the 30x marketing model nobody is using yet
|
marketers spend most of their week making small decisions, not writing. which lead is worth contacting. which message fits that lead. which ad creative to test. is this search term coming from a buyer. is this ad fatiguing. does the landing page match the promise. each one is small. thousands per week makes them expensive. last week i wrote about JEV, the decision model from TypeSafe AI. quick recap: an LLM generates an answer word by word. JEV takes a situation and a fixed list of possible answers, picks one, and returns a confidence score. an LLM generates language. JEV evaluates choices. that shape, many inputs, one standard, a judgment per input, is exactly what marketing work looks like. here's what it looks like in practice: our own case. we score every video idea before we film it. we pulled 248 shorts and 415 titles from our niche with real views and engagement, built a baseline, and now every draft gets scored against it before filming. a weak idea gets rejected in a score instead of in the edit. that judgment layer used to be the most expensive part of our stack. now it costs cents, so we score everything instead of only the script we were already sure about. watch the full walkthrough: seven marketing workflows at 30x for under $3. one marketer runs JEV across the whole meta ad library, tags every live ad by hook, format and offer, sorts google search terms by buyer intent, catches ad fatigue, and scores every form fill 0 to 100 against the ideal customer.
the same client seo run, from $250 to $25. same output, not a better prompt. just 30x faster on the boring parts.
700 leads scored in 40 seconds for $0.09. including how each outreach message would perform and which leads got mismatched to the wrong message.
there's a non-marketing one too: a browser agent finding flights in 7 seconds for $0.0039. same pattern, a stream of small decisions at every step.
the bigger idea: LLMs are great at creating. JEV is great at deciding. and marketing needs a lot more decisions than people realize. JEV doesn't replace chatgpt or claude. they keep writing. the decision layer