Good morning, AI enthusiasts. OpenAI has “a lot of good stuff” lined up this week, according to Sam Altman—and its first release is a step back…in name only.

A newly launched GPT-4.1 (?) family features million-token context windows, improved coding abilities, and significantly lower prices across the board — potentially laying a new foundation for the fast-approaching era of agentic AI development.

In today’s AI rundown:

• OpenAI’s dev-focused GPT-4.1 family

• ByteDance’s efficient Seaweed video AI

• Create conversational branches to explore ideas

• Google’s AI to decode dolphin speech

• 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

OPENAI

🤖 OpenAI’s dev-focused GPT-4.1 family

Image source: OpenAI

The Rundown: OpenAI just released GPT-4.1, a new API-only model family built specifically for developers — featuring major improvements in coding abilities, instruction following, and the ability to process up to 1M tokens of context.

The details:

• The new API-only lineup includes GPT-4.1, 4.1 mini, and 4.1 nano, significantly outperforming GPT-4o on key developer tasks.

• All three models support 1M token contexts, enough for 8 full React codebases, while being 26% cheaper than GPT-4o for typical queries.

• The models also show gains in real-world tasks like frontend development, with evaluators preferring 4.1's web interfaces 80% of the time over GPT-4o.

• Pricing is reduced across the board, with GPT-4.1 coming in 26% cheaper than GPT-4o and 4.1 nano appearing as OpenAI's fastest and cheapest model yet.

Why it matters: The only thing moving backwards is OpenAI’s naming convention — but GPT-4.1 is a major leap forward for devs. With a massive context window, lower costs, and sharper focus, it sets a new foundation for the agentic coding and may be a precursor to the company’s rumored Agentic Software Engineer.

TOGETHER WITH HUBSPOT

🧠 100+ ChatGPT prompts to revolutionize your workflow

The Rundown: HubSpot’s free, comprehensive  “How to Use ChatGPT at Work” guide provides 100+ ready-to-use prompts to help professionals boost efficiency and adopt AI-driven workflows.

Inside, you’ll find:

• A quick crash course to master ChatGPT in under 30 minutes

• Practical industry use cases to spark real-world inspiration

• 100+ prompts to streamline tasks and accelerate productivity

• Expert tips to tackle common AI roadblocks with confidence

Get your free copy and join 10,000+ professionals leveling up with AI.

BYTEDANCE

🎥 ByteDance’s efficient Seaweed video AI

Image source: ByteDance

The Rundown: ByteDance introduced Seaweed, a hyper-efficient 7B-parameter video generation model that is competitive against much larger models like Kling 1.6, Google Veo, and Wan 2.1, despite using significantly less compute resources.

The details:

• Seaweed features multiple generation modes, including text-to-video, image-to-video, and audio-driven synthesis, with outputs going up to 20 seconds.

• The model ranks highly against rivals in human evaluations and excels in image-to-video tasks, massively outperforming models like Sora and Wan 2.1.

• It can also handle complex tasks like multi-shot storytelling, controlled camera movements, and even synchronized audio-visual generation.

• ByteDance says Seaweed has been fine-tuned for applications like human animation, with a strong focus on realistic human movement and lip syncing.

Why it matters: Between Wan (Alibaba), Kling, and now ByteDance’s Seaweed, China is absolutely crushing the AI video leaderboards. This byte-sized (pun intended) release also shows that scale isn’t the only path to top-tier video generation, opening up efficient, limitless creativity with readily available, near-SOTA video models.

AI TRAINING

🔀 Create conversational branches to explore ideas

The Rundown: In this tutorial, you will learn how to use Google AI Studio's new branching feature to explore different ideas by creating multiple conversation paths from a single starting point without losing context.

Step-by-step:

1. Visit Google AI Studio and select your preferred Gemini model from the dropdown menu.

2. Start a conversation and continue until you reach a point where you want to explore an alternative direction.

3. Click the three-dot menu (⋮) next to any message and select "Branch from here."

4. Navigate between branches using the "See original conversation" link at the top of each branch.

Pro tip: You can create branches at key decision points to compare different AI approaches to the same problem without starting over.

