Howdy! It’s Barsee again.

Happy Tuesday, AI family, and welcome back to AI Valley.

Today’s climb through the Valley reveals:

• Japan’s new tech enables human-to-robot teleoperation in real-time

• ElevenLabs’ new voice-first personal assistant

• Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

FLOW

Stop typing. Start Flow-ing ✨

The fastest way to move work forward isn't a new shortcut—it's your own voice.

Wispr Flow now runs natively on Mac, Windows, and iPhone, so the same friction-free dictation follows you from the commute to your desk and back again.

Why teams are switching:

• 4× faster than typing. Speak at 140 wpm and let Flow handle the punctuation, re-phrasing, and formatting.

• Real-time AI edits & commands. Mumble, back-track, or change your mind, Flow cleans it up before you hit Send.

• Personal dictionary that travels with you. Learn names, acronyms, and industry jargon once, then sync them to every device.

• Tone awareness for every app. Sound crisp in Gmail, casual in Slack, and precise in VS Code without lifting a finger.

• 100+ auto-detected languages. Toggle effortlessly between English, Spanish, Polish, and more.

• Built for privacy & scale. SOC 2-ready and trusted by thousands of professionals who can't afford typos—or downtime.

One hotkey. Any workflow.

• Mac: Press Fn to start speaking in Notion, Figma, or anywhere the cursor blinks.

• Windows: Tap Ctrl + Win and dictate Jira updates, code comments, or long emails.

• iPhone: Whisper into the mic on the subway—Flow silences background noise and outputs publication-ready text.

Everything you say appears instantly across devices; your notes, drafts, and custom terms stay perfectly in sync.

Ready to Flow?

Grab the latest build for your platform of choice (or all three) and feel the speed bump on your very next message.

→ Download Flow free

(PS: Already using Flow on one device? Install it on the rest in minutes and watch your productivity compound.)

_{*This is sponsored}

PEAK OF THE DAY

Japan’s new tech enables human-to-robot teleoperation in real-time 🦾

A major breakthrough in human-robot interaction has been introduced by Japanese startup H2L, with their "Capsule Interface" system that allows real-time, full-body control of robots with unparalleled precision.

The system’s focus on muscle activity lets robots not only mimic a user’s movements but also the strength behind them.

Here’s what you need to know:

• The Capsule Interface uses advanced muscle displacement sensors to capture both movement and applied force, instantly translating a user’s physical actions into exact robot motions.

• Unlike traditional motion tracking systems (IMUs/exoskeletons), the Capsule Interface detects small changes in muscle tension, preserving the nuances of strength and effort during interactions.

• The system has been demonstrated controlling a Unitree H1 humanoid robot, performing complex tasks such as cleaning, lifting boxes, and engaging with humans with lifelike physicality.

• The Capsule Interface operates through chair or bed-mounted systems that require no special training. Users can control robots simply by using natural movements while seated or lying down.

• With a 300Hz sampling rate for muscle activity and less than 5ms latency, the system ensures smooth, embodied control of robots.

• The system also integrates display and audio components, offering complete immersive feedback during robot operations.

Why it matters:

H2L’s technology represents a major leap in human-robot collaboration, making advanced robotic control easier and more intuitive. This system is the first to offer force-aware teleoperation, allowing for more nuanced, full-body control of robots without complex equipment. It opens up robotic control to a wider audience and has potential applications in remote work, healthcare, and dangerous environments by enabling users to perform physical tasks remotely with force feedback and helping mobility-limited individuals carry out tasks independently.

THROUGH THE VALLEY

1. Court filings reveal OpenAI and io's early work on an AI device

Court documents show that OpenAI and Jony Ive’s startup, io, are working on a new AI hardware device. While some thought it would be a wearable, OpenAI says the prototype isn’t wearable. They’ve explored different formats, from desktop to portable devices. These filings came during a lawsuit from Google-backed iyO, which claims OpenAI copied its product. OpenAI denies this. The device is still a year away and aims to offer an interface beyond phones and laptops.

