Gemini 2.0 Pro, Diffusion model for video restoration, OmniHuman , o3-mini, Deep Research in ChatGPT and Open-source DeepResearch, GitHub Agent mode, Arena-Price Plot, Pikadditions and more

Gemini 2.0 Pro, Diffusion model for video restoration, OmniHuman , o3-mini, Deep Research in ChatGPT and Open-source DeepResearch, GitHub Agent mode, Arena-Price Plot, Pikadditions and more

Hi. Welcome to this week's AI Brews for a concise roundup of the week's major developments in AI.

In today’s issue (Issue #92 ):

1. AI Pulse: Weekly News & Insights at a Glance

2. AI Toolbox: Product Picks of the Week

---

🗞️🗞️ AI Pulse: Weekly News & Insights at a Glance

🔥 News

1. Google’s Gemini 2.0 updates [Details]:

   1. The following models are now available in the Gemini API via Google AI Studio and in Vertex AI:

      1. Gemini 2.0 Pro, an experimental update to Google’s best model yet for coding and complex prompts with context window of 2 million tokens

      2. Gemini 2.0 Flash-Lite, a new variant that is cost-efficient

      3. Gemini 2.0 Flash is now generally available, with higher rate limits, stronger performance, and simplified pricing.

   2. Google is also rolling out a version of 2.0 Flash Thinking model in Gemini app that can interact with apps like YouTube, Search and Google Maps

   3. Imagen 3 image generation model available through the Gemini API for paid users, with a rollout to the free tier coming soon.

2. OpenAI:

   1. o3-mini model in the reasoning series launched in two variants o3-mini and o3-mini-high, available in both ChatGPT and the API [Details].

   2. Deep Research in ChatGPT, a new research agent, powered by o3 model, that conducts multi-step research on the internet for complex tasks. It’s similar to Google’s Deep Research tool available in Gemini Advanced [Details].

   3. Canvas sharing is now live in ChatGPT, allowing users to share, interact with, or edit canvases.

3. Topaz Labs introduced Project Starlight: the first-ever diffusion model for video restoration that can enhance old, low-quality videos to high-resolution [Details].

4. A new paper from the Anthropic Safeguards Research Team describes a method that defends AI models against universal jailbreaks. You can try out a demo and attempt to jailbreak a version of Claude 3.5 Sonnet that is guarded using this new technique. Anthropic is now offering $10K to the first person to pass all eight levels, and $20K to the first person to pass all eight levels with a universal jailbreak [Details].

5. Hugging Face developed "Open Deep-Research" in 24 hours as an open-source alternative to OpenAI's "Deep Research," enabling AI agents to autonomously perform tasks like web browsing and summarization. It took the #1 rank of any open submission on the GAIA leaderboard [Details | Demo].

6. Replit launched iOS and Android apps for its AI Agent-based tool for building software projects. Replit Agent is now accessible to everyone with a free tier [Details].

7. ByteDance researchers have developed an AI system, OmniHuman, that transforms single photographs into realistic videos of people speaking, singing and moving naturally. It generates full-body videos that show people gesturing and moving in ways that match their speech, surpassing previous AI models that could only animate faces or upper bodies [Details | Demos].

8. GitHub adds agent mode for GitHub Copilot in VS Code. It is capable of iterating on its own code, recognizing errors, and fixing them automatically. It can suggest terminal commands and ask you to execute them. It also analyzes run-time errors with self-healing capabilities [Details].

9. Chatbot Arena introduced Arena-Price Plot, an interactive plot of price vs. performance trade-offs for LLMs [Link]

10. Pika launched a new feature Pikadditions that lets you add any object or character from a reference image to any video [Link].

11. Mistral launched iOS and Android apps for its ‘le Chat’ assistant along with new platform features and pricing tiers[Details].

🔦 Weekly Spotlight

1. The End of Programming as We Know It by Tim O’Reilly [Link].

2. Artificial Analysis State of AI: China Q1 2025 [Link].

3. Deep Dive into LLMs like ChatGPT - a 3.5 hour video by Andrej Karpathy [Link].

4. AI Voice Agents: 2025 Update by Andreessen Horowitz [Link].

5. Open Deep Research: An Open-Source clone of Open AI's Deep Research experiment. Instead of using a fine-tuned version of o3, this method uses Firecrawl's extract + search with a reasoning model to deep research the web [Link].

6. DeepResearch by Jina AI an open-source alternative to Google and Open AI’s Deep Research [Link]

---

🔍 🛠️ AI Toolbox: Product Picks of the Week

1. Wildcard: a developer platform that combines powerful API integrations with intelligent tool selection for AI agents

2. ReelMagic: all-in-one creative storytelling platform with character and narrative consistency, access multiple AI models, and a creative AI agent that is your production co-pilot. You can now create videos of upto 3 Minutes

3. Nowadays: an AI-powered event planning copilot that takes the hassle out of organizing corporate events.

4. Pickle: Your AI body double for video calls.

Last week’s issue

Thanks for reading and have a nice weekend! 🎉 Mariam.