PRESENTED BY UNSTRUCTURED

🛠️ Simplifying RAG with Unstructured + AstraDB

The Rundown: Unstructured just introduced a new feature that made building a knowledge graph easier than ever: Custom Prompting for NER to build your nodes and edges.

Join Unstructured’s live session to learn how to:

• Use Unstructured API to load data into Astra DB

• Utilize the Custom Prompting Feature & OSS Graph Retriever library

• Leverage the Graph Retriever for dynamic retrieval

Register here to join live or to get the recording.

AI RESEARCH

🐬 Google’s AI to decode dolphin speech

Image source: Google

The Rundown: Google unveiled DolphinGemma, a specialized AI model designed to analyze and generate dolphin vocalizations — designed in collaboration with researchers at Georgia Tech to potentially uncover patterns in their communication.

The details:

• DolphinGemma leverages Google's Gemma and audio tech to process dolphin vocalizations, trained on decades of data from the Wild Dolphin Project.

• The AI model analyzes sound sequences to identify patterns and predict subsequent sounds, similar to how LLMs handle human language.

• Google also developed a Pixel 9-based underwater CHAT device, combining the AI with speakers and microphones for real-time dolphin interaction.

• The model will be released as open-source this summer, allowing researchers worldwide to adapt it for studying various dolphin species.

Why it matters: While previous attempts at dolphin communication have struggled, combining decades of research with modern AI could finally open the door for new understanding of how these intelligent creatures communicate. If successful, DolphinGemma could open new frontiers in understanding animal intelligence.

QUICK HITS

🛠️ Trending AI Tools

• 🧠 ChatGPT - New memory feature that remembers all previous conversations

• ❎ Grok 3 - xAI’s top model, now also with new memory capabilities

• 🎨 Canva Visual Suite 2.0 - Create across all design types with AI

• 🤖 Appsmith Agents - Secure, embedded agents powered by your data

💼 AI Job Opportunities

• 🧾 Weights & Biases - Deal Desk Manager

• 💼 Horizon3 - Enterprise Account Executive

• 🛎️ Rad AI - Customer Support Engineer

• 🧑‍💻 Perplexity AI - AI Software Engineer

📰 Everything else in AI today

NVIDIA announced its first-ever U.S. AI manufacturing effort, partnering with TSMC, Foxconn, and others to begin chip and supercomputer production in Arizona and Texas.

OpenAI is reportedly planning to release two new models this week, with o3 and o4-mini capable of creating new scientific ideas and automating high-level research tasks.

Amazon CEO Andy Jassy published his annual shareholder letter, saying that genAI will "reinvent virtually every customer experience we know."

Meta announced plans to train AI models on EU users’ public content, offering an opt-out form and noting the importance of incorporating European culture into its systems.

Hugging Face acquired Pollen Robotics and introduced Reachy 2, a $70k open-source humanoid robot designed for research and embodied AI applications.

LM Arena launched the Search Arena Leaderboard to evaluate LLMs on search tasks, with Google’s Gemini-2.5-Pro and Perplexity’s Sonar taking the top spots.

NATO awarded Palantir a contract for its Maven Smart System to enhance U.S. battlefield operations with AI capabilities, aiming to deploy the platform within 30 days.

COMMUNITY

🎥 Join our next live workshop

We just did a full workshop on n8n, where we teach you how to create your own AI assistant that speeds up your workflow and helps you get more done, led by Dr. Alvaro Cintas, The Rundown’s AI professor.

Watch it here. Not a member? Join The Rundown University on a 14-day free trial.

🤝 Share The Rundown, get rewards

We’ll always keep this newsletter 100% free. To support our work, consider sharing The Rundown with your friends, and we’ll send you more free goodies.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team