OpenAI’s AI Chemist Finds a Lab-Tested Way to Improve Drug Discovery Chemistry
OpenAI and Molecule.one connected GPT-5.4 to an autonomous chemistry platform that ran 10,080 reactions and found a TEMPO-based way to improve a difficult Chan-Lam coupling used in medicinal chemistry. The result is narrow, but it shows AI starting to work inside the experimental loop, not just around it.
Matter 1.6 Tries to Fix the Smart Home’s Setup Problem
Matter 1.6 adds NFC-based setup, Joint Fabric for shared multi-platform control, smarter thermostat suggestions, and better device status reporting. The update could make smart homes less fragmented, but only after device makers and platform owners actually ship support.
Mastra npm Compromise Turns AI Agent Frameworks Into a Supply-Chain Target
Attackers republished more than 140 Mastra npm packages with a poisoned easy-day-js dependency, exposing AI agent developers to an install-time remote payload. Teams that installed affected @mastra packages on June 17 should treat developer machines and CI runners as compromised.
G7 AI Summit Turns Model Access Into a Sovereignty Fight
AI leaders are gathering at the G7 in France as Europe, Canada, and other allies question how much critical AI infrastructure should depend on U.S. model labs, cloud providers, chips, and export-control decisions.
HPE Turns Juniper Into the Network Layer for AI Factories
HPE used Discover 2026 to fold Juniper deeper into its AI data center strategy, adding QFX switches for inference and AMD Helios, Mist and Marvis AIOps updates, and SASE controls for self-driving networks.
MLPerf Training 6.0 Makes AI Infrastructure a Sparse-Model Race
MLCommons released MLPerf Training 6.0 with new DeepSeek V3 and GPT-OSS 20B benchmarks, while NVIDIA, AMD, and cloud providers used the round to show how AI training is moving toward sparse models, rack-scale systems, and verified cloud capacity.
Android 17 Starts Rolling Out With Bubbles, Tighter Permissions and Delayed Gemini Tools
Google has started rolling out Android 17 to Pixel devices, with floating app Bubbles, Screen Reactions, stronger permission controls, anti-theft protections, and gaming updates. The more ambitious Gemini Intelligence features are still due later this summer on select advanced devices.
Databricks Turns the Lakehouse Into an Operating Layer for AI Agents
Databricks used Data + AI Summit 2026 to launch Lakehouse//RT, Genie One, CustomerLake, Unity AI Gateway updates, and a Panther acquisition. The moves show the company trying to make the lakehouse a governed operating layer for real-time apps, agents, marketing, and security operations.
SpaceX’s $60B Cursor Deal Turns AI Coding Tools Into Infrastructure
SpaceX’s SEC filing confirms a $60 billion all-stock deal to buy Anysphere, the company behind Cursor. The acquisition turns AI coding tools into a strategic infrastructure bet tied to compute, developer distribution, and xAI’s coding-agent ambitions.
LiteSpeed cPanel Flaw Puts Shared Hosting Servers on CISA’s Patch Clock
CISA added CVE-2026-54420, an actively exploited LiteSpeed cPanel plugin flaw, to its Known Exploited Vulnerabilities catalog. Shared hosting providers running CloudLinux or CageFS should move to the fixed plugin versions, check cPanel logs, and treat suspicious access as a possible root-level incident.