A new open-source voice model claims to beat the best

Read time: under 4 minutes

Welcome back, Superhuman. AI is changing the internet in unexpected ways. We are on track to have more AI-generated content on the internet than human-generated content by the end of the decade. Will the internet still be a place where humans can hang out and connect? I asked one of the most interesting people in AI for the answer.

Today’s Insights

• Nvidia and xAI team up, brain-inspired AI, and a realistic voice model

• 5 big insights on the future of AI

• Tutorial: How to add objects to any image with Gemini 2.0 Flash

• 5 new AI tools to boost your productivity

• News, memes, what’s trending on socials, and more

TODAY IN AI

1. Nvidia and xAI sign onto massive AI infrastructure project: OpenAI’s Project Stargate might have finally met its match. Nvidia and xAI announced they’re joining Microsoft, BlackRock, and investment firm MGX’s initiative to put as much as $100B into new AI infrastructure projects in the US. Microsoft could use the new data centers to power its rumored in-house model, MAI-1, which would likely put even more strain on its already-fraying relationship with OpenAI.

2. Researchers recreate mammalian brain to help AI learn faster: One of the things that sets animals apart from AI is that they can learn from real-life experiences, not just data. With that in mind, Microsoft is teaming up with Swiss startup Inait to turn 20 years of neuroscience research into a full-scale simulation of the mammalian brain. The model uses 18 million lines of code to give LLMs “true cognition” — bringing more complex reasoning to fields like finance and robotics.

3. Open-source voice model allegedly beats top rivals: An SF-based startup with backing from a16z just released a new text-to-speech model that it says sounds just as realistic as offerings from OpenAI and ElevenLabs — except it’s completely open-source. Coming in four different sizes, Canopy Labs’ Orpheus can generate “high-quality, aesthetically pleasing” speech, while incorporating sighs, laughs, and other expressions that add to the realism. Check out some demos of the model in action here.

PRESENTED BY SANA

Do real work with agentic AI

Sana’s all-in-one AI platform lets you build no-code custom agents for any team or role grounded in all of your company’s data.

Imagine working with AI that can:

• Find the exact info you need

• Answer and brainstorm any idea

• Recap meeting actions and to-dos

• Automate repetitive tasks

• Solve complex, multi-step problems

Plus: Extensible APIs. Enterprise-grade security. AI solutions engineering and strategy support.

Try an AI agent and see the difference

THE FUTURE OF AI

AI is making human content even more valuable — Google AI’s Logan Kilpatrick

Logan Kilpatrick is one of the most unique people in AI — he’s one of the only individuals in the world to have worked at both OpenAI and Google AI. Few people in AI have the kind of front row exposure to what’s happening on the bleeding edge that Logan has.

So we sat down with him to learn what’s happening in AI and what’s next. Here are 5 big things we learned from him:

1. There’s never been a better time to be a human than today. The amount of AI content on the internet is trending to exceed the amount of human content on it. This divergence will make human content even more valuable in the coming years.

2. And there's never been a better time to build stuff than today. AI models like Gemini 2.0 Flash prove that building with AI is getting faster, better, cheaper — all of this is great news for people who want to build new products and work with new technologies like agents.

3. Dedicated human community spaces will emerge as AI content overruns the internet.  We need humans to be able to connect and if the current places in which people do that on the internet aren't accessible to humans anymore, we’ll have to create new spaces of our own to connect with each other.

4. Live API is the most exciting thing to watch out for in the coming months. The Live API essentially enables a live AI co-presence. So instead of constantly prompting the AI, you can just share your screen or camera and it will work alongside you in real time.

5. AI models are smart enough — they just have poor memories. On a lot of tasks, AI models are smart enough to do what we want them to do, they’re just limited by how much they can remember about a given task. To fulfil the real promise of AI, the next big technical step is to create models with more memory and larger context windows.

We also talked about how AI will coach you on your job, pick up tasks in the background, create novel scientific breakthroughs, and more.

👉 Watch the full conversation here

THE AI ACADEMY

How to add objects to any image with Gemini 2.0 Flash

• Go to AI Studio’s website and sign up.

• Select ‘Gemini 2.0 Flash (Image Generation) Experimental’ as your model.

• Upload your image and just prompt to specify the object, position, and style.

Sample Prompt: “Add a [enter an object] in [enter position] in a [enter style].”

