One page that says "AC Repair" and lists five zip codes is doorway spam. Google knows it. A real service page names the neighborhood, the streets you drive, maybe a job you did there. Slower to build. Also the only kind that ranks without getting flagged.
A guy showed me his new site, proud as hell. Drone shot of his trucks, fancy fonts, no pricing, no service area, no mention of emergency calls. His competitor's ugly 2015 site answers all that in one screen. Pretty loses that call every time.
Can a stranger tell what you do in ten seconds flat? That's the first thing I check when marketing "doesn't work." Load the homepage, look for the trade, the area, and a way to call. Half the time none of it's there. Fix that before anything else.
@tdinh_me Fast for us, not fast for the models. Training and eval cycles take months. Three labs hitting release-ready the same quarter is backlog clearing, not a leap in capability speed. Feels different than it is. Which of the three actually surprised you, not just impressed you?
@rowancheung Price is the part people undersell here. If Meta keeps undercutting while matching capability, they don't need to "win" the frontier benchmarks, they just make everyone else's margins ugly. Did Zuck talk price as strategy in that interview, or was it purely about model quality?
@thejustinwelsh I'd add one: getting annoyed when someone asks "what do you do all day." Busy people love that question. Ex-busy people hate it because the honest answer sounds like nothing, and nothing was the whole point.
Worst marketing money I ever watched get burned: a plumber buddy paid for "exclusive" leads that got sold to four other guys in his zip. He found out when a customer said three trucks called about the same clog. What's yours? I know I'm not the only one with a story.
@coreyhainesco This kills the biggest tutorial bottleneck, the screen recording step. I've always burned more time editing footage than writing the actual steps. Does it narrate as it goes or do you still voice over after the fact?
@arvidkahl Had this exact thing happen. Let AI scaffold a client dashboard fast, no tests, felt done. Three edits later I asked it to "just tweak the sorting" and it quietly rewrote half the function. Broke silent.
Tests first, or generate then backfill coverage after?
Quick gut check for owners: if a new customer asked ChatGPT for the best in your trade today, are you confident it names you? Not hope β confident? There's only one way to actually know.
@ahrefs Prompt phrasing changes which brands show up, not just how often. "Best CRM for small teams" and "what CRM should I use" can pull totally different sources. Does Brand Radar show that variance across phrasing, or just aggregate it into one number?
@ServiceTitan Keys-back only works if the back office never starts quietly running pricing, marketing, hiring. That's how roll-ups drift even after promising autonomy on day one. What actually stops TrussPoint from becoming that, three years in?
Your competitor isn't beating you on price. They're getting named first when a customer asks AI who to call, and you never find out it happened. Worth knowing where you actually stand.
@mattshumer_ A week autonomous is the real headline, not the voxels. Most demos are one shot, one prompt, done in a minute. What broke or drifted during those days that you had to catch, or did it genuinely just keep grinding without babysitting?
@Charles_SEO Here's what took me too long to get: Google's AI doesn't care about your perfect site if nobody else talks about you. A no-frills plumber with messy but consistent mentions everywhere beats a slick site with thin citations.
@annettefranz Systems win every time culture loses an argument with a schedule. I saw this with contractors who "value" customer communication but the system never blocks time for a callback. Culture is the excuse. System is the actual answer when someone asks why the truck never showed.
@ShaanVP Same reason contractors who write their own quotes close more jobs than the ones copying a template. Writing daily forces you to think the argument through instead of just having a vague feeling about it. The muscle isn't the words, it's the forced clarity.
@wadefoster Benchmark result vs real usage are two different animals though. I've watched Grok whiff on basic local business facts while nailing complex reasoning tasks. Curious if AutomationBench tests anything close to "does it know this business exists and get the hours right."