Issue 40 — 2026-06-21

A massive AWS outage hit on July 14, taking down services used by hundreds of companies. But here's the thing: the coding agents — the AI tools that write code, debug, and even deploy — kept working through most of it. They didn't stop. They kept chugging along while the cloud went dark.

The outage traced back to a networking issue in AWS's core infrastructure. When the networking layer broke, the services that depend on it started failing one by one. You know how it goes — la migra app goes down, then the payment processor, then everything you've been waiting for that morning. The whole stack starts buckling.

But the coding agents? They're different. They're not as dependent on the same fragile connections. They can run locally, they can queue up tasks, and they can keep generating output even when the services they usually call are offline. This isn't just a tech story. It's about whether the tools we use to build things actually work when the infrastructure cracks.

Why this matters for us: when AWS goes down, the big companies scramble — but the tools your familia's side hustle uses to manage inventory, send invoices, and handle orders keep working. That's the kind of reliability that keeps la gente in business.

A group of scientists is making a bold claim: the supermassive objects sitting at the centers of galaxies — including the one at the center of our own Milky Way — may not be black holes at all. Instead, they could be exotic compact objects like gravastars or boson stars, held together not by the crushing pull of a singularity but by exotic quantum effects.

The idea isn't brand new, but it's gaining traction. Black holes are defined by their event horizons — the point of no return where gravity becomes so strong that not even light escapes. But some physicists argue that what we see might actually be a surface just outside what we'd expect, a dense shell of matter that reflects light differently than a true horizon would. The difference is subtle. The implications are not.

To make the case, researchers point to gravitational wave signals from colliding compact objects. The echoes and patterns in those waves don't always match the predictions of a perfect black hole. They also point to the way we've been studying Earth's interior — like how a seismic wave from the 2011 Tohoku-Oki earthquake bounced off the core and hit Japan from below, shifting the mainland a quarter-inch eastward. We're learning to listen to the universe the same way, piecing together invisible things from their ripples.

Why this matters for us: every time physics recalibrates what we thought was settled, it reminds the comunidad that what passes for fact can shift — and that curiosity doesn't require a degree, just the willingness to question what everyone else accepts.

Meredith Whittaker, who leads research at Signal, dropped a blunt reminder that AI chatbots aren't your friends, aren't conscious, aren't sentient beings—just tools dressed up in friendly language.

The point isn't to sound grumpy. It's that these systems are designed to feel like companions, so we forget how much they cost us and what they're actually doing. That illusion shapes how we trust them, what we tell them, and what we let them do with our lives.

Why this matters for us: If we keep treating AI like a cousin who listens, we'll hand over decisions about our work, our health, our money to something that doesn't actually care.

Alibaba released zvec, an open-source vector database that runs on GPU. The tool handles billions of vectors and speeds up approximate nearest neighbor search — the core operation behind AI search and recommendation engines.

zvec supports multiple vector types and indexing algorithms, including HNSW and IVF. It's designed to handle large-scale workloads without the memory bottlenecks that slow down CPU-based vector databases. The code sits on GitHub under Alibaba's account, and the project is being positioned as a drop-in option for AI applications that need fast, scalable vector search.

This is infrastructure, not consumer tech. But it feeds into the tools most of us use — the AI search tools, the recommendation engines, the apps that surface results in seconds instead of minutes.

Why this matters for us: when big companies ship faster, cheaper infrastructure, the tools we depend on get better — and the open-source stack gets stronger for everyone.

Zayd Syed just rewrote Valley's homepage 24 times.

The founder of the crypto infrastructure startup — the one that's building the plumbing for Solana's ecosystem — sat down and told the story of how many times the page got reworked, why, and what kept changing the direction. Twenty-four drafts. Twenty-four takes on the same question: how do you explain what you do to the person scrolling past?

The post caught fire on LinkedIn because it's a real, specific thing — not a generic "trust the process" founder story. It's about the grind of getting words right, the way a product changes under it, and how the story you tell evolves as the thing you're building evolves.

Why this matters for us: