This week we explore:
Google's compression breakthrough could make local AI models a reality
Apple turns Siri into an AI Switzerland
Meta builds a computational model of human perception
Smart glasses finally solve a real problem
Why Rec Room's death matters for the VR industry
Artificial Intelligence
Google's TurboQuant Could Make Consumer AI Hardware Actually Useful
If Google's new TurboQuant compression really works as advertised, we're looking at a fundamental shift in how AI gets deployed. Instead of paying OpenAI monthly fees and sending your data to the cloud, you could run GPT-4 level models on a basic Mac Mini sitting on your desk. That's the kind of change that reshapes entire industries overnight.
The algorithm makes large language models smaller and faster without losing quality, meaning that 16GB Mac Mini can suddenly run incredible AI models completely locally, free, and secure. This isn't just a technical achievement it's potentially the beginning of the end for cloud AI monopolies.
Apple Admits Defeat: Siri Becomes Gateway to Better AI Assistants
Apple is essentially white-flagging its AI assistant war by turning Siri into a hub for ChatGPT, Claude, and other third-party AI models in iOS 18. It's a brilliant pivot from trying to build the best AI to becoming the best AI platform. It lets users access superior assistants while keeping them locked in Apple's ecosystem.
You will be able to install any AI app and toggle it on in settings, effectively making Siri a wrapper for actually competent AI. After years of trying to make Siri compete and failing spectacularly, Apple chose the smart play: if you can't beat them, become the platform they all want to be on.
Meta's AI Can Now Predict Your Brain's Response to Audio and Video
This is potentially huge for understanding how humans actually process information and could revolutionize everything from entertainment to mental health treatment. Meta's TRIBE v2 can predict brain activity from audiovisual inputs using data from 700+ people, essentially creating a computational model of human perception.
If it works as advertised, we're looking at AI that doesn't just mimic human responses but actually predicts the neural mechanisms behind them. The implications for personalized content, therapeutic interventions, and human-computer interaction are staggering.
Google's Voice AI Could Actually Kill Call Centers This Time
Google's new voice API promises sub-second latency and 90+ language support, potentially automating away entire call center operations with just a Python script. While the hype around AI replacing human jobs has been overblown before, voice interfaces are finally hitting the quality threshold where small businesses might actually ditch human receptionists.
The real test isn't the tech specs it's whether a salon owner in Ohio can set this up without calling their nephew who 'knows computers.' If Google nails the user experience, this could be the first widespread AI job displacement we actually notice.
Microsoft's TRELLIS Makes 3D Asset Creation Stupidly Fast
This could completely flip how game studios and AR companies build 3D content. Instead of hiring armies of 3D artists or waiting weeks for asset creation, you just feed TRELLIS a single photo and get production-ready 3D models in seconds.
The fact that Microsoft open-sourced this 4B parameter model means every indie developer now has Hollywood-level 3D generation tools. This is the kind of democratization that creates entirely new categories of creators and destroys traditional creative industry gatekeepers.
OpenAI and Anthropic Just Accidentally Created the Perfect AI Workflow
This is huge for developers who've been bouncing between AI coding tools like ping-pong balls. OpenAI built a plugin that lets Claude users tap into Codex for code review and testing without leaving their workspace. It's the first real sign that AI companies might actually collaborate instead of just building walled gardens.
You can now build with Claude and have Codex double-check everything without switching apps. If this trend continues, we might actually get AI tools that work together instead of forcing you to pick sides in some corporate AI war.
Quick Hits
Anthropic's Alleged Source Code Leak Shows AI Security Theater
A viral thread claimed Anthropic accidentally leaked Claude's entire source code, but there's zero credible evidence this happened. The story reads like AI misinformation designed to go viral, yet thousands shared it without verification revealing how easily fake tech drama spreads in our hyper-connected world.
