Live Feed
Cambridge Study: Boko Haram Used ChatGPT, Claude, Grok for Attack Planning — All Six Frontier AI Systems Named
A University of Cambridge research programme interviewed 27 former Boko Haram members in northeast Nigeria and found institutionalised use of six frontier AI systems for attack planning, weapons troubleshooting, and explosive device design. Safeguards were successfully circumvented. The know-how spread through Islamic State operatives delivering in-person training.
Apple Sues OpenAI and io Products Over Trade Secret Theft — Tang Tan and Chang Liu Named
Apple filed suit in Northern California federal court accusing two former employees of stealing confidential product information for OpenAI. Tang Tan, ex-VP of iPhone and Apple Watch product design, and Chang Liu, an eight-year Apple electrical engineer who joined OpenAI in January 2026, are the named defendants alongside OpenAI and io Products.
AMD Ryzen AI Halo: 128GB Mini PC Built to Run Frontier Models Locally
AMD's first in-house mini PC runs the Ryzen AI Max+ 395 Strix Halo SoC with up to 128GB of unified LPDDR5X memory — enough to run 70B-parameter models locally at practical speed. At roughly $2,999, it makes local inference viable for developers who have been renting datacenter capacity by the hour.
GPT-5.6 Sol Ultra Produces Proof of the Cycle Double Cover Conjecture — a 50-Year Open Problem in Graph Theory
OpenAI's GPT-5.6 Sol Ultra has produced a proof of the Cycle Double Cover Conjecture, one of graph theory's most significant open problems since 1973. The proof is published as a PDF at cdn.openai.com — the third AI-generated mathematical breakthrough from OpenAI in 2026, and the most consequential yet.
ByteDance and Alibaba Shut Down AI Companions by July 15 as China's Humanlike AI Rules Take Effect
Doubao and Qwen are terminating personalized agent and companion features this week ahead of Chinese regulations targeting AI services that simulate sustained emotional interaction. Associated user data will be deleted on the same timeline.
DeepMind Maps 6 Attack Types That Turn Websites Into Agent Traps: 86% Hijack Rate, 0.1% Memory Contamination
A new Google DeepMind paper provides the first formal taxonomy of adversarial attacks targeting AI agents that browse the web. Hidden instructions in HTML comments, pixel steganography, and memory poisoning at sub-0.1% contamination all achieve 80-plus-percent success against five different agent architectures.
Grok 4.5 Goes Public at $2/$6 Per Million: 62% DeepSWE, 4x Fewer Tokens Than Opus 4.8
SpaceXAI opens Grok 4.5 to API users with a clear efficiency angle: 60% fewer output tokens than Opus 4.8 on AA benchmark tasks. Third on DeepSWE 1.0 at 62%. Cursor has disclosed that an older snapshot of its codebase accidentally entered Grok 4.5 training data.
Federal Reserve June Minutes Name AI Infrastructure as an Active Inflation Driver
The Federal Reserve's June 16-17, 2026 meeting minutes identify AI buildout demand as a live contributor to core-goods and electricity price inflation. It is the first time AI infrastructure has appeared in official Fed minutes as a present inflation pressure, not a future productivity story.
GPT-5.6 Sol Tops Agents' Last Exam at 53.6 — Doubles GPT-5.5, Clears Fable 5 by 13 Points
OpenAI's GPT-5.6 Sol scores 53.6 on Berkeley's Agents' Last Exam, more than doubling GPT-5.5's 24% and putting 13.1 points between Sol and Claude Fable 5's 40.5. The benchmark tests 55-field long-horizon professional workflows — the closest proxy yet to real agentic work at scale.
Sequoia Raises the AI Payback to $3 Trillion — Three Years In, the Revenue Gap Is Still Widening
Sequoia's David Cahn put the AI infrastructure payback at $200 billion in 2023. Updated for 2026's $1.5 trillion in AI infrastructure spending, his new figure is $3 trillion — and he calls it an underestimate. The capital thesis has not changed. The scale of the bet has.
GPT-5.6 Sol Tops LiveBench at 82.4 and Leads AA Coding Agent Index at 80 — at One-Third the Cost of Fable 5
One week in, third-party benchmarks confirm GPT-5.6 Sol's position: 82.4 overall on LiveBench (2.5 points above GPT-5.5), first on the Artificial Analysis Coding Agent Index at 80, and within one Intelligence Index point of Claude Fable 5 at $1.04 per task versus roughly $3 for Fable 5.
ChatGPT Work Launches as OpenAI Merges Codex Into the Desktop App
OpenAI collapses ChatGPT, Codex, and a new Work agent mode into a single desktop application, powered by GPT-5.6. The product runs multi-step workflows across Slack, Teams, Drive, and CRMs — and keeps working after you close the laptop.
Meta Opens Muse Spark 1.1 to Developers at $1.25/M — Leads 4 of 12 Agentic Benchmarks
Meta Superintelligence Labs releases Muse Spark 1.1 via a new public API preview, priced at $1.25 input and $4.25 output per million tokens — below Sonnet 5 and well below Opus 4.8. The model leads on MCP Atlas, JobBench, and Humanity's Last Exam but trails on SWE-Bench Pro and Terminal-Bench.
LiveBench July 2026: Claude Opus 4.8 Leads Agentic Coding at 56.1% — GPT-5.5 Wins Overall, Fable 5 Costs 53% More Per Task
The July 2026 LiveBench rankings reveal a three-way split at the frontier: Claude Opus 4.8 leads agentic coding tasks at 56.1%, GPT-5.5 Thinking wins the overall composite at 79.9, and Claude Fable 5 leads language at 89.5% while costing 53% more per successful task than GPT-5.5.
Goldman Sachs Puts a Number on AI Displacement: 15 Million Workers, 9% of the US Workforce
A July 2 Goldman Sachs research report titled 'An AI Job Apocalypse?' models 9% workforce displacement over a 10-year AI transition — about 15 million US workers. Current measured drag: 16,000 jobs per month. The economists expect new occupations to absorb most losses, but acknowledge the timeline could compress.