Issue 29 — 2026-06-10

OpenAI has submitted a confidential filing for an initial public offering, according to a CNBC report. The company has been keeping a low profile through the process, which means the market hasn't had a chance to price in the deal yet. That changes now.

The filing kicks off what's likely to be one of the biggest IPOs in recent memory. OpenAI built on the back of the generative AI boom — ChatGPT, its flagship product, became a household name almost overnight — and has spent heavily on compute, talent, and data. Wall Street has been watching this moment for a while, and the confidential filing is the signal that the window is opening.

Why this matters for us: if OpenAI goes public and its valuation runs hot, the ripple effects hit everything from how AI tools get priced to who gets to own the infrastructure behind them.

Apple announced a new AI architecture that reworks how its devices handle machine learning — shifting processing from the cloud to the device itself. The company is consolidating AI workloads across its silicon, letting the iPhone, Mac, and Apple Watch run more models locally without waiting for servers.

The move is a direct response to the AI arms race. While competitors stack up cloud compute and train bigger models, Apple is betting on efficiency. The architecture packs neural processing into existing chips rather than forcing consumers to upgrade hardware. That matters because it means older devices get smarter without a new purchase — and it reduces Apple's dependency on expensive cloud infrastructure.

This is the kind of engineering play that tends to get overlooked. While everyone's talking about AI models, Apple is quietly reworking the plumbing. The architecture doesn't promise to beat OpenAI in benchmarks. It promises to make AI feel normal on the devices we already own.

Why this matters for us: Apple's push to run AI on the device means less reliance on big tech servers, which is good news for families, small businesses, and anyone tired of paying monthly subscriptions for features that should come free.

Google Search was supposed to be dead. OpenAI's ChatGPT launched with a lot of noise — it was going to replace search, swallow the internet, eat lunch. Instead, Google is still the default for billions of people.

The real story isn't that ChatGPT flopped. It's that Google figured out how to integrate AI into search without breaking what works. Users still type into a box. They still click links. They still get what they need without having to retrain their brains.

This matters because the big tech predictions always sound more dramatic than they are. The companies that last aren't the ones with the flashiest demos. They're the ones that keep doing the thing people actually use.

Why this matters for us: when Silicon Valley says "search is dead," la gente still types into Google and gets what they need — the hype dies faster than the habit.

Apple dropped Core AI documentation, opening up the same machine learning framework that powers Siri, Face ID, and the rest of Apple Intelligence. The move means developers can now build on the same models Apple uses in its own devices.

This isn't Apple's first shot at developer tools — the company has been quietly building out ML infrastructure for years. What's different now is that the framework is fully documented and accessible, not locked inside Apple's ecosystem. Developers can train models, run inference, and deploy apps using Core AI without needing to be an Apple developer.

For the comunidad, this matters because Apple has historically kept its best tools for its own devices and services. Now that Core AI is open, smaller developers — including many from Brown and Black communities — can build AI-powered apps without needing a Silicon Valley startup budget. The same models that power your iPhone now power your app.

Why this matters for us: Apple opening its AI framework means more tools for developers who don't need Apple's permission to build.

William James, the 19th-century philosopher who also co-founded American psychology, put forward one of the most counterintuitive ideas in the field: emotion doesn't follow thought. It drives it. The body feels first, and the mind catches up.

James and his brother Henry, both studying the nervous system, noticed that people's emotional experiences weren't neatly separated from their physical sensations. When we laugh, our body is already laughing; when we feel afraid, our body is already bracing. The feeling comes through the flesh before the thought lands.

This matters today because we keep trying to outthink our way out of problems — more analysis, more planning, more data. James was saying 130 years ago that we should pay attention to the body and the gut. The feeling is not the enemy of clear thinking. It's the signal.

Why this matters for us: When the body feels something before the mind knows it, we have to trust what la gente already know — that the gut is part of the calculation, not a distraction from it.

Apple has been quietly building out its Core AI framework — the machine learning toolkit that powers everything from Siri's understanding to the camera app's portrait mode. Now, developers are using it to build smarter apps directly on the device, without shipping data to distant servers.

The framework handles on-device inference, which means apps run faster, work offline, and respect privacy. For the thousands of engineers in our communities building products — the dev who just moved to Austin, the auntie's son coding from his mom's garage in Pico-Union — Core AI is the foundation for apps that actually work when the WiFi drops.

Apple's documentation page now covers the full stack: Core ML models, Core Vision for image processing, Core Speech for voice, and Core Location for spatial data. All of it designed to run on the chips Apple already put in your hands.

Why this matters for us: every time Apple ships AI that works on-device, it means less dependency on Silicon Valley's cloud monopolies, and more power for the developers in our communities who are building apps for la gente.

Mozilla is flipping the script on how AI gets trained. Instead of feeding models one massive dataset and calling it a day, they're building ground truth as an ongoing process — something that evolves as data gets used, questioned, and corrected.

The idea is simple but rarely practiced: datasets rot. They get stale, biased, or just wrong. Treating ground truth as a living system means catching errors, updating labels, and keeping the data honest over time. It's the difference between building a house once and maintaining it.

Why this matters for us: if AI keeps eating stale data, it keeps serving us stale answers — especially about the communities and languages we rely on.