Chamath Says Only Musk Builds World-Changing Infrastructure
Venture capitalist Chamath Palihapitiya delivered a brutal reality check: while Elon Musk revolutionized rockets, satellites, and electric vehicles, the rest of tech has been busy perfecting wireless earbuds and 15-second dance videos. It cuts to the core of whether today's tech giants solve humanity's biggest problems or just optimize ad revenue.
The ops hire that onboards in 30 seconds.
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.
Spatial Computing
Apple Vision Pro Gets Its First Speed Dating App
AuraTap brings the brutal efficiency of modern dating to spatial computing with 2-minute face-to-face encounters in VR. It's essentially Chatroulette meets Bumble, but now you're wearing a $3,500 headset to awkwardly make small talk with strangers.
The real test isn't whether people will use it, but whether Apple's walled garden approach can handle the inevitable chaos of unmoderated human interaction. Dating apps are messy, unpredictable, and full of people behaving badly which is exactly the opposite of Apple's controlled experience philosophy.
Google's AI Can Now Code VR Apps From Text Prompts
This is the breakthrough that makes spatial computing accessible to everyone, not just Unity wizards with CS degrees. Google Research just dropped Vibe Coding XR, which lets Gemini AI generate full WebXR applications from simple text descriptions and is complete with physics and interactivity.
Suddenly, creating VR experiences becomes as easy as chatting with an AI assistant. This could unleash a wave of spatial computing content from creators who never would have touched a 3D engine otherwise.
One of VR's biggest social platforms is calling it quits because they never cracked the code on making virtual hangouts profitable. Rec Room's shutdown is a brutal reminder that building cool VR experiences and building sustainable businesses are completely different challenges.
The VR market isn't mature enough to support even well-funded pioneers. Until headset adoption hits smartphone-scale numbers, VR social platforms will keep running into the same wall: not enough users willing to pay for virtual experiences.
Smart Glasses Finally Crack Real-Time Accessibility with Live Captions
This is the kind of breakthrough that makes AR glasses actually useful for millions of people right now. Real-time live captions on smart glasses could be transformative for deaf and hard-of-hearing users, turning what's often been a gimmicky tech category into genuine assistive technology.
It's a reminder that the most compelling AR applications might be about enhancing human capabilities rather than overlaying digital nonsense. Sometimes the best tech is invisible. It just makes life work better for people who need it most.
Our Vision
This week's stories reveal a tech industry finally hitting its stride after years of promising more than it could deliver. Google's TurboQuant and Microsoft's TRELLIS represent the kind of fundamental breakthroughs that actually change how work gets done, not just how it gets marketed. Meanwhile, Apple's decision to turn Siri into an AI platform shows even the most stubborn companies are learning to play well with others when survival is on the line.
But the contrast between AI's accelerating capabilities and VR's continued struggles tells a deeper story about technology adoption. AI is succeeding because it solves immediate, obvious problems like writing better code, generating content, automating boring tasks. VR struggles to find its way because it's still looking for problems that justify wearing a computer on your face. Rec Room's death isn't just about one platform failing; it's about an entire category that hasn't found its sustainable business model yet.
The real winners this week are the developers and creators who suddenly have access to tools that were impossible just months ago. When you can generate production-ready 3D models from photos, run GPT-4 locally on a Mac Mini, and code VR apps with natural language prompts, the barriers between having an idea and shipping a product basically disappear. That's the kind of democratization that creates the next wave of innovation.
The question isn't whether AI will change everything anymore. It's whether we're building the right foundations for humans to thrive alongside these new capabilities, rather than just replacing them entirely.
How did you like this week's edition?
Are you tracking agent views on your docs?
AI agents already outnumber human visitors to your docs — now you can track them.
Private Credit on Your Terms
Percent's secondary marketplace lets accredited investors buy into eligible deals or indicate interest in selling existing positions. Secondary market access in private credit is still rare. 16.72% current weighted average coupon. Terms start at 3 months. New investors can receive up to $500 credit.
Alternative investments are speculative. Secondary liquidity not guaranteed. Past performance not indicative. Terms apply.





