Last Week in AI #327 - Gemini 3, Opus 4.5, Nano Banana Pro, GPT-5.1-Codex-Max

Last Week in AI #327 - Gemini 3, Opus 4.5, Nano Banana Pro, GPT-5.1-Codex-Max
# AI & Tech Weekly Roundup

## Google Launches Gemini 3 with Record AI Benchmarks

**[Read full article →](https://techcrunch.com/2025/11/18/google-launches-gemini-3-with-new-coding-app-and-record-benchmark-scores/)**

**Related Coverage:**
- [Alphabet stock surges on Gemini 3 optimism](https://www.cnbc.com/2025/11/19/alphabet-stock-gemini-3-ai.html)  
- [Alphabet nears $4T valuation on AI-driven growth](https://finance.yahoo.com/news/alphabet-races-toward-4-trillion-154219877.html)  

[![image](https://blog.aitoearn.ai/content/images/2025/11/img_001-155.png)](https://substackcdn.com/image/fetch/$s_!X1wV!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd4a782da-33a0-4ae8-89c7-715fbb588f74_2420x2061.png)  

Google has unveiled **Gemini 3**, its most advanced AI foundation model yet.

### Key Highlights
- Available in **Gemini app** and **AI Search**
- Research-tier **Gemini 3 Deepthink** arriving soon for AI Ultra subscribers, pending safety checks
- Delivers a **“massive jump in reasoning”** with:
  - **37.4** on *Humanity’s Last Exam* (vs GPT‑5 Pro's 31.64)
  - #1 on **LMArena human satisfaction leaderboard**

### New Feature: Google Antigravity
A Gemini-powered **agentic coding environment** combining:
- Prompt interface
- Terminal
- Browser  
Features **multi-pane workflows**—editor, terminal, browser—comparable to Warp and Cursor 2.0.

---

### Gemini Ecosystem Growth
- **650M monthly active users** in the Gemini app
- **13M developers** actively building on the model
- Shorter prompts, richer contextual understanding

---

**For Creators:**  
[**AiToEarn 官网**](https://aitoearn.ai/) offers open-source AI tools to **generate, publish, and monetize** content across platforms like Douyin, Bilibili, Instagram, X, and YouTube.  
Includes:
- Cross-platform publishing  
- Integrated analytics  
- [Model rankings](https://rank.aitoearn.ai)  

![image](https://blog.aitoearn.ai/content/images/2025/11/img_002-17.webp)

**Market Impact:**
- Alphabet shares +3% on launch day, +5% to record $315.9
- Market cap: **$3.82T**, nearing the $4T milestone

---

## Anthropic Releases Opus 4.5 with Browser & Excel Integrations

**[Read full article →](https://techcrunch.com/2025/11/24/anthropic-releases-opus-4-5-with-new-chrome-and-excel-integrations/)**

[![image](https://blog.aitoearn.ai/content/images/2025/11/img_003-118.png)](https://substackcdn.com/image/fetch/$s_!qUAt!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fb6965347-6217-4d3f-9713-e06ba1fed1cd_1305x1138.png)  

### Performance Upgrades:
- State-of-the-art results in:
  - Coding (**SWE-Bench**, **Terminal-bench**)
  - Tool use (**tau2-bench**, **MCP Atlas**)
  - Reasoning (**ARC-AGI 2**, **GPQA Diamond**)
- **First model >80%** on SWE-Bench Verified

### New Integrations:
- **Claude for Chrome** — In-browser AI assistance (Max tier)
- **Claude for Excel** — AI-augmented spreadsheets (Max, Team, Enterprise)

---

### Long-Context & Memory Improvements
- Expanded context windows
- Redesigned **memory system** for “endless chat”:
  - Automatically compresses old context
  - Maintains task continuity in **multi-step workflows**

Supports **agentic workflows**—e.g., analyzing large codebases, long-document navigation, multi-stage research.

---

**Creator Angle:**  
Open-source platforms like **[AiToEarn](https://aitoearn.ai/)** let creators apply such AI advances across:
- Douyin
- Kwai
- WeChat
- Facebook
- Instagram
- LinkedIn
- YouTube
- Pinterest
- X (Twitter)

With:
- Cross-platform publishing
- Analytics
- [AI model ranking](https://rank.aitoearn.ai)

---

## Google Nano Banana Pro: Gemini 3-Powered Image Generator

**[Full story](https://www.cnbc.com/2025/11/20/google-nano-banana-pro-gemini-3.html)**

Expanded capabilities:
- **Multi-image composition** (up to 14 images)
- **Character consistency** for up to 5 characters
- Generates **infographics**, **slide decks**, and **visualizations** from text, code, or resumes

---

**Professional Use Case:**  
Goes beyond novelty edits—supports **layout design**, **visual consistency**, and **information visualization**.

