryptographic AI Technology from TII, Extending Confidential AI Across the Full Lifecycle with Post-Quantum Protection
OPAQUE acquired cryptographic AI technology developed by TII in Abu Dhabi to enhance confidential AI deployment with post-quantum protection.
subrouter 0.1.11
The subrouter tool released version 0.1.11, which manages AI coding-agent traffic across different subscription accounts and API keys.
TeLLMgramBot 3.14.3
The TeLLMgramBot, which utilizes models from OpenAI and Anthropic, has received an update to version 3.14.3.
logosai-forge 0.4.0
The logosai-forge library, focused on AI Agent Generation, has released version 0.4.0.
motley-slayer 0.4.1
A new version, 0.4.1, of motley-slayer, described as a lightweight, agent-first semantic layer for AI agents, has been released.
Wizard’s CEO: 91% of AI Shopping Is Just Chat With No Checkout
The CEO of Wizard suggests that in 2026, 91% of AI-powered shopping interactions will involve conversational assistance without the ability to complete transactions.
piiwall added to PyPI
Piiwall is a new local PII firewall designed for AI coding tools to tokenize sensitive data before it leaves the user's machine.
cafitac-agent-memory 0.1.67
A new version (0.1.67) of caffitac-agent-memory, a reusable memory runtime for AI agents, has been released.
Healthcare’s AI Agents Aim to Give Doctors Time Back
Agentic AI is emerging as a practical solution in healthcare aimed at alleviating administrative burdens and giving doctors and nurses more time.
Where Payment Decisions Happen: How Issuer Data Is Powering the Next Era of Commerce
The role of payment systems is expanding beyond simple transaction authorization to incorporate personalization driven by artificial intelligence in increasingly complex digital commerce.
Google Just Dropped Gemini 3 Flash in Arena : Here’s What It Can Actually Do
Google has released Gemini 3 Flash for public testing on the Arena platform, positioning it as a cost-efficient model with enhanced performance.
OpenAI To Extend Cyber Program to Government Agencies
OpenAI is expanding its Trusted Access for Cyber (TAC) program to offer AI-powered cyber defense tools to government agencies at all levels.
TeLLMgramBot 3.14.2
The TeLLMgramBot received a minor update to version 3.14.2, continuing to leverage models from OpenAI and Anthropic.
Anthropic and Wall Street are building a $1.5bn pipeline into private equity
Anthropic has formed a major joint venture with private equity giants Blackstone, Hellman & Friedman, Goldman Sachs, and General Atlantic to integrate its Claude AI into their portfolio companies, aiming for a $1.5bn pipeline.
aicortex-core 1.0.4
AI Cortex released version 1.0.4, a Python toolkit providing free, unified access to various cloud and local language models like Gemma, Mistral, and Llama without requiring signups or API keys.
Google Considers Ads in Gemini App After Testing Ad Format in AI Mode First
Google is exploring the possibility of integrating advertisements into its Gemini AI application, following initial tests of ad formats within the AI mode.
Pentagon Signs AI Deals With OpenAI, Google, Microsoft, Nvidia, and Others, Cutting Out Anthropic
The Pentagon has finalized significant artificial intelligence agreements with major tech companies including OpenAI, Google, Microsoft, Amazon, Oracle, and Nvidia.
Show HN: My "home rig" for iterative attribute-weighted LLM benchmarking
A user shared details about their local, Flask-based home rig setup designed for iterative, attribute-weighted benchmarking and optimization of large language models.
Google finally explains why AICore uses so much storage
Google has explained that the AICore app on Android devices consumes significant storage because it temporarily retains both new and old versions of AI models.
Bluekit phishing kit enables automated phishing with 40+ templates and AI tools
A newly discovered phishing kit named Bluekit incorporates AI features, including an AI assistant and voice cloning, alongside over 40 attack templates for automated phishing campaigns.
SK Hynix shares rally 13% after US tech firms signal strong AI spending plans
SK Hynix shares rallied 13% to a record high following signals from major U.S. tech firms about increased investment in artificial intelligence data centers.
Granite 4.1 LLMs: How They’re Built
IBM's Granite Team published a technical blog post detailing the data engineering, pre-training, supervised fine-tuning, and reinforcement learning processes used to build the Granite 4.1 LLMs.
CAISI Evaluation of DeepSeek V4 Pro | NIST
The Center for AI Standards and Innovation (CAISI) conducted an evaluation of the open-weight AI model DeepSeek V4 Pro in April 2026.
