Morning Brief

Berman's right about one thing: most people who evangelise AI coding loops never tell you what they actually are. So here's the plain version. A loop is your coding agent working on its own toward a fixed goal, human pulled out of the middle. Two ingredients: a trigger (manual, scheduled, or fires on an event) and a goal. Goals split two ways. Verifiable: "every page under 50ms," something you can test deterministically. Or LLM-as-judge: "refactor until you're happy," where the model calls it done. Here's the owner's lens. The verifiable loops are where the money is. The page-speed one is the cleanest example going: define the test condition, let it grind, walk away. Real outcome, measurable. The judge loops are where it gets wobbly, and they burn tokens like a bonfire (some run for days). My confessed mistake: early on I let an agent loop on a vague "improve this" goal overnight. Woke up to a fat bill and changes I couldn't verify. Useless. The discipline: only loop on goals you can measure. If you can't write the test, you can't trust the result. That's not an AI rule. That's a P&L rule. What's the first job in your stack you'd actually trust to run unattended?

I tried to "do" stillness for years. Booked it in like a gym class. Twenty minutes of sitting cross-legged, feeling smug, then straight back into the storm the second my phone buzzed. Treated peace like a thing I visited rather than a thing I was. Eckhart Tolle put a crack in that for me. His point: the calm isn't the twenty minutes on the cushion. It's the awareness sitting behind your thinking, the gap between one thought ending and the next one starting. The deal closes, that bit of you is fine. The deal collapses, that bit of you is fine. The paint changes. The canvas doesn't. And before that sounds like a fridge magnet: this isn't going dead inside to dodge pain. It's the opposite. There's a quiet warmth under the racket, not just an absence of one. Here's the bit that holds up Monday morning. You don't have to stop thinking (good, because you can't). You just notice there's a watcher in there, and rest your attention on the gap instead of getting dragged along by every passing worry. Still cocking this up most days, mind. When did you last catch yourself watching your own thoughts instead of being bossed by them?

Right then. Kevin O'Leary did a run through six "accessible luxury" watches, all under or near five grand, and pitched them as appreciating assets with an insurance plug at the end. 150% payout on total loss. Lovely. Here's the bit founders should clock, because it's the oldest move in consumer goods: the price ladder. TAG F1 at $1,950 is the door. Speedmaster sits at the top as the "what should I get?" default. Every product in between is a rung that makes the next one feel reasonable. I ran this badly once. I built a brand with three SKUs all priced within a tenner of each other, congratulated myself on "simplicity," and watched everyone buy the cheapest one. No anchor. No trade-up. I'd flattened my own ladder and called it focus. The thing isn't variety. It's the spread. A real entry point, a clear hero, and a top that makes the middle look sensible. Owners get this in their bones: margin lives in the mix, not the bestseller. So look at your range. If you removed your most expensive SKU tomorrow, would your middle still sell? If not, you don't have a ladder. You have a list.

The first time someone said "Christ consciousness" to me I nodded along like I knew what they meant. I had no idea. Turns out neither did they. So I respect anyone trying to pin the phrase down. The framing I keep bumping into goes like this: the divine isn't somewhere else, it's the intelligence threaded through everything, and "sin" isn't a bloke on a cloud keeping score, it's just missing the mark on a spectrum, cause and effect. Here's where I get off the bus though. The model also throws in opal lakes, mermaids, and evil spirits lobbing "mental vibratory rays" at each other. That's not insight. That's a fantasy novel with a karma plugin. So I do what I do with all of it: keep the bit that holds up Monday morning, bin the rest. And the bit that holds up is dead practical. You reap what you sow. Aim badly, miss the mark, adjust, try again. No cloud. No scorekeeper. Just consequences. You don't need a cosmology to take responsibility for your own aim. You just need to stop blaming the weather. Where's your line between a useful frame and a load of woo dressed up as physics?

Right then. Bryant Chou, Webflow's old CTO, is back with a new one: Ploy. The pitch is your website plus a marketing brain bolted on. It rebuilds your site, learns your brand into a proper design system, plugs into 50-odd tools, runs your ads, writes your copy, and gets you found inside ChatGPT, Perplexity and Claude. 12% of his YC batch are already on it. It's clever kit. The demo (rebuilding ancient founder sites better than the founders ever wrote them) is genuinely impressive. Here's the bit that gets lost in the noise. A faster, prettier, AI-discoverable site does not fix a broken business. I learned this the expensive way: we once threw budget at a beautiful rebuild while our LTV to CAC was upside down. Traffic went up. We still bled money. The site wasn't the problem. The unit economics were. So treat tools like this as plumbing, not strategy. The question isn't "can I ship a site in 75 seconds." It's "what does a new customer actually cost me, and what are they worth." Ploy can do the front door. It cannot decide whether the house is worth building. What's the number that actually breaks your P&L right now? Tooling, or maths?