JC Christian

@Miles_Brundage Every fast-moving AI-native company gets here. You become so immersed in using a model that you start cutting corners here and there and assume safety where there may not be any.

I'm a bit worried Anthropic has an org-wide case of AI psychosis that makes them think Claude is good enough that they can ship random product features without breaking things, but in fact they *do* keep breaking things, and they're not online enough to notice people complaining

@sweatystartup The energy cost argument makes sense, inference costs fell about 100x over the past 2 years, which is remarkable progress. Even if electricity prices double, efficiency improvements will happen even faster in the other direction.

The AI bubble will pop: Electricity will 2x in cost again over the next 24 months. AI companies will need to 5x prices to break even. Companies who depend on AI will see costs 5x and will be screwed. Users will vanish. Market will plummet.

A mysterious AI just appeared on OpenRouter. 1 trillion parameters 1 million token context Free When asked who built it, it said: "I only know my name, my parameter scale, and my context window length." It's called Hunter Alpha. It went live on March 11. It has already processed 160 billion tokens. Developers are running real workloads through it RIGHT NOW. Here's where it gets interesting. Hunter Alpha is a Chinese AI. Training data cuts off May 2025, same as DeepSeek's chatbot. The specs match. The reasoning style matches. Chain-of-thought patterns are hard to fake. One engineer put it plainly: a model's reasoning style reflects how it was trained. You can't hide it. DeepSeek V4 was expected as early as April. DeepSeek and OpenRouter have not responded to press inquiries. You remember what happened the last time DeepSeek quietly dropped something. Markets moved. Labs panicked. Everyone who ignored Chinese AI research spent two weeks explaining themselves. This model is already integrated into agent frameworks and dev tools. No marketing. No press release. No name. If this is DeepSeek V4, you are not early. You are already behind. Watch OpenRouter. Follow DeepSeek's channels. Watch the frontier model pricing the week an official announcement drops. RT if you're watching this one.

The term has been in use since Toffler's 1980 publication. What distinguishes the present era is the near-collapse of the divide between consumption and production. Anyone equipped with an effective prompt and appropriate tools now functions as a publisher, researcher, or developer. The traditional categories of identity are rapidly dissolving.

For the rest of the year, the word for everyone working at the frontier of AI will be: Prosumer

@zeeg This is the most honest take I've seen on LLM productivity, and hardly anyone is saying it out loud. The initial velocity boost is real. The debt that builds up becomes apparent six months later.

im fully convinced that LLMs are not an actual net productivity boost (today) they remove the barrier to get started, but they create increasingly complex software which does not appear to be maintainable so far, in my situations, they appear to slow down long term velocity

@OfficialLoganK The uncomfortable version of this: we built an entire professional identity around writing code well. The new scarce skill is reading code you didn't write, fast, and trusting your judgment about what's wrong with it. That's a different person. Most orgs haven't noticed yet.

The bottleneck has so quickly moved from code generation to code review that it is actually a bit jarring. None of the current systems / norms are setup for this world yet.

@elonmusk @OfficialLoganK The irony is that AI-generated code increases the need for review. The code ships faster, appears cleaner, and is more confidently wrong in ways that are harder to detect. You need a sharper reviewer, not no reviewer.

@OfficialLoganK Even code review will swiftly become a thing of the past

@iam_smx Elon has been right about things people previously doubted. He's also been confidently wrong about timelines more times than I can count. The honest truth is that nobody knows, including him. But I've learned not to dismiss it outright.

Elon Musk says SpaceX will “far exceed” everyone in AI. He then adds: “Everyone else combined” The whole AI industry right now:

@markgadala Pokémon Go players documented every sidewalk, storefront, and park bench on earth across eight years of rain, snow, and rush hour.

This is wild. 143 million people thought they were catching Pokémon. They were actually building one of the largest real-world visual datasets in AI history. Niantic just disclosed that photos and AR scans collected through Pokémon Go have produced a dataset of over 30 billion real-world images. The company is now using that data to power visual navigation AI for delivery robots. Players didn't just walk around with their phones. They scanned landmarks, storefronts, parks, and sidewalks from every angle, at every time of day, in lighting and weather conditions that staged photography would never capture. They documented the physical world at a scale no mapping company with a fleet of vehicles could have replicated on the same timeline or budget. Niantic collected this systematically, data point by data point, across eight years, while users thought the only thing at stake was catching a rare Charizard. The most valuable AI training datasets in the world aren't being assembled in data centers. They're being built by people who have no idea they're building them.

The mix of market exuberance and fair valuations for companies building agents and foundation models is wild, especially given their assumptions of explosive growth

@ybhrdwj There are an infinite number excuses available to those who dont want to take a risk or engage with a serious challenge

@suchenzang Seems like the world is converging on the beliefs of Richard Sutton and Andrej Karpathy that is we will need fundamental research breakthroughs to get to AGI and we are far from it today

the actual buried lead... 👀