Alternatively, you can also upload the object’s image and prompt to add it to the existing image.

• Click ‘Generate’ and wait for the AI to process the edit.

• If needed, refine your request (e.g., adjust placement or details).

• Once satisfied, download the enhanced image, and share or integrate it into your projects.

PRESENTED BY GENESYS

AI: The Key to Personalized, Faster CX

AI is no longer just an option—it's a strategic necessity. With 70% of CX leaders recognizing AI as a business imperative, the time to invest is now.

Learn how AI-powered technologies are transforming customer experiences and driving loyalty.

Read the State of CX report

AI & TECH NEWS

Everything else you need to know today

💵 Pricey Prompts: OpenAI released a new version of its o1 model, o1-pro, aimed at developers. It delivers “consistently better responses,” but it’s also its most expensive model yet.

🖼️ Visual Ventures: xAI has added image-generation capabilities to its API, powered by the "grok-2-image-1212” model. Priced at $0.07 per image, it sits right between rivals Black Forest Labs ($0.05) and Ideogram ($0.08).

🧠 Thought Theater: NotebookLM rolled out an "Interactive Mindmaps" feature to help users explore ideas through playful visualization — a possible shift toward personalized AI experiences.

📰 News Bytes: An Italian newspaper published the world's first entirely AI-generated issue — a 4 page insert where journalists were limited to chatbot interactions.

💰 Funding: Analytics giant Dataminr secured $85M in pre-IPO funding from NightDragon and HSBC, aiming to expand globally and fuel new product development. Meanwhile, AI startup Prezent raised $20M to scale its generative AI-powered slide-deck platform, aiming to overhaul business presentations for more than 150 Fortune 2000 companies.

PRODUCTIVITY

5 AI Tools to Supercharge Your Productivity

✅ Vidu AI: Create high-quality videos from text and images with AI.

✅ MaxAI: Chat with any webpage to summarize, explain, analyze, write, and more.

✅ You. com*: Get 1 year free of you. com Pro, the best of AI. Access 20+ AI models in one, centralized easy-to-use AI chat platform, including the latest models from OpenAI, DeepSeek, Anthropic, and Gemini.

✅ Mermaid: Build complex collaborative diagrams from text and data in seconds.

✅ Miro: An AI-powered collaborative workspace that helps teams move faster from idea to outcome.

^{* indicates a promoted tool, if any}

PROMPT OF THE DAY

Review Progress

Prompt: Review my productivity over the past [week/month] by analyzing my goals and how much I have achieved. My goals were: [list goals]. Here’s my progress so far: [describe progress]. Compare this with my previous performance data from [specific time period, if available]. Identify trends, strengths, and areas for improvement. Based on this analysis, provide actionable steps to optimize my productivity and help me stay on track with my goals.

SOCIAL SIGNALS

What’s trending on socials today

🎨 Vibe Win: This Claude 3.7 UI workflow is turning out to be a gold mine for vibe coders. They can use it to design entire UIs screen by screen, make project plans, and break them down into components.

🛌 In Your Dreams: An AI-generated simulation of what it feels like to “fly” while dreaming has got Redditors smashing the like button.

🩺 Medical Marvel: A patient’s chronic illness had a physician stumped for years. After feeding the patient’s biomedical data into Grok, the AI suggested a previously unconsidered treatment which reportedly cured the patient.

📝 Thought Tuning: A Reddit user claims to have "reverse-engineered" ChatGPT’s thought process to enhance its responses. The technique involves instructing the AI to analyze, critique, and consider multiple perspectives before answering.

AI-GENERATED IMAGES

Digital Bloom

Midjourney Prompt: Flowers grew out of a white computer, surrounded by flowers and grass in the style of cinema4d rendering and in the style of clay sculpture, kawaii aesthetic, spring garden, bright colors, with a dreamy atmosphere, blooming colors in bright daylight, virtual reality, fantasy world, --ar 9:16

Acquire new customers and drive revenue by partnering with us

Superhuman is the world’s biggest AI newsletter for businesses and professionals with 1M+ readers and 2M+ followers on socials working at the world’s leading startups and enterprises. Companies like Amazon, Hubspot, and Salesforce feature their products in Superhuman. You can learn more about partnering with us here.

🧞 Your wish is my command

Got more feedback or just want to get in touch? Reply to this email and we’ll get back to you.

Thanks for reading.

Until next time!

Zain & the Superhuman AI team