Better prompts. Better AI output.

AI gets smarter when your input is complete. Wispr Flow helps you think out loud and capture full context by voice, then turns that speech into a clean, structured prompt you can paste into ChatGPT, Claude, or any assistant. No more chopping up thoughts into typed paragraphs. Preserve constraints, examples, edge cases, and tone by speaking them once. The result is faster iteration, more precise outputs, and less time re-prompting. Try Wispr Flow for AI or see a 30-second demo.

Start flowing free

Hey, quick AI catch-up today.

This one feels like the industry is trying to answer the same question from three angles: Where does the assistant live? and what should it be?

Today:

• DeepMind and Anthropic CEOs on Life "The Day After AGI"

• Claude Gets a Massive "Moral" Update

• Apple is Secretly Building an AI Pin

• OpenAI’s Mystery Earbuds Are on Track for 2026

• Siri Gets a Brain Transplant to Fight ChatGPT

• OpenAI Wants to Put AI in Classrooms Worldwide

the day after AGI…

In a rare, candid interview titled The Day After AGI, AI leaders Demis Hassabis (DeepMind) and Dario Amodei (Anthropic) explore the fast-approaching future of artificial general intelligence. They warn that exponential progress could outpace our ability to adapt, especially in jobs, economics, and governance. Both affirm AGI is likely within the decade—Demis estimates a 50% chance. Coding and other high-skill tasks are already being automated, and junior roles may vanish first. 

While they reject doomerism, they stress urgent focus on safety, policy, and meaning in a post-work world. The upside: curing disease, scientific breakthroughs, and abundant resources—if we proceed wisely.

WATCH THE VIDEO ON YOUTUBE

Apple is (reportedly) building an AI “pin” the size of an AirTag

The headline rumor: Apple is working on an AirTag-sized, circular wearable “pin” that uses cameras + microphones to understand what’s around you.

What’s being reported about the device itself:

• Thin, flat, circular body (aluminum + glass), roughly AirTag-sized

• Two cameras (standard + wide-angle), three mics, speaker, a physical button, and wireless charging

• Still early; could land as soon as 2027 (and could still be canceled)

Why this one matter:

Apple doesn’t need to invent a new category from scratch — it just needs to make the “AI helper in your day” idea feel normal. The last wave of standalone AI pins didn’t exactly inspire confidence (RIP Humane AI Pin), so Apple entering this lane is a real signal that wearables are the next battleground.

OpenAI says its first device is “on track” for 2026 — and the earbud rumors are getting louder

OpenAI’s Chris Lehane said the company is “on track” to unveil its first device in the second half of 2026 — but he wouldn’t say whether it’s a pin, earpiece, or something else.

Meanwhile, reporting/leaks swirling around the same time suggest it could be earbuds (codename “Sweet Pea”), potentially with:

• A custom 2nm chip and more on-device processing (less cloud dependence)

• Very aggressive shipment talk (one report mentions 40–50 million units in year one — take that as “ambitious target,” not destiny)

My read: if OpenAI really goes “earbuds-first,” that’s a bet that the winning interface is always-there audio + context, not a screen. But the details are still foggy — and after the Humane story, everyone’s going to be skeptical until there’s a real demo.

Anthropic published “Claude’s new constitution” — and released it under CC0

Anthropic published a new constitution for Claude: a detailed description of the values and behavior they’re trying to train into the model written primarily for Claude itself (as guidance the model can learn from).

A few things that stood out:

• They say the constitution is central to training and directly shapes behavior, even if real outputs don’t always perfectly match the ideals.

• They shifted from a list of principles to explaining the “why” behind behaviors, aiming for better generalization and judgment in novel situations.

• They publish it CC0 (public domain equivalent) so anyone can reuse it. That’s a big transparency move.

• Their priority order is explicit: broadly safe → broadly ethical → compliant with Anthropic guidelines → genuinely helpful.

Why this one matter: when an AI system starts to feel like it’s “with you” all day (earbuds/pin/glasses), the question of what it’s optimizing for stops being academic. So I actually love seeing the “values spec” become something the public can read, critique, and compare across labs.

🧠RESEARCH

OmniTransfer: All-in-one Framework for Spatio-temporal Video Transfer

This paper introduces OmniTransfer, a tool for editing videos. It allows users to modify a video's visual style or movement—like swapping faces or adding effects—within a single system. By intelligently separating "looks" from motion, it creates smoother, more realistic results without needing complex manual instructions, outperforming older video editing methods.

UniX: Unifying Autoregression and Diffusion for Chest X-Ray Understanding and Generation

UniX is a new AI for chest X-rays that handles two jobs: diagnosing health issues and creating realistic X-ray images. It uses two specialized parts working together—one for reading and one for drawing. This team approach improves diagnosis and image quality significantly while using less computer memory than similar models.

Advances and Frontiers of LLM-based Issue Resolution in Software Engineering: A Comprehensive Survey

This paper reviews how advanced AI programs are being used to fix computer software bugs automatically. It looks at different methods for teaching AI to solve coding problems, from basic prompts to complex training. The authors analyze current success rates on test comparisons and suggest ways to build better autonomous coding assistants.

🛠️TOP TOOLS

Each listing includes a hands-on tutorial so you can get started right away, whether you’re a beginner or a pro.

Celebrity AI – Create AI Videos, Images & Songs in Seconds - multi‑modal creator tool that lets you generate short AI videos, AI images, and full songs with lyrics and vocals—inside one interface.

Census AI – Ask U.S. Census Data in Plain English - natural‑language interface to U.S. census and demographic data.

ChaiGPT – Social AI Chat & Role‑Play Companions - conversational AI where you chat with AI characters or create your own.

📲SOCIAL MEDIA

🗞️MORE NEWS

Apple’s Siri Overhaul Apple is upgrading Siri into a smarter chatbot named "Campos" to rival OpenAI. Arriving in 2026, this update embeds conversational AI deep into iPhones and Macs, letting your devices understand and act on complex requests.

OpenAI’s Global Education Plan OpenAI is partnering with nations to integrate AI into classrooms. This initiative aims to modernize school systems, helping students learn faster and preparing workforces for a future where working with smart computers is essential.

Amazon’s Health Assistant Amazon’s One Medical now offers a smart health helper in its app. This AI explains complex lab results, schedules visits, and manages prescriptions, giving patients 24/7 guidance without waiting for a human doctor.

Meta’s Glasses for Good Meta is funding researchers and groups who use smart glasses for social good. These grants encourage creating helpful apps for their wearable tech, focusing on how AI on your face can solve real-world problems.

Microsoft’s Robot Brains Microsoft created "Rho-alpha," a new brain for robots. It combines vision and language to help machines understand physical tasks, aiming to make robots that can adapt to messy, real-world environments just like people do.

Stargate Community Fund OpenAI will pay for its own power upgrades to support its massive "Stargate" data center. The company promises to fund new energy grids so their huge electricity use doesn’t spike bills for local residents.