21 Aug 2025

AI Glasses Hear All; New Manhattan Project?; NFL's AI Handoff

🇺🇸 America's AI "Manhattan Project" is Here

Get ready for a major leap in AI development... The U.S. National Science Foundation (NSF) and NVIDIA are pouring a hearty $152 million into the Open Multimodal AI Infrastructure (OMAI). The project, led by the Allen Institute for AI (Ai2), is being dubbed "America's Manhattan Project for AI."

What it is: A national AI core facility to build open multimodal AI systems for science that anyone can inspect and reuse.
Why it matters: This move democratizes access to frontier-grade AI, challenging the dominance of closed-off Big Tech stacks. It's a game-changer for researchers and could reshape enterprise IT budgets and compliance strategies.

Read more: AI Secret, The Rundown AI

🏈 NFL Goes All In on AI

The National Football League is becoming a high-tech lab. The NFL and Microsoft have expanded their partnership, deploying 2,500 Surface Copilot+ PCs with on-device AI processing across all NFL teams.

What's New: Coaches get real-time formation analysis, operations managers get AI dashboards and even marketing teams are testing Copilot.
The Big Picture: This goes beyond tablets. The NFL is rewriting its entire decision-making process. Coaches will offload film breakdowns to AI, analysts get machine-parsed player patterns, and front offices might start letting AI weigh in on draft picks.
Why It Matters: The NFL's $18 billion machine is getting a complete operating system upgrade. Legacy play-calling and gut-feel scouting are about to lose their competitive edge.

Read more: The Rundown AI, The Neuron

📱 Google's Pixel 10 is an AI Powerhouse

Pssst, Apple: Wake up! Google just dropped the Pixel 10 lineup, and it's packed with over 20 new AI features ready to roll, today. Powered by the new Tensor G5 chip, the phone is a beast, boasting 60% faster AI processing and a 4-billion parameter Gemini Nano model running entirely on-device for max privacy.

Killer Features:
- Conversational Photo Editing: Edit your pics with natural language. Yes, the rumored "nano-banana" model is here!
- Magic Cue: A proactive AI assistant that surfaces context across your apps to suggest replies and actions.
- Live Translate: Real-time call translation in 10 languages that preserves the speaker's actual voice.
Why it matters: Google is leaving Apple in the dust when it comes to on-device AI. This could be a major win in the ongoing smartphone wars.

Read more: The Rundown AI, The Neuron

👓 "Always On" AI Glasses That Witness Everything

Here's a story that's got privacy advocates up to their lenses with concern. Two Harvard dropouts, AnhPhu Nguyen and Caine Ardayfio, are launching Halo X, a pair of "always-on" AI smart glasses that listen to, record, and transcribe every single conversation around you. Think of it as having infinite memory. You may recall their viral video of the I-XRAY prototype.

There are many pendants in market today (Limitless, Plaud, etc.) that already record and transcribe. Glasses add a new form to the mix.

What Makes Halo Different

Always Listening: These glasses record every word you hear, and maybe some you miss, transcribe it in real-time, then delete the audio file (they claim).
Contextual AI: If someone mentions something complex, explanations pop up on your glasses display instantly.
Memory Bank: The AI can recall and summarize past conversations, essentially giving you a photographic memory for everything you've heard.
Stealth Mode: Designed to look like normal glasses – no one may know you're recording them.

The Business Side

Funding: Just raised $1 million led by Pillar VC
Price Point: $249 for preorders starting today
Market Play: Positioning as a privacy-focused alternative to Meta's smart glasses

The Privacy Problem

This isn't just about cool tech, it's about the future of privacy in public spaces. Eva Galperin from the Electronic Frontier Foundation warns that normalizing "always-on recording devices" erodes our expectation of privacy in conversations.

Legal Issues: Several U.S. states require consent from all parties before recording. The founders' response? "We trust our users to get consent." That's... not exactly reassuring.

Past Controversies: The I-XRAY glasses noted above, were a facial recognition system for Meta Ray-Ban glasses that could identify and dox random strangers without consent. They never released the code, but the demo was unsettling to some.

Why It Matters

We're at a crossroads where broad, AI-powered surveillance is going beyond the fixed Ring camera and becoming consumer-friendly, mobile and affordable. While Halo promises to help users "cheat" at everything (remember Cluely?) from job interviews to school exams, it also represents a fundamental shift toward a world where every conversation might be recorded and analyzed.

The question isn't whether the technology is impressive (it is) – it's whether we're ready for a society where privacy in public spaces becomes obsolete.

Read more: Techcrunch

🥩 Tyson Foods Gets a Brain

Think the meat industry is all brawn, no brains? Wrong (And we're not talking about brains for your palate.) Tyson Foods, the giant behind 20% of America's meat, just gave its B2B arm a major AI upgrade. Using AWS Bedrock and Claude 3.5, Tyson has rolled out a conversational search assistant that's shifting how chefs and operators interact with their products.

The Gist: No more clunky catalogs. Chefs can now find recipes, distributors, and promos with human-like precision.
The Real Impact: This isn't just a fancy search bar. Tyson is creating a data-native empire, gathering direct intel from over a million operators for the first time. It's a live AI feedback loop.

Read more: AI Secret

📈 Market Buzz & The AI Winter Debate

Is the AI hype train about to derail? The market got a little jittery this week, with the S&P 500 and Nasdaq taking a dip after an MIT analysis claimed that 95% of generative AI pilots aren't moving the needle on profits. This, coupled with Meta downsizing its AI department and Sam Altman himself calling AI a "bubble," has sparked fears of an impending "AI Winter."

The Big Question: Can AI deliver on its sky-high valuations? It all comes down to "sample efficiency" - how fast AI can learn new skills. If it can't learn fast enough, the bubble could burst. But if it can, we're just getting started.
The Counterpoint: AI infrastructure is currently propping up the US economy, adding an estimated $93-163 billion this year. Without it, growth would be nearly flat.

Read more: The Neuron, TLDR AI

✨ Freshly Minted: AI Tools to Watch

For developers:

Warestack: Create release protection rules in plain English to automatically flag or block code changes.
ChartDB v2: Instantly create an editable ER diagram from your database schema with a single query.

For the Creatives:

Kira.art: A creative AI agent that lets you generate and edit images through conversation – no prompts or layers required!
Luma AI: Transforms your text into videos and lets you restyle any video's background, characters, or setting in post-production.

For Everyone Else:

Obsidian "Bases": Turn any set of notes into a powerful database with filtering and formulas.
VoiceType: Write up to 9x faster using your voice instead of the keyboard.

Read more: AI Secret, The Neuron

💻 Quick Hits

Surya, the Sun-Staring AI: NASA and IBM have launched Surya, an AI model that can predict solar flares with 16% greater accuracy, giving us an extra hour of warning time to protect our satellites and power grids.
GPT-5 Gets a Personality Makeover: OpenAI has rolled out a friendlier, "warmer" version of GPT-5 after users complained it was too formal. Don't worry, internal testing shows it's not just being a suck-up.
Claude Can Hang Up on You: Anthropic's Claude Opus 4 and 4.1 can now end conversations as a last resort against abusive interactions. It's an interesting exploration into AI welfare and moral status.
DeepSeek V3.1 is Here: This model update brings improved intelligence and a larger context window.

Read more: TLDR AI, The Rundown AI, The Neuron