Three trillion parameters. 896 expert networks. A 17-place jump to number one in front-end coding overnight. I want to slow down on Kimi K3 for a minute, because the headline numbers are getting all the attention and the actual architecture story is more interesting than the leaderboard. Start with the benchmark reality. Kimi K3, built by Beijing-based Moonshot AI, is now sitting at number one in Arena's front-end coding rankings. First place across six out of seven domains —

I'm writing this by talking... Hands crossed. Sitting back. Just saying words out loud into a console that's going to turn them into an article, generate a cover image, produce audio, post it to my site, push it to LinkedIn. The whole thing. I'm not touching a keyboard. Last week I wrote a Mac app in twenty minutes. It connects over Bluetooth to my Ember mug and shows me the temperature in my menu bar. Not my mug. I didn't build the mug. Didn't invent the Bluetooth stack. I j

Everyone has been staring at the benchmark leaderboards. Meanwhile, the real fight moved somewhere else entirely. Apple just filed a 41-page lawsuit against OpenAI. Not over AI. Over manufacturing. The specific targets are Tang Tan — OpenAI's hardware chief, formerly Apple's VP of product design — and a former Apple engineer named Chang Lu. Apple's allegation is a coordinated extraction campaign: employees being actively pushed to download confidential files on their way out

I want to say upfront that "I built" is doing a lot of heavy lifting in that sentence. What I actually did was look at my coffee mug, have a thought, type that thought into Glaze, and go do other stuff. About fifteen or twenty minutes later, I had a finished, working macOS application sitting in my menu bar. It connected to my Ember Mug over Bluetooth, showed me the temperature, and sent me a push notification when the battery got low. I'm not sure "I built" is even the right

I want to be clear about something upfront: I claim zero talent or skill in what I'm about to describe. This is a broken clock story. The clock just happened to be right at a useful moment. Labor Day weekend 2024. Raining in South Florida. Bored enough to do something stupid. I grabbed a Xiao ESP32-C3 — tiny little microcontroller, about $5, probably the smallest ESP32 package I've ever seen — and eight buttons. Wired four of them to standard shortcuts: copy, paste, cut, ente

Let me explain what happened... I found a vibe-coding tool for Mac called Glaze. I had 120 free tokens....I had a mildly annoying workflow problem involving prompts and clipboard management and too many browser tabs and about eight thousand open Notes documents I never look at. I told the AI to build me a small floating window. That’s where things got away from me. Here’s How the Evening Actually Went The original ask was pretty simple. Make me a floating sidecar window that

The Receipt OpenClaw — one of the most widely used open-source agent frameworks in the world — just shipped a feature called Dreaming. By default, it runs at 3:00 AM. Every night. Whether you’re awake or not. What it does, technically, is background memory consolidation: three phases (Light, REM, Deep) that sort recent short-term signals, reflect on themes, and promote the most durable insights into long-term memory. The system scores candidates on relevance, frequency, recen

MIT's State of AI in Business report found that 95% of enterprise generative AI pilots fail to deliver measurable ROI. Not “underperform.” Fail — no return the company can point to. Separately, Gartner surveyed companies on their AI-driven workforce reductions and found that 80% had cut headcount, with zero correlation between the size of the cut and any resulting improvement in ROI. Cutting people didn't produce better numbers. It just produced fewer people. Then there's the

Ben Broca launched Polsia in late 2025. One founder. No employees. A system of AI agents that plans, codes, markets, and runs companies around the clock. Within a month, he crossed $1 million in annualized revenue with over 1,000 companies running on the platform. A month after that, $3 million ARR, 3,000 companies. Last count, something like $10 million run rate, thousands of businesses on the platform, and a $30 million Series A at a $250 million post-money valuation — rais

Late Tuesday night, Sam Altman posted two sentences on X: “GPT-5.6 Sol, along with Terra and Luna, will launch publicly this Thursday. We're expanding preview access globally now.” By morning it was a standard tech news cycle — pricing tables, benchmark charts, the usual “which tier should you use” explainers. Sol at $5 input / $30 output per million tokens, Terra at $2.50/$15, Luna at $1/$6. Sol is OpenAI's strongest model yet, meaningfully better at coding, biology, and cyb

