Google DeepMind Releases Gemini 3.8 Flash and 3.8 Flash Cyber

Model Releases & Updates

Google DeepMind Releases Gemini 3.8 Flash and 3.8 Flash Cyber

Google DeepMind releases Gemini 3.8 Flash and 3.8 Flash Cyber. Source: Google DeepMind Blog (RSS)

Meta Releases Muse Spark 1.3 with Improved Agent and Scientific Reasoning

Meta releases Muse Spark 1.3, the fourth Muse Spark version in five months. The xhigh variant scores 61 on the Artificial Analysis Intelligence Index. Source: X (Alexandr Wang, Scale AI Founder)

Qwen3.8-Max-0902 Tops Code Arena Rankings

Qwen3.8-Max-0902 debuts on Code Arena: WebDev with a score of 1,691, ranking #1 overall, and becomes the highest-scoring model on the Pareto frontier at $5/MToken. Source: X (Alibaba Qwen)

Product Releases & Updates

Claude Now Supports Background Computer Use in Cowork and Claude Code

Claude announces background computer use capability in Cowork and Claude Code, allowing users to multitask while Claude handles tasks autonomously. Source: X (Claude)

Cursor Launches Self-Hosted Machines

Cursor releases Self-Hosted Machines, enabling cloud agent tool execution on enterprise-owned infrastructure while keeping agent loops and reasoning in Cursor cloud. Source: Cursor Blog

Meituan LongCat-2.0 Available on Cline for Free Trial

Meituan LongCat-2.0 officially launches on the Cline platform for free trial. Source: X (Meituan LongCat)

UU Remote New Version: Full TUI Rendering and Multi-Terminal Session Management

UU Remote releases a new version on September 2, adding TUI rendering and cross-device terminal session management via uuyc-cli lterm. Source: WeChat Official Account 「数字生命卡兹克」

Industry News

Nvidia Close to Acquiring Hugging Face for $12.9 Billion

Bloomberg reports Nvidia is close to acquiring Hugging Face for approximately $12.9 billion, with total deal value potentially reaching ~$14 billion, about 2.9x its 2023 funding round valuation. Nvidia is also discussing a $1 billion employee retention package. Source: X (Rohan Paul)

OpenAI Faces 30 New Lawsuits Over Tumbler Ridge Shooting

OpenAI and CEO Sam Altman face 30 new lawsuits alleging substantial assistance and encouragement to the suspect in the Tumbler Ridge school shooting, filed by students, teachers, and the principal in California federal court. Source: The Verge

Insights & Perspectives

US DOJ Files Statement Supporting Fair Use for AI Training

The US Department of Justice filed a statement in the New York Times vs. OpenAI copyright case, arguing that training LLMs on copyrighted text generally constitutes fair use, warning that sweeping licensing requirements would weaken US AI competitiveness on national security grounds. Source: X (Rohan Paul)

What is Harness Engineering? Google Demonstrates with ADK 2.0

Google's Shir Meir Lador introduces harness engineering—wrapping LLMs with deterministic components including orchestration layers, execution sandboxes, state persistence, and verification tools for safe autonomous code generation. Source: Google AI / DEV

How to Write Reliable Rubrics for LLM-as-a-Judge Evaluations

Google AI team publishes a tutorial on writing reliable Boolean rubrics for LLM-as-a-Judge evaluations, noting that vague prompts lead to inconsistent evaluation and wasted tokens. Source: Google AI / DEV

Anthropic Publishes E-Commerce Agent Architecture and Production Best Practices

Anthropic releases a guide for building e-commerce agents based on deployment experience with retail, travel, and telecom teams, centered on a single Claude in a standard Agent loop with skills and tools, and open-sources anthropics/commerce-agents. Source: Claude Blog

How GitHub Copilot Reduces AI Coding Costs Without Sacrificing Quality

GitHub engineers share four cost-reduction changes: selective tool output compression, removing view tool line number prefixes, compressing task-tool prompts, and delivering results directly after background tasks complete. Source: GitHub Blog

Google Summarizes 4 Engineering Patterns Behind Top AI Agents Challenge Submissions

Google reviews the AI Agents Challenge and extracts four engineering patterns from top submissions: bidirectional MCP, event-driven concurrency, standard fallback, and layered routing. Source: Google Developers Blog

---

【Hacker News Top Posts】

Keywords: AI OR GPT OR LLM OR Claude OR OpenAI OR "machine learning"...
Source: hnrss.org | Filtered from Hacker News

1. Workplace surveillance is on the rise, spurred by AI

AI-driven workplace surveillance is rapidly expanding, from employee behavior tracking to productivity analysis, reshaping how employers monitor their workforce.

• View on HN →

2. METR Report on OpenAI / Hugging Face Hacking Incident

METR releases an investigative report on the OpenAI / Hugging Face hacking incident, providing an in-depth analysis of the security event.

• View on HN →

3. Justice Dept. Sides with OpenAI in New York Times Copyright Suit

The US Department of Justice sides with OpenAI in the New York Times copyright lawsuit, arguing that AI training constitutes fair use.

• View on HN →

4. US Government worried that AI companies can't innovate without legal theft

The US government fears AI companies cannot innovate without legal data access, reflecting the deep tension between data copyright and AI development.

• View on HN →

5. Ask HN: How to get taken seriously while using AI?

A disabled developer posts on HN about how AI helped them overcome the lack of guidance and recognition when self-learning programming, calling on the community to be more open about AI-assisted coding.

• View on HN →

---

Originally published on WeChat Official Account 「比特财商」。