CORE AI
Live
BREAKING
SoftBank secures $40B bridge loan to double down on OpenAI OpenAI kills Sora — all compute redirected to robotics research GPT-5.4 scores 75% on OSWorld-V, beats human baseline of 72.4% Gemini 3.1 Pro hits 77.1% on ARC-AGI-2 with 1M token context window Apple's new Siri runs on Google's 1.2 trillion parameter Gemini model Google Lyria 3 Pro — most advanced AI music model — rolling out now Shopify merchants now sell inside ChatGPT via Agentic Storefronts OpenAI Safety Bug Bounty launched for AI abuse and misuse scenarios SoftBank secures $40B bridge loan to double down on OpenAI OpenAI kills Sora — all compute redirected to robotics research GPT-5.4 scores 75% on OSWorld-V, beats human baseline of 72.4% Gemini 3.1 Pro hits 77.1% on ARC-AGI-2 with 1M token context window Apple's new Siri runs on Google's 1.2 trillion parameter Gemini model Google Lyria 3 Pro — most advanced AI music model — rolling out now Shopify merchants now sell inside ChatGPT via Agentic Storefronts OpenAI Safety Bug Bounty launched for AI abuse and misuse scenarios
Saturday, 28 March 2026 // Edition 0328
The Intelligence
Doesn't Stop
Every breakthrough, benchmark, and model drop — curated for the minds who can't afford to miss what's next in AI.
0
Stories Today
0
Model Moves
0
$ Capital Deployed
Top Stories — 28 March 2026
Showing all stories
Featured
🔥 BreakingIndustry
SoftBank Lines Up $40 Billion to Double Down on OpenAI — The AI Arms Race Just Got Financial
SoftBank has secured a $40B bridge loan — with JPMorgan, Goldman Sachs, Mizuho, SMBC, and MUFG — to fund further investment in OpenAI running to March 2027. The signal is strategic: the frontier AI race is no longer just about model quality. Balance-sheet scale is now a moat. For startups, the bar just got sharply higher.
Fortune / TechStartups · Mar 27
5 min read
01
NewModels
GPT-5.4 Scores 75% on OSWorld-V — Above the Human Baseline of 72.4%
OpenAI's new flagship combines reasoning, coding, and agentic workflows with a 1M-token context window. On OSWorld-V — which simulates real desktop productivity tasks — it edged past human performance. GPT-5.1 models are now retired. The shift from AI as a chat tool to AI as an autonomous coworker is no longer theoretical.
OpenAI / NeuralBuddies
Mar 20
02
HotModels
Gemini 3.1 Pro Hits 77.1% on ARC-AGI-2 With a 1 Million Token Context Window
Google DeepMind's most advanced Pro-tier model handles multimodal reasoning across text, images, audio, video, and code in a unified stream. Available via Gemini API, Vertex AI, and Google Antigravity — it's the new benchmark to beat for frontier multimodal performance as of February 2026.
Google DeepMind / LLM Stats
Feb 2026
03
Industry
Apple's Rebuilt Siri Is Secretly Powered by Google's 1.2 Trillion Parameter Gemini Model
Apple confirmed its entirely rebuilt Siri — launching with iOS 26.4 — uses Google Gemini running on Apple's Private Cloud Compute. The new Siri has on-screen awareness and cross-app integration. It's one of tech's most unexpected alliances ever, and it redraws the competitive AI map entirely.
Crescendo AI / TechStartups
Mar 2026
04
NewAgents
Shopify Merchants Now Sell Inside ChatGPT and Gemini via Agentic Storefronts
Shopify's Agentic Storefronts let merchants sell across ChatGPT, Google AI Mode, Microsoft Copilot, and Gemini — with pricing, checkout, and inventory synced from Shopify admin at no extra cost. Even non-Shopify brands can join via a new Agentic plan. Commerce just slipped into the chat window permanently.
The AI Marketers · Mar 26
Mar 26
05
Research
Donald Knuth Published "Claude's Cycles" After Anthropic's AI Solved His Open Graph Theory Problem
In early March 2026, legendary computer scientist Donald Knuth published a paper opening with "Shock! Shock!" — after Claude Opus 4.6 solved a Hamiltonian cycle problem he'd been working on for weeks. Knuth, the father of algorithm analysis, called it a "dramatic advance in automatic deduction and creative problem solving."
Radical Data Science / Anthropic
Mar 10
06
WatchSafety
OpenAI Launches Safety Bug Bounty — AI Abuse and Misuse Scenarios Now In Scope
OpenAI's new Safety Bug Bounty expands its security program to cover AI-specific abuse vectors and misuse scenarios. Researchers can now report safety issues directly. The move comes as regulators in the EU and US tighten scrutiny on frontier labs — and signals safety becoming a genuine competitive surface.
OpenAI Newsroom · Mar 25
Mar 25
07
Policy
Google Warns Quantum Computers Could Break Today's Encryption by 2029
Google is urging governments and financial institutions to accelerate post-quantum cryptography migration, warning that "store now, decrypt later" attacks are already a real threat. Post-quantum readiness is moving from theoretical hygiene to urgent infrastructure planning for anyone holding long-lived sensitive data.
The Guardian / TechStartups · Mar 27
Mar 27
08
HotIndustry
OpenAI Kills Sora — Shuts Down Its AI Video App and Redirects Compute to Robotics
OpenAI confirmed the shutdown of Sora, its text-to-video platform that launched February 2024, went viral, and reached the top of Apple's App Store after Sora 2. The compute is now being redirected entirely toward robotics research — a clear declaration about where OpenAI sees the next major frontier.
OpenTools AI / The AI Marketers · Mar 26
Mar 26
09
NewModels
Google Launches Lyria 3 Pro — Its Most Advanced AI Music Generation Model
DeepMind's Lyria 3 Pro is rolling out to paid Gemini subscribers, developers via Google AI Studio, and globally to Producer AI users. It marks a major leap in music generation fidelity and control. AI music tools are rapidly moving from party tricks to professional creative infrastructure for studios and artists.
Google / Crypto Integrated · Mar 25
Mar 25

No stories match your search.

About Core AI
Core AI is an independent intelligence feed for people who take artificial intelligence seriously. We track every model release, benchmark record, funding round, and policy shift — and translate them into clear, sharp editorial that respects your time.
We publish every day. We cite primary sources. We don't run press releases — we run perspective.
Get in touch
hello@coreai.news
Signal over noise
Every story must clear one bar: does this change how you understand AI? If not, it doesn't make the feed.
Primary sources always
We link to original papers, lab blogs, and official statements — not aggregators or rewrites of rewrites.
No hype, no panic
AI coverage tends toward breathless extremes. We aim for the calm, clear middle — accurate and context-rich.