Anthropic published a paper a few hours before this piece went out, and the internet already knows what it's going to do with it: turn “Claude has a mechanism for conscious access” into “Anthropic thinks Claude is conscious.” That headline will spread fast. It's also not what the paper says, and the actual finding is more useful than the sci-fi version. The Receipt Anthropic's researchers built a tool they call a JSpace lens — JSpace, short for Jacobian space, the mathematica

I filed a provisional patent in November 2024 for something I called "System and Method for Digital Business Cards." The idea was simple: tap a card, get someone's contact info on your phone instantly, no app required. I built it, called it Tap2Space, put it on Wix, and ran it as a real business for about two years. Over 100 real clients, custom-designed cards, a physical NFC card and keychain fob included. It worked. People paid for it. Some of them are still using the cards

A video's been making the rounds this week breaking down a trio of new SpaceX-adjacent trademarks — Starmind, Stargaze, Starfall — and connecting them to a real FCC filing: SpaceX asking permission to launch up to one million satellites as an orbital data center constellation. The video's framing is clean and mostly right: Starlink becomes the communications layer, the new constellation becomes the compute layer, laser links between satellites become the internal data fabric.

It's Sunday, July 5th, the day after the 4th, and I have exactly until midnight to burn through almost a month's worth of vibe-coding tokens before they reset and vanish into the ether, because apparently rollover credits are a courtesy the phone companies figured out decades ago and half the SaaS world still hasn't caught up to. I'm writing this fast. I have things to build. Bear with me. The Setup One of my accounts runs on a use-it-or-lose-it token model. No rollover. You

I watched a clip on TikTok this week of a woman explaining her business to a host, and I need to talk about it, because it's one of the best real-world examples I've seen of exactly what I've been saying since 2023 — and it's also proof of something most people building with AI right now don't want to hear. What She Actually Built She sells custom GPTs — the ones you build on a standard $20/month ChatGPT account — to coaches, founders, and consultants. She's doing around $100

I want to show you something my team and I just shipped, because I think it's the clearest proof yet of what I've been building toward with ARIA. Over the past week we launched four media properties: The Daily Quantum The Daily MicroCap The Daily Biologics Petides Direct Four completely different brands. Four different audiences — quantum computing professionals, micro-cap investors, biotech operators, and pet owners researching peptide therapy. Look at them side by side and

The stat making the rounds is real: Meta employees consumed 73.7 trillion AI tokens in about 30 days, which works out to roughly $221 million a month and $2.65 billion a year at list price. That's not exaggerated. It's in an internal memo, first reported by The Information, sent to roughly 6,000 employees warning that internal AI usage costs are approaching billions of dollars in 2026. But the number by itself misses the actual story, and the actual story is better than the n

Fourteen months ago I wrote a piece called "Why We NEED a Billion Robots." The thesis was simple: the AI conversation was obsessed with the wrong half of the workforce. Everyone was watching chatbots write emails while ignoring the fact that nobody had figured out how to automate a plumber, an electrician, or a warehouse picker. I argued that solving machine intelligence was the easy part — manufacturing that intelligence into a body that could actually do physical work was t

While everyone was watching Washington argue about which labs get export licenses and who controls access to frontier AI, something genuinely significant slipped out with almost no fanfare. An open source model just learned to build its own brain. It's called Ornith-1.0. It was built by a US company called DeepReinforce. It dropped on June 25th. And if you're not paying attention to what it actually does — not just its benchmark scores, but how it achieves them — you're missi

I need to start with a confession. I make fun of this format constantly. The listicle. The countdown. The "5 reasons you're leaving money on the table" TikTok with the guy pointing at floating text. The LinkedIn carousel that says "most people don't know this 🧵" and then numbers things you already knew. I have mocked this format in conversation, in articles, probably in my sleep. And yet here I am. Because sometimes the format fits the content, and I'd rather be a little hyp