AI News

Latest AI news, breakthroughs, and industry trends.

Qwen3.8-2.4T Open Weights: Frontier AI Goes Open Source

Qwen3.8-2.4T Open Weights: Frontier AI Goes Open Source

Alibaba opens Qwen3.8-2.4T-A95B weights — the 2.4T-param MoE behind Qwen3.8-Max, now 91% smaller (397GB) for local runs.

Read more →
DeepSeek V4 Pro 0813 Is Live — The Real Story Isn't the Benchmark, It's Tool-First Execution

DeepSeek V4 Pro 0813 Is Live — The Real Story Isn't the Benchmark, It's Tool-First Execution

DeepSeek V4 Pro GA: DeepSWE jumps 12.8 to 62.7, Agent beats Fable 5, 3x Flash pricing, tool-first execution path.

Read more →
AI Daily Briefing – Aug 13: Brin Takes Over Gemini as 3.5 Pro Scrapped

AI Daily Briefing – Aug 13: Brin Takes Over Gemini as 3.5 Pro Scrapped

Wednesday's AI roundup: Google's co-founder steps in to steady the Gemini team, Chinese embodied intelligence posts a new global efficiency record, and AI video officially moves from single clips to full narrative. Meanwhile, money keeps flooding into robots — from fingertips to data pipelines. Top 3 Highlights 1. Brin takes over Gemini team; Gemini 3.5 Pro reportedly scrapped Google co-f …

Read more →
AMD Acquires Taalas: 17K Tokens/s Inference Chips

AMD Acquires Taalas: 17K Tokens/s Inference Chips

AMD acquires Taalas, etching model weights into silicon for 17K tokens/s inference. How one-model-per-chip works.

Read more →
HugAgentOS: ZJU's Self-Evolving Agent OS Explained

HugAgentOS: ZJU's Self-Evolving Agent OS Explained

HugAgentOS, ZJU's open-source self-evolving agent OS, turns agent experience into reusable skill via three engines.

Read more →
2026 AI Trends: Agents in Production, Cheaper Inference & More

2026 AI Trends: Agents in Production, Cheaper Inference & More

2026 AI trends: agents in production, cheaper inference, multimodal action, open-weight models — plus next steps for teams.

Read more →
Chain-of-Thought Extraction Attack on Claude & GPT

Chain-of-Thought Extraction Attack on Claude & GPT

Two API calls recover hidden chain-of-thought from Claude, GPT, and Gemini using their cheapest models. Here is the fix.

Read more →
Codex Import: What Moves From Claude Code, What Breaks

Codex Import: What Moves From Claude Code, What Breaks

The most expensive part of switching AI coding tools was never the subscription — it was accumulated setup. Here's what Codex import moves from Claude Code and what breaks in the one-way migration.

Read more →
What Is Manus? The AI Agent Platform, Explained

What Is Manus? The AI Agent Platform, Explained

Manus keeps making headlines — most recently for regaining independence from Meta and moving HQ to Singapore. Behind the drama is a genuinely useful product; here's what it is and what it can do.

Read more →
Manus Regains Independence from Meta: Data Deletion Alert

Manus Regains Independence from Meta: Data Deletion Alert

Manus is officially back on its own: the AI agent pioneer acquired by Meta is resuming operations as an independent company in Singapore. But the split comes at a hard cost — data deletion before Aug 23.

Read more →
MindTopo: VLM Spatial Reasoning Fails in Action

MindTopo: VLM Spatial Reasoning Fails in Action

Ask a vision-language model what's inside a box and it answers correctly. Ask it to act on that structure — and it fails. MindTopo benchmark reveals VLMs' spatial reasoning gap in action.

Read more →
Meta Returns to Open Source: Muse Glimmer 30B Runs Locally

Meta Returns to Open Source: Muse Glimmer 30B Runs Locally

Meta placed its biggest bet yet on the open-vs-closed AI debate: Muse Glimmer, a 30B-parameter Apache 2.0 model built for local agentic work, released August 10 with a Zuckerberg manifesto.

Read more →
Scroll to top