AI's Getting Smarter, But Are We?

Scott Alexander — the Substack writer behind Astral Codex Total — just dropped a long-form essay collecting his AI opinions. TLDR summarizes it as a big-picture read on where AI is headed, what he thinks will happen, and why it matters. This is the kind of thing that sticks around and gets passed along by people who actually use it.

Alexander's core move in the piece is straightforward: he lays out his actual positions on AI capability, timeline, and the people and companies shaping the space — not just the usual hot takes. He's not writing from Silicon Valley; he's writing from a place of reading the evidence and being honest about what he doesn't know. The piece covers his views on how fast AI is moving, which models and companies matter most, and what he thinks the big bets are. It's the kind of summary that saves you from reading the full essay yourself.

For la gente who want to understand AI without the jargon and the noise, this is a good read. It's not another VC pitch deck in essay form. It's someone laying out what they actually think, with reasons, and leaving room for being wrong.

Why this matters for us: as AI reshapes work and tech, we need voices that cut through the hype so la gente can understand what's actually coming.

NASA's Deep Space Network — a global system of giant antennas that talk to spacecraft across the solar system — just passed its first real test on Artemis II. The network, which handles everything from Mars rovers to deep-space probes, had been showing its age. Engineers nearly lost the signal during testing, but once the crew launched, the DSN held up.

The story here isn't that NASA's space program is great. It's that the infrastructure actually works this time. The DSN has been the backbone of American deep-space exploration for decades, and it's starting to show wear. Artemis II — the mission that sent astronauts around the Moon for the first time since Apollo — proved that the system can still handle the load.

Why this matters for us: When government infrastructure holds up, it means the money we pay in taxes is actually doing something — and it means the jobs in engineering, manufacturing, and tech stay in the country instead of getting outsourced or lost to foreign competition.

TLDR Product's Kasper Jung is asking whether product management systems should lean on AI agents to do the coding work. The piece walks through the current state of LLM evaluation and benchmarking — how we're measuring whether these models can actually deliver, not just churn out code that looks right.

The question isn't theoretical. Product management has always been about choosing what to build and how to know it's worth building. Now the tools are adding agents that can write code, test it, and iterate. The real test is whether those agents can be trusted with the kind of judgment that separates shipping something good from shipping something that works.

Why this matters for us:
When the tools that run the companies we work for start coding with agents, the people who understand the work — not just the people who understand the tools — get to decide what actually ships.

Waymo rolled out a premier subscription tier for $29.99 a month, letting riders use the robotaxi service as many times as they want without paying per ride. The launch gives the company a way to lock in frequent riders—taxi workers, parents running errands, folks who can't drive themselves—into a predictable monthly bill instead of a ride-by-ride tab.

This is a real shift for a company that started by building out driverless cars and now wants to be part of the daily commute. The subscription model mirrors what Uber and Lyft have been pushing for years, but with a key difference: no surge pricing, no surge anxiety. You tap in and go.

The move fits into the wider story of AI moving from the boardroom to the street. AI has been hyped as the thing that replaces workers, but as Normal Tech noted this week, the engineers who build and maintain these systems are still very much needed—just doing different work. The same applies to autonomous driving: the cars are learning, but the infrastructure, the routes, the safety layers, all of it still needs people to keep it running.

Why this matters for us: La migra app of ride services is becoming a $30/month monthly habit, which means the folks who actually use these services every day are the ones who'll feel the squeeze when prices shift.

Zed, the editor everyone's been talking about, has a new project: DeltaDB. The team is positioning it as a database built for metrics and logs, the kind of data that piles up fast when you're running services at scale.

The pitch is straightforward — Zed's team has seen how metrics and logs are handled today, and they've decided to build a better way. The name DeltaDB suggests they're leaning into incremental updates, which matters when you're tracking what's happening across dozens or hundreds of services.

Why this matters for us: when the tools we use to build and ship get better, the people who depend on them — the engineers, the ops teams, the small shops running lean — get a leg up too.

A patient in the UK has received the first human dose of a cellular reprogramming therapy that aims to turn back the clock inside your cells. The treatment works by introducing Yamanaka factors — four proteins that reset cells to a younger state — into the bloodstream. It's called a "high-risk" therapy because the science is still new and the stakes are real: get it wrong and you could end up with cells that have lost their identity, not just their age.

The therapy is being tested on age-related conditions, with early data suggesting it can reduce cellular age markers in people with diabetes. The world-first patient has already received their dose, which marks the beginning of a human trial that has been years in the making. If it holds up, this could be the first therapy to target aging itself rather than the diseases that come with it.

This isn't something for Silicon Valley alone — la gente in the comunidad are the ones who carry the weight of aging without the luxury of a wellness retreat. When a therapy like this works, it buys time for families, for abuelos who want to stay in their homes, for the ones who have been told to "just accept getting older."

Why this matters for us: if this therapy scales, it could mean more of our elders stay active and independent — not just surviving, but living — which is a win for every family that's had to watch someone they love fade too early.