🤖 The White House Received Its First Robot Guest

Melania Trump turned heads at a global education summit with a vision for the classroom of tomorrow. Showcasing a humanoid AI teacher, she invited audiences to imagine a future where learning is fully personalized and available on demand at home. Let’s dive in!

In today’s insights:

LATEST DEVELOPMENTS

Evolving AI: Melania Trump debuted a humanoid robot at a global education summit, pitching AI as the future of personalized learning.

The event was part of Melania Trump's "Fostering the Future Together" global coalition summit, where she invited attendees to imagine a classroom led by a humanoid AI educator capable of teaching literature, science, philosophy and history on demand at home. The initiative brought international leaders together to discuss how AI and educational technology can empower children. Figure AI's CEO Brett Adcock called it a historic moment, saying he was "proud to see F.03 make history as the first humanoid robot in the White House." Earlier that same day President Trump appointed Meta's Mark Zuckerberg, Oracle's Larry Ellison and Nvidia's Jensen Huang to a new AI policy council, framing artificial intelligence as a defining arena of competition with China. The thinking that AI can automate and accelerate learning has been gaining momentum in tech circles, with experiments like Alpha School (a private network using AI to teach children at an accelerated pace) drawing growing attention and White House endorsement.

The US-China race in humanoid robotics is moving fast, and education is now part of the battleground. China made up the majority of the roughly 16,000 humanoid robots sold in 2025 and hosted the first-ever World Humanoid Robot Games in Beijing last August, bringing together 280 teams from 16 countries. Against that backdrop, walking a robot down the White House red carpet is less of a gimmick and more of a statement. But the actual research tells a more cautious story: studies consistently show higher student engagement with humanoid robots, but learning gains remain small to moderate, and every systematic review flags cost, technical reliability, and ethics as serious hurdles. The push to frame AI as a personalized teacher for every child also sidesteps a question no one in the room seemed to ask: if the robot delivering lessons carries biases baked in during training, who is responsible for what it teaches?

Viktor is an AI coworker that lives in Slack, right where your team already works.

Message Viktor like a teammate: "pull last quarter's revenue by channel," or "build a dashboard for our board meeting."

Viktor connects to your tools, does the work, and delivers the actual report, spreadsheet, or dashboard. Not a summary. The real thing.

There’s no new software to adopt and no one to train.

Most teams start with one task. Within a week, Viktor is handling half of their ops.

Add Viktor to Slack for free.

Evolving AI: Meta's FAIR team has released TRIBE v2, a foundation model that predicts how the human brain responds to video, audio, and text stimuli using fMRI data.

TRIBE v2 is a tri-modal brain encoding model built by Meta's FAIR team. It takes video, audio, and text as inputs and predicts high-resolution fMRI responses across the entire brain, covering over 20,000 cortical vertices and nearly 9,000 subcortical voxels. Rather than learning perception from scratch the model uses frozen feature extractors from AI systems and passes their outputs through a shared Transformer that processes 100-second windows of naturalistic stimuli. A subject-specific prediction block then maps those representations to individual brain activity. One of the most striking findings is that TRIBE v2 follows a log-linear scaling law. Beyond prediction TRIBE v2 was also tested on virtual neuroscience experiments. It successfully identified classic brain regions like the fusiform face area and Broca's area purely through digital simulation.

Neuro-AI has had a genuinely big 12 months. BCIs are moving from labs to real patients, memory prostheses are being tested in humans, and researchers are starting to treat the brain less like a mystery and more like a system that can be modeled and simulated. TRIBE v2 fits right into that shift. Rather than running expensive fMRI sessions to test every new hypothesis, neuroscientists could use a model like this as a digital stand-in, running virtual experiments before ever putting someone in a scanner. That alone could speed up research into conditions like aphasia or sensory processing disorders significantly. But the deeper thing worth sitting with is what TRIBE v2 revealed about AI without anyone planning it. Its internal representations naturally sorted themselves into the same functional networks the brain uses. That is not a party trick. It suggests that today's large AI models may already be organized in ways that mirror human perception, not because anyone designed them that way, but because they were trained on the same kind of messy, multimodal world we live in.

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

Download the free guide

Evolving AI: Mistral has released Voxtral TTS, an open source text-to-speech model built for enterprise voice applications.

French AI company Mistral released Voxtral TTS on Thursday targeting voice AI assistants and enterprise use cases like customer support and sales. The release puts Mistral in direct competition with ElevenLabs, Deepgram, and OpenAI. The model supports nine languages including English, French, German, Spanish, Dutch, Portuguese, Italian, Hindi, and Arabic. It can adapt to a custom voice using a sample of less than five seconds and captures characteristics like subtle accents, inflections, and irregularities in speech flow.

Earlier this year Mistral also launched two transcription models: one for batch processing and one for low-latency real-time use. Voxtral TTS adds a speech generation layer to that foundation. Mistral's VP of Science Operations, Pierre Stock, said the company aims to build an end-to-end platform handling multimodal streams of audio, text, and image input and output.

Voice AI is a hot corner of tech right now. Production deployments grew 340% year over year and Gartner predicts conversational AI will cut contact center labor costs by $80 billion in 2026 alone. The race is on and the main battleground is shifting from who sounds most human to who can do it cheapest at the edge. ElevenLabs is among the priciest options at around $0.30 per 1,000 characters which makes sense for audiobooks and creative work but falls apart fast when you're routing thousands of support calls through it. Mistral is betting that open source plus edge deployment is the smarter play for enterprise scale and the timing feels right.

Vibe coding is hot right now. This video from Google shows how you can build a full website just by describing the look and feel, no coding needed. If you’ve ever wanted to build something without touching code, this is worth a watch.

QUICK HITS

❌ Judge rejects Pentagon's attempt to 'cripple' Anthropic.

💬 Google Gemini now lets you import your chats and data from other AI apps.

🔞 OpenAI drops plans to release an adult chatbot.

📈 Claude AI Maker Anthropic Considers IPO as Soon as October.

 *partner link