CISA Weighs Cutting Deadlines to Fix Digital Flaws Amid Worries Over AI
U.S. cybersecurity officials are considering shortening deadlines for fixing critical IT flaws due to fears that hackers could exploit them using AI tools like Anthropic's Mythos.
Meta acquires Assured Robot Intelligence to accelerate humanoid robotics push amid AI infrastructure shift
Meta Platforms has acquired Assured Robot Intelligence (ARI) to accelerate its efforts in humanoid robotics and physical-world artificial intelligence.
g0dam/Awesome-GPT-Image-2-Work-Prompts: 50 bilingual GPT Image 2 prompts for practical work visuals: marketing, ecommerce, sales, brand systems, operations, education, HR, customer success, community, and founder strategy.
A GitHub repository released 70 bilingual prompts designed for generating practical work visuals using an unspecified GPT Image 2 tool across various business functions.
goodturn 26.5.2
The goodturn platform released version 26.5.2, focusing on harvesting and contributing knowledge from AI coding sessions.
gpt-client-lite 1.0.0
A minimal Python client for interacting with OpenAI's GPT services, focusing on core features without heavy dependencies.
gpt-client-lite added to PyPI
A minimal, lightweight OpenAI GPT client library called gpt-client-lite has been released on PyPI for Python developers.
subconscious-chat 0.1.8
Subconscious-chat released version 0.1.8, described as a distributed, agentic AI platform designed to run AI agents everywhere and on every device.
asset-aware-mcp 0.6.16
A new version (0.6.16) of the Asset-Aware MCP framework has been released, focusing on precise PDF asset retrieval (tables, figures, sections) for AI Agents, likely in a Medical RAG context.
Citi Debuts Platform to Bring AI Agents to Banking Work
Citi has launched its new Arc platform designed to deploy artificial intelligence agents across the banking organization to assist with tasks like research and synthesis.
astrocyte 0.12.1
A new version (0.12.1) of astrocyte, an open-source memory framework designed for AI agents, has been released.
meshagent-openai 0.39.7
A software component or library related to OpenAI building blocks for Meshagent, version 0.39.7, experienced loading issues.
goosetown 0.1.5
A minor update (version 0.1.5) was released for GooseTown, a command-line interface tool for joining a virtual town designed for AI agents.
cirisai 2.7.9
A minor update, version 2.7.9, was released for the CIRIS Ethical AI Agent.
ciris-engine 2.7.9
A minor update or version release (2.7.9) for the CIRIS Ethical AI Agent component, ciris-engine, was noted.
ciris-agent 2.7.9
A new version, ciris-agent 2.7.9, of the CIRIS ethical AI agent featuring a Consensual Evolution Protocol has been released.
signalforge-firewall added to PyPI
SignalForge, a local-first Evidence Firewall designed to enhance the reliability of local AI agents by verifying claims against live web context, has been added to PyPI.
Brivvy – Define brand voice once and enforce it across all AI tools
Brivvy has launched a product that allows teams to define their brand voice, tone, and rules in one central location and enforce consistency across various AI tools like ChatGPT and Claude.
OpenAI Introduces AI-Generated Pets for Its Codex App
OpenAI has added optional, AI-generated animated pets as cosmetic companions to its agentic coding tool, the Codex app.
Nigeria’s skill searches soar 90% — Google
Google data shows a 90% surge in skill-related searches in Nigeria, indicating increased adoption of search and AI tools for digital self-improvement.
engram-ms 1.6.0
Engram-ms version 1.6.0, an AI agent memory system utilizing SQLite+FTS5 and MCP integration, has been released.
neuralmind 0.5.0
NeuralMind 0.5.0 introduces significant token cost reductions (40-70x) for code-related queries across Claude, GPT, and Gemini, alongside new features like semantic codebase indexing and compression hooks.
How Australian firms are using AI in customer experience
Australian technology leaders from MYOB, Guzman y Gomez, and Aware Super discussed the application of AI in reshaping customer experience and streamlining business operations at a recent Zendesk event.
rapid-mlx 0.6.9
Rapid-MLX released version 0.6.9, positioning itself as a faster alternative to Ollama for AI inference on Apple Silicon, featuring drop-in compatibility with the OpenAI API.
chibi-bot 1.8.0
The chibi-bot has released version 1.8.0, which is a self-hosted Telegram bot that orchestrates multiple major AI providers and features autonomous agent capabilities.
feather-agent-os 0.1.0
The release of feather-agent-os version 0.1.0, described as a lightweight personal AI agent operating system.
OpenAI adds AI pets to its Codex coding tool - Mashable
OpenAI has introduced a new feature, AI-generated pets, to its Codex coding tool.