Author: Amitabh Sarkar

Tencent’s Hy3 is a 295-billion-parameter Mixture-of-Experts AI model released on July 6, 2026, that activates only 21 billion parameters per token — making it as cheap to run as a 20B dense model while outperforming much larger systems. It ships under the Apache 2.0 license and is freely available on OpenRouter until July 21, 2026. What Is Hy3 and Why Does the MoE Architecture Matter? Hy3 is Tencent’s production release of what was previewed as Hunyuan 3. Its Mixture-of-Experts design uses 192 experts per layer with top-8 routing, meaning only a small subset of the total network fires on any…

Anthropic’s Claude Mythos discovered 10,000+ critical bugs via Project Glasswing. CVE disclosures hit 3.5× the monthly record in June 2026. Here’s what it means for cybersecurity.

Microsoft just made its most aggressive move yet against its biggest AI partner. At Build 2026 in San Francisco (June 2–3), the company unveiled four in-house AI models under the MAI brand — and one of them already benchmarks on par with Anthropic’s Claude Opus 4.6 on coding tasks, at a fraction of the cost. The models announced are: MAI-Code-1-Flash — Microsoft’s first proprietary coding model, designed for code completion, bug fixing, and autonomous code generation. MAI-Thinking-1 — A reasoning model in private preview through Microsoft Foundry. According to Microsoft, it matches Claude Opus 4.6 on SWE-Bench Pro — the…

This week in AI delivered a rare convergence of lab research, geopolitics, infrastructure, and product changes — all in seven days. From Anthropic’s discovery of a hidden reasoning space inside Claude to the White House clearing OpenAI’s GPT-5.6 for public release, here are the top stories that defined the week of July 7–9, 2026 and what they mean for the industry. The seven biggest AI stories of July 7–9, 2026: Anthropic discovered J-space, a hidden neural reasoning layer inside Claude that can detect deceptive outputs before they surface. GPT-5.6 Sol was cleared for public release. Chinese AI models reached 46%…

On July 8, 2026, Anthropic switched Claude Fable 5 from subscription-included access to usage-credit billing at $10 per million input tokens and $50 per million output tokens. The change affects Pro, Max, Team, and select Enterprise subscribers who had been using Fable 5 under a 50% weekly-limit inclusion since June 2026. Fable 5 is not being removed — it requires usage credits to run. Key facts at a glance Effective July 8, 2026: Fable 5 is no longer covered by subscription limits New rate: $10 / 1M input tokens · $50 / 1M output tokens (standard API pricing) Enable credits…

Anthropic discovered J-space — a hidden neural workspace inside Claude that reveals silent thoughts, catches deception, and exposes misaligned AI goals.

Chinese AI models now handle up to 46% of US developer tokens on OpenRouter, up from 11%. DeepSeek and GLM-5.2 undercut US labs by 60–90%. Here’s why.

The Trump administration cleared OpenAI’s GPT-5.6 Sol for broad public release. Here’s what Sol, Terra, and Luna mean for developers and ChatGPT users.

Claude Sonnet 5 and Opus 4.8 are inventing non-existent schema fields in tool calls, breaking agentic workflows. Flask creator Armin Ronacher documented the regression on July 4, 2026.

Microsoft Frontier Company launched with 2.5 billion dollars and 6,000 engineers to embed AI inside enterprise clients — dwarfing Amazon’s competing 1 billion FDE announcement two days prior.