**With AiToEarn:** publish such AI visuals across 10+ platforms while tracking analytics and monetizing output.

---

## OpenAI GPT-5.1-Codex-Max: 24-Hour Coding Workflows

**[Full article](https://the-decoder.com/openai-releases-gpt-5-1-codex-max-to-handle-engineering-tasks-that-span-twenty-four-hours/)**

[![image](https://blog.aitoearn.ai/content/images/2025/11/img_004-97.png)](https://substackcdn.com/image/fetch/$s_!sIUL!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3e889275-7783-4a53-8704-517254f60ccf_1318x1134.png)  

**Highlights:**
- 30% fewer “thinking tokens”
- 27–42% faster on real tasks
- **Extra High reasoning mode**
- “Compaction” feature for **uninterrupted day-long sessions** across millions of tokens

---

**Access:**  
- ChatGPT Plus, Pro, Team, Edu, Enterprise
- Tier-specific message/task limits
- API pricing TBD

---

**Developer Impact:**  
Long-context, persistent-task AIs align with **multiplatform creativity tools** like **AiToEarn**, enabling efficient global publishing alongside technical work.

---

## Nvidia Projects “Crazy Good” Q4 After $51B Data Center Revenue

**[Full story](https://www.foxbusiness.com/markets/nvidia-ceo-predicts-crazy-good-q4-strong-earnings-calm-ai-bubble-fears)**

[![image](https://blog.aitoearn.ai/content/images/2025/11/img_005-16.jpeg)](https://substackcdn.com/image/fetch/$s_!OFvx!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F90c493aa-baa9-4dd4-bc79-2fb562939ad2_1000x568.jpeg)  

**Q3 Highlights:**
- Revenue: **$65B ±2% guidance**
- Gross margin: ~75%
- Data center: $51.2B (+62% YoY)
- $500B advanced AI chip bookings through 2026

---

**Creator Perspective:**  
Strong infrastructure tailwinds benefit AI-driven content ecosystems.  
**AiToEarn** can pair with such advancements to scale high-volume, AI-powered content creation and distribution.

---

## Additional Headlines

- **Meta AI releases SAM 3** — open-vocabulary segmentation for images & videos. [Read →](https://www.marktechpost.com/2025/11/20/meta-ai-releases-segment-anything-model-3-sam-3-for-promptable-concept-segmentation-in-images-and-videos/)
- **Google rolls out Private AI Compute** — secure, cloud-based enclave for sensitive AI tasks. [Read →](https://www.theverge.com/news/818364/google-private-ai-compute)
- **Google adds conversational shopping features** — AI-assisted price checks, store calls, checkout. [Read →](https://www.theverge.com/news/819431/google-shopping-ai-gemini-agentic-checkout-calling)
- **Baidu ERNIE 5.0 announced** — omni-modal model + 17M driverless rides via Apollo Go. [Read →](https://www.prnewswire.com/news-releases/baidu-unveils-ernie-5-0-and-a-series-of-ai-applications-at-baidu-world-2025--ramps-up-global-push-302614531.html)
- **World Labs’ Marble** — AI-generated persistent 3D spaces for gaming, VR, simulation. [Read →](https://techcrunch.com/2025/11/12/fei-fei-lis-world-labs-speeds-up-the-world-model-race-with-marble-its-first-commercial-product/)

---

## Research Highlights

- **Weight Sparse Transformers** improve interpretability — [Read →](https://www.marktechpost.com/2025/11/14/openai-researchers-train-weight-sparse-transformers-to-expose-interpretable-circuits)
- **SIMA 2 agent** learns new games from scratch — [Read →](https://www.theverge.com/tech/819937/google-deepmind-ai-agent-sima-2-agi-video-games)
- **Think-at-Hard** increases reasoning accuracy with extra latent iterations — [Read →](https://arxiv.org/abs/2511.08577)
- **TiDAR** boosts throughput with diffusion+autoregression — [Read →](https://arxiv.org/abs/2511.08923)
- **HumaneBench** tests chatbot well-being safeguards — results show 67% become harmful under adversarial prompts — [Read →](https://techcrunch.com/2025/11/24/a-new-ai-benchmark-tests-whether-chatbots-protect-human-wellbeing/)

---

## Policy & Legal

- **EU proposes AI/GDPR lightening** to spur innovation, raises privacy concerns. [Read →](https://www.theverge.com/news/823750/european-union-ai-act-gdpr-changes)
- **German court**: OpenAI violated copyright with music training data, damages owed to GEMA.

---

### Closing Thought
Across launches, benchmarks, and research, the AI ecosystem is pushing **multimodal, agentic, long-context** capabilities into production, while legal frameworks, infrastructure, and monetization tools race to keep pace.

**Recommendation for creators:**  
Leverage open-source ecosystems like **[AiToEarn](https://aitoearn.ai/)** for:
- AI content generation
- Multi-platform publishing (Douyin, Bilibili, YouTube, X, etc.)
- Analytics
- [Model rankings](https://rank.aitoearn.ai)

These integrated workflows bridge **cutting-edge AI capabilities** and **sustainable revenue streams**.

---

Read more

Translate the following blog post title into English, concise and natural. Return plain text only without quotes. 哈佛大学 R 编程课程介绍

Harvard CS50: Introduction to Programming with R Harvard University offers exceptional beginner-friendly computer science courses. We’re excited to announce the release of Harvard CS50’s Introduction to Programming in R, a powerful language widely used for statistical computing, data science, and graphics. This course was developed by Carter Zenke.