OxiClean. ShamWow. P90X. Insanity. NutriSystem. Those things ran for years. Millions in media spend. and here's the wild part: most of them follow the exact VSL structure everyone teaches today. Problem → Agitation → Unique Mechanism → Demonstration → Proof → Offer → Urgency they figured it out in 1997 with a stain remover. before funnels were a thing. before VSL was even a term. so here's a 30-minute research session worth running: 1. ask Claude to find a brand in your niche (or a similar one) that ran infomercials in the 90s or early 2000s 2. find every infomercial they ran, pull the YouTube URLs 3. drop them all into NotebookLM 4. have it extract their USPs, unique mechanisms, and overall flow then go steal one or two of the tactics for your own VSL. you're borrowing a structure that got pressure-tested with real dollars against real audiences for years. p.s. here's my personal fav infomercial of all time: https://www.youtube.com/watch?v=VgCge6_f4B0&t=1235s ... RIP Billy Mayes

had a good question come up this week. someone running about 100 new ads a week through a small testing budget, and their media buyer started shutting a lot of them off because they'd never get a fair test at that spend. that's the right call, and here's how i think about it. the main thing to keep in mind is that the whole point of testing creative is to hit your KPIs at the spend level you're trying to run. so before you ask how many ads you should be testing, the real question is whether you're already hitting your numbers at the budget you want to be at. if you are, there's no reason to flood a bunch of new ads in and spread your budget thin. all that does is add noise and starve the winners that are already carrying you. you end up worse off than when you started. new creative becomes the lever when you're trying to scale past where you are now, maintain your KPIs, or hit something your current setup isn't getting you. that's when it makes sense to lean into volume and test more aggressively. otherwise your testing plan should revolve around your current goals rather than testing creative for the sake of it. as a rule of thumb, i don't like to test more than about 30% of my budget on new creative. the rest stays behind what's already working. so if your numbers are healthy, stop trying to test everything. protect the winners, test inside that 30%, and save the heavy creative pushes for when you're actually trying to scale.

we don't really use landing page builders anymore. no clickfunnels. no leadpages. we barely touch the gohighlevel builder. we build the whole page with AI now and they're coming out better than what most 5k/mo "funnel agencies" ship. I sat in on our team power hour today and watched one of our guys run the whole thing start to finish. the build is the easy part now. it's a couple things nobody warns you about that actually decide whether the page makes money. here's the part i actually care about. post andromeda, the ad is the targeting. you don't pick the audience anymore, the creative does. and the page you send to is part of that signal. which means a unique landing page per angle isn't a nice to have. it's a whole new audience pocket. old world, building a custom page for every angle was a joke. nobody's spinning up 8 different pages for 8 different callouts. so everybody ran every angle to the same generic page and left pockets on the table. now i can build a page that mirrors the ad in like 5 minutes. so the "are you a [persona]" ad goes to a page that opens with that exact persona. the dark mode chart ad goes to a page that looks like the ad. the angle and the page match all the way down. same offer. same funnel. the only thing changing is the angle and the page it points to. and every winner becomes a brand new pocket meta couldn't reach before. that's the unlock. not "we make pages faster." it's "we can finally run a dedicated page per angle, which is exactly what andromeda rewards." the actual process, step by step: **1. scrape the brand.** open claude in the browser, point it at the client's site. it pulls copy, images, logos, exact colors, exact fonts. dumps it all into one markdown file. this is the part that used to take an hour of copy pasting. now it's a few minutes. (the claude plugin is great at this) **2. generate the page off your skills.** feed that file into your landing page build. we run it through a CRO skill (handles the form, the structure, the conversion stuff) and a copy skill. set the visual direction here and be specific. "apple inspired" or "comical and loud" or "clean editorial." this is what keeps it from looking like every other AI page.

I've been pretty obsessive lately with testing every new AI tool - especially when it comes to tools that can help me drive more new creatives in my ad accounts. Up until recently, Google's Nano Banana 2 was the king at image generation. That is, until yesterday. ChatGPT dropped their new images 2 model, and it's REALLY good. I did a blind split test in the youtube video below, pitting the two models against each other.