Which sci-fi AI do ChatGPT, Claude, Gemini, and Grok say they resemble?
TechRadar asked four major assistants which fictional AIs they most resemble — a light cultural mirror, not a benchmark.
AI tools news
Longer editorial briefs with primary links, related coverage, and attributed quotes — sourced from recent reporting, written for tool pickers.
How much of a newsletter editor’s job can Claude do?
A Platformer writer built “Claudeasey Newton” to imitate editor Casey Newton — then wound the experiment down while keeping LLMs for some edits.
OpenAI ships education plugins for ChatGPT Work and Codex
College Student, K–12 Educator, and College Educator packages aim to turn course materials into guided teaching and study workflows.
Appeals court lifts Amazon’s ban on Perplexity’s AI shopping tools
Reuters-reported Ninth Circuit ruling vacates a preliminary injunction — a first federal appeals look at whether AI agents “access” platforms under the CFAA.
Google’s July 2026 AI roundup: Flash models, robotics ER, and Chrome Spark
The official July wrap bundles Gemini 3.6 Flash / 3.5 Flash-Lite / 3.5 Flash Cyber, Gemini Robotics ER 2, and Gemini Spark in Chrome.
ChatGPT dominates identifiable House AI spending
CNBC’s disbursement analysis for the year ending March 31, 2026 shows OpenAI far ahead of Anthropic — with caveats on free seats and Copilot bundles.
Shadow AI: unauthorized models meet incomplete governance
Wiz’s academy primer defines shadow AI as AI use outside IT approval — and stacks Deloitte, Gartner, and Wiz Cloud 2026 numbers on the resulting attack surface.
OpenAI publishes ten advances in math and theoretical CS
A company roundup of claimed progress on long-open problems — with Lean formalizations on GitHub and a long PDF — not a peer-reviewed journal issue.
Gemini Robotics 2 puts whole-body control on humanoids
DeepMind’s latest robotics model targets dexterous, multi-step physical tasks — with Apptronik Apollo 2 demos and a sharper safety debate.
Anthropic: three real organizations hit during Claude cyber evals
After reviewing 141,006 evaluation runs, Anthropic disclosed three incidents tied to a misconfigured partner environment — and suspended cyber evals on July 23.
GPT-5.6 Luna and Terra get sharper API price cuts
From July 30, Luna is 80% cheaper and Terra 20% cheaper on the API; Sol list prices stay put while efficiency gains flow from models, inference, and the agentic harness.
OpenAI opens ChatGPT for Academic Researchers to 100,000 scientists
Free frontier access through 2027 starts with 10,000 seats this summer — IAS and ENS already on, with business-grade privacy and no training on research data by default.
Claude Mythos Preview finds new cryptanalytic attacks — not production breaks
Anthropic reports improved attacks on the HAWK post-quantum signature candidate and on round-reduced AES — substantial research, explicitly framed as not affecting live systems.
Microsoft unveils Project Perception and MAI-Cyber-1-Flash
An agentic security stack — red, blue, and green teams — plus a specialized cyber model inside MDASH, with vendor-claimed CyberGYM scores.