2. Apple explores buying AI‑search startup Perplexity

Apple is in talks to acquire Perplexity, an AI-powered search startup valued at approximately $14 billion, according to sources close to Bloomberg. While discussions are still in their infancy, Apple has yet to extend an offer or initiate contact, and key executives like Eddy Cue and Adrian Perica are already involved. Should this deal go through, it would mark Apple’s largest acquisition ever, eclipsing its $3 billion purchase of Beats in 2014. Acquiring Perplexity would bolster Siri and Safari, enabling Apple to reduce its reliance on Google’s search engine.

3. MIT researchers create AI that teaches itself by generating training data

MIT researchers have unveiled SEAL (Self-Adapting Language Models), a groundbreaking framework that allows AI systems to evolve by generating their own training data and instructions for refining their internal mechanisms. This self-sustaining approach, powered by reinforcement learning, enables AI models to “self-edit” and continuously improve. In tests, SEAL outperformed even the latest GPT-4.1 model in knowledge integration and few-shot learning tasks. This pioneering system could revolutionize enterprise AI by enabling models to adapt dynamically, absorb company-specific knowledge, and personalize interactions. While challenges like "catastrophic forgetting" still need to be addressed, SEAL represents a major leap forward in autonomous, self-improving AI.

4. OpenAI cautions next-gen AI models may aid bio-weapon development by unskilled users

OpenAI warned that future models could enable “novice uplift” — letting people with little or no biology background create dangerous biological agents. While it doesn’t expect new bioweapons to be invented, it fears AI could help amateurs replicate known threats. OpenAI is stepping up testing and safeguards, especially for successors to its o3 model, which are expected to hit high-risk thresholds. The company says near-perfect detection and rapid response are essential to prevent misuse. It plans to work closely with U.S. national labs and host an event with nonprofits and researchers to address these growing risks collaboratively.

TRENDING TOOLS

• Pally: AI relationship management for all your pals, from all your socials.

• Monday: Chrome extension that blocks distracting sites, and to access them, you convince an AI that you're gonna use the site for something productive.

NEW LAUNCH

ElevenLabs’ new voice-first personal assistant that doesn’t just talk, it acts 🗣️🤖

ElevenLabs has released 11ai Alpha, a voice assistant designed to do more than just respond to commands. It can carry out tasks and help professionals manage their workflows.

Here’s what you need to know:

• 11ai goes beyond answering questions. It helps you complete tasks using voice commands across your productivity tools.

• The assistant connects with tools like Linear (Project Management), Slack (Team Communication), Perplexity (Real-time Research), and Notion (Knowledge Management) through Model Context Protocol (MCP).

• Real-world use cases: "Add my tasks to Linear and schedule time blocks in Google Calendar," "Research our prospect’s latest funding and update their CRM profile," and "Summarize yesterday’s engineering Slack threads and create follow-up tasks."

• It offers enterprise-grade performance with low latency (<500ms), multilingual support, and HIPAA-compliant security.

• You can choose from over 5,000 voices or create your own voice clone for more natural interactions.

Why it matters:

11ai makes voice assistants more useful by focusing on real tasks instead of just answering questions. It connects to tools you already use and helps you get things done faster. For enterprises, MCP allows for custom integrations and better control over permissions. This change shows where voice technology is evolving, taking on tasks and helping businesses run more smoothly.

THINK PIECES / BRAIN BOOST

• A Reddit user used this AI workflow to fully automate videos on an Instagram account and got ~5M views in 3 weeks.

• How people are using AI agents, MCP, and Claude to outcompete 99.9% of people using AI today.

• If you want to get into AI automation, a guy who's built 100+ workflows gives you the 10 things he wishes he knew before he started.

THE VALLEY GEMS

What’s trending on social today:

1/ Movie-making and content creation costs are set to plummet.

2/ This is one of the funniest humanoid robot videos I have seen this year.

Presented by Proton

Secure credentials. Easy admin control. No loose ends.

Proton Pass lets you onboard and offboard team members instantly. Grant or revoke access in one click, enforce password policies, and track activity — all from a simple dashboard. No training required. Trusted by 50,000+ businesses, Pass provides easy-to-use account security for fast-moving teams.

Simplify team security with Proton Pass

THAT’S ALL FOR TODAY

Thank you for reading today’s edition. That’s all for today’s issue.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee

👍️ New reader? Subscribe here

Thanks for being here.

REACH 100K+ READERS

Acquire new customers and drive revenue by partnering with us

Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.

If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.