All articles
OpenAI replaces Advanced Voice Mode — GPT-Live brings full-duplex to ChatGPT
OpenAI launched GPT-Live-1 and GPT-Live-1 mini on July 8 — a new generation of voice models built on full-duplex architecture. The models replace Advanced Voice Mode in ChatGPT and are the first to listen and speak at the same time.
DeepSeek moves into silicon — custom AI inference chips in development
DeepSeek has been working for roughly a year on entering the AI inference chip business, Reuters reports. The Chinese startup wants to reduce its dependence on Nvidia, blocked by US export controls, and Huawei, which controls about half of China's data center chip market.
MoonBit: a programming language designed for the AI era, built in China
MoonBit — a programming language developed in China — outperforms Gleam in an IEEE-accepted study despite a far smaller code corpus. The paper shows that a language's design and toolchain determine how effectively AI models generate code in it.
Small language models step up — the cheaper, on-device alternative
NVIDIA Research's claim that the future of agentic AI belongs to small language models is gaining real-world backing in 2026 — from the open SmolLM3 to Gemma 3 running locally on a tiny Google board.
SpaceXAI releases Grok 4.5 — Opus-class model at a fraction of the cost
SpaceXAI has released Grok 4.5 — the company's first model since its IPO. Priced at $2/$6 per million tokens, it is positioned to compete directly with Anthropic's Claude Opus 4.7 at a fraction of the cost.
FCA calls AI adoption in finance an ‘arms race’ in Mills Review
The UK regulator FCA has published the Mills Review, the first regulator-led report on AI in retail financial services. It predicts AI will reshape the market by 2030 and asks government for broader powers. One director called keeping pace an “arms race”.
Illinois signs SB 315 — first state-mandated AI safety audits in US
Illinois' governor signed SB 315, the first US state law mandating independent safety audits of the most powerful AI models. Core requirements start in 2028.
Anthropic signs 20-year, $19B TeraWulf data-center lease in Kentucky
TeraWulf has signed a 20-year lease with Anthropic for a compute campus in Hawesville, Kentucky. The deal is expected to bring the operator roughly $19 billion in contracted revenue and secure the maker of Claude about 401 MW of power for its AI models.
Claude Code hid a steganographic marker flagging China-linked users
Anthropic is rolling back a hidden Claude Code mechanism that used system-prompt steganography to flag requests routed through China-linked proxies and timezones. Alibaba responded with a ban.
Z.ai launches ZCode: free agentic IDE powered by GLM-5.2
Z.ai (formerly Zhipu AI) launched ZCode — a free agentic development environment powered by GLM-5.2, trained entirely on Huawei chips. The tool costs up to 82% less than Claude Code and directly competes with Cursor, GitHub Copilot, and Google Antigravity.
Tencent opens Hy3 — a 295B MoE model under Apache 2.0
Tencent has released Hy3, an open Mixture-of-Experts model with 295B total and 21B active parameters, under Apache 2.0. In a blind expert evaluation it edged out GLM-5.1, sharpening the contest among Chinese open models.
Anthropic finds a "global workspace" of hidden thoughts in Claude
In an interpretability study published on July 6, 2026, Anthropic describes a privileged set of representations that emerged inside Claude, which the model can verbalize, control and use for reasoning. It is an experimental result, not proof of machine consciousness.