AI talks to dolphins, robot marathon, new deepfakes, o3 & o4-mini, new AI video model
The week in AI: dolphins, robots, deepfakes and OpenAI's smartest models yet
It's been another whirlwind week in AI development with breakthroughs that range from the practical to the downright mind-bending. From communication with dolphins to marathon-running robots, here's a breakdown of the most significant AI developments that caught my attention.
Google's Dolphin Gemma: AI-powered animal communication
Google has developed "Dolphin Gemma," a lightweight AI model designed to analyze and potentially communicate with dolphins. The system can run on a smartphone and process dolphin vocalizations in real-time, including whistles, clicks, and buzzes.
What makes this particularly interesting is that Dolphin Gemma can not only identify patterns in dolphin sounds but also generate new sounds in "dolphin language." The model was trained using Google's Soundstream technology, which converts audio into tokens similar to how text prompts are processed by language models.
At just 400 million parameters, it's compact enough to run on a Google Pixel phone. Google plans to open-source the model this summer, potentially allowing researchers to adapt it for other animal species. While we may not have AGI yet, the ability to communicate with other species could be a genuinely revolutionary application of AI.
Character animation breakthroughs
Two new tools are making waves in the character animation space:
Uni Animate DIT is a plugin for the open-source video generator One 2.1 that allows you to animate any photo using reference pose videos. It works by transferring motion from a reference video to your character, handling everything from basic movements to complex finger gestures. The tool supports both realistic photos and stylized characters, even generating plausible views of parts of characters not visible in the original image. It requires 14GB of VRAM to run.
Instant Character by Tencent lets you add reference characters to your images with remarkable fidelity. Using Flux as its base model, it can transfer characters to new scenes while preserving their appearance with impressive accuracy. In comparison tests with GPT-4o and other tools, Instant Character provided the most accurate character transfers. It's available as a free Hugging Face demo, with instructions for local installation also available.
Recent Videos
Hermes Agent Master Class
https://www.youtube.com/watch?v=R3YOGfTBcQg Welcome to the Hermes Agent Master Class — an 11-episode series taking you from zero to fully leveraging every feature of Nous Research's open-source agent. In this first episode, we install Hermes from scratch on a brand new machine with no prior skills or memory, walk through full configuration with OpenRouter, tour the most important CLI and slash commands, and run our first real task: a competitor research report on a custom children's book AI business idea. Every future episode will build on this fresh install so you can see the compounding value of the agent in real time....
Apr 29, 2026Andrej Karpathy – Outsource your thinking, but you can’t outsource your understanding
https://www.youtube.com/watch?v=96jN2OCOfLs Here's what Andrej Karpathy just figured out that everyone else is still dancing around: we're not in an era of "better models." We're in a different era of computing altogether. And the difference between understanding that and not understanding it is the difference between being a vibe coder and being an agentic engineer. Last October, Karpathy had a realization. AI didn't stop being ChatGPT-adjacent. It fundamentally shifted. Agentic coherent workflows started to actually work. And he's spent the last three months living in side projects, VB coding, exploring what's actually possible. What he found is a framework that explains...
Mar 30, 2026Andrej Karpathy on the Decade of Agents, the Limits of RL, and Why Education Is His Next Mission
A summary of key takeaways from Andrej Karpathy's conversation with Dwarkesh Patel In a wide-ranging conversation with Dwarkesh Patel, Andrej Karpathy — former head of AI at Tesla, founding member of OpenAI, and creator of some of the most popular AI educational content on the internet — shared his views on where AI is headed, what's still broken, and why he's now pouring his energy into education. Here are the key takeaways. "It's the Decade of Agents, Not the Year of Agents" Karpathy's now-famous quote is a direct pushback on industry hype. Early agents like Claude Code and Codex are...