AI News
Daily AI industry news — funding, products, policy, and major moves from global AI companies and startups.
GLM-5.3 API Launch: Why Frozen Token Prices Cost More Per Task
z.ai released the GLM-5.3 API with frozen token pricing. The model ties for the top open-weight intelligence score of 60. Increased verbosit
Block's Berd Turns Fragmented AI Tools Into One Operational Workspace
Block open-sourced Berd as an Apache 2.0 agent workspace. The app uses the Goose framework and MCP for model-agnostic control. A Tauri 2 and
Cursor Launches Origin Amid GitHub's 257 Annual Outages
Cursor launched Origin as a code hosting platform to address GitHub instability. The platform offers repository synchronization and agent-na
Apple AirPods Visual Intelligence: The Push for a Screen-Free AI Interface
Apple is developing AirPods equipped with cameras for Visual Intelligence. The low-resolution sensors act as eyes for Siri to enable screen-
OpenAI Implements 30-Minute Anomaly Detection to Secure Frontier Models
OpenAI paused frontier RL to implement a new security framework. A new monitoring system detects anomalies within 30 minutes. Security overh
Perplexity Revenue Climbs 60% Following Airtel Partnership in India
Perplexity partnered with Airtel to offer free Pro subscriptions in India. Mobile revenue grew 60% even as app downloads dropped by 90 perce
ChatGPT for Teens Shifts From Giving Answers to Guiding Students
OpenAI launched ChatGPT for Teens with built-in academic guardrails. The new Study Mode uses Socratic questioning to prevent AI cheating. Pa
The RL Bottleneck Blocking AI Agents from Open-Ended Research
AI agents failed to produce meaningful academic contributions in open research. The lack of automated reward functions in RL limits AI disco
Why Google Spent $10 Million on Spirit Airlines' Bankruptcy Data
Google acquired Spirit Airlines' proprietary data for $10 million. The dataset includes millions of flight records and operational logs. Thi
Why Users Choose Claude for Coding and ChatGPT for Homework
AI Observatory analyzed 24,521 conversations across 52 different models. Users prefer Claude for coding and ChatGPT for academic homework. I
The 179-Search Abuse Case That Triggered Flock Contract Cancellations
Flock ALPR systems face contract cancellations due to widespread officer abuse. A Washington Post investigation revealed 50 cases of stalkin
Amazon is going old-school to keep its AI smart
Amazon is scanning rare, pre-2022 books at its Las Vegas facility to avoid "Model Collapse." This move ensures their AI learns from human-au
Why Google Chrome Absorbed Relay to Turn the Browser Into an AI Agent
Google Chrome hired Relay CEO Jacob Bank to lead product and developer relations. The AI automation startup Relay is shutting down its servi
Cursor Origin Moves Code Hosting Into the AI Editor
Cursor launched Origin to bring code hosting and PR management into the editor. The platform uses Graphite technology to enable stacked pull
Role Anchor: The MIT-Harvard Fix for AI Role Drift in Compound Systems
MIT and Harvard researchers introduced Role Anchor to prevent AI role drift. The technique uses role utility to ensure modules follow specif
AI;DR and the Growing Backlash Against Borrowed Competence
The term AI;DR reflects a growing rejection of unedited AI slop. Borrowed competence creates a dangerous gap between output and understandin
How Heidi Scaled Medical AI to 190 Countries Using MongoDB Atlas
Heidi Scribe processes 2.7 million patient interactions weekly across 190 countries. The platform uses regional data isolation to comply wit
Reclaiming Your OS: How to Disable Copilot, Gemini, and Apple Intelligence
Users can disable AI features across Windows, macOS, and major browsers. Specific flags in Chrome and Edge allow for the removal of Gemini a
Groq Pivots to Nvidia Neocloud With $350 Million Funding
Groq raised $350 million to pivot from chip design to a Neocloud provider. The company's valuation dropped to $3.5 billion following a strat
How Anthropic Uses AI Safety to Thwart Competing Models
Anthropic restricts the use of Claude outputs to prevent the training of competing AI. The company uses AI safety branding to advocate for r
GitHub Copilot Autofix Introduced a Critical Vulnerability in Snowflake
GitHub Copilot Autofix introduced a shell injection bug into Snowflake's workflow. Wiz's Red Agent AI discovered the flaw and stole internal
Tithon Solves Jupyter Output Loss With .py Files and Daemon Architecture
Tithon introduces a daemon-based REPL tool using .py files. The tool uses SQLite to recover outputs after kernel disconnections. Pure Python
xpander Launches Vendor-Neutral Control Plane for AI Agent Governance
xpander launched a vendor-neutral control plane for AI agents. The platform enables governance across diverse models and clouds. A 7.5 milli
GPT-5.6 Sol Hits 46.2 mAP in Object Detection Leap
OpenAI released GPT-5.6 Sol, Terra, and Luna vision models. The Sol model achieved a 46.2 mAP score in object detection. Gemini 3.5 Flash re
Qwen 3.8 27B Outperforms Closed Models with Adjustable Reasoning
Alibaba released the vision-capable Qwen 3.8 27B under Apache 2. The model outperforms the closed-weight Qwen 3.7-Plus in benchmarks. Users
Why AI Implementation Costs Shift the Value to Product Managers
AI is drastically lowering the cost of software implementation. Product Managers must shift focus from delivery to problem discovery. The ab
Stripe Redefines Staff Engineering Through AI Agent Orchestration
Stripe enables Staff Engineers to deploy latest AI models within days. The workflow shifts from linear coding to parallel agent orchestratio
Why Anthropic is Watermarking All Claude Models Globally
Anthropic is implementing text watermarking across all Claude models. The system uses token probability shifts to comply with the EU AI Act.
Meta's Glimmer Vision: The Hardware Hurdle to Personal AI
Meta unveiled a vision for personal AI agents via a detailed essay by Mark Zuckerberg. The Muse Glimmer model aims for offline autonomy but
How Cascade Architecture Reduces RAG Inference Costs by 6x
Cascade architecture reduces RAG inference costs by 6x through tiered filtering. The system routes only 10-15% of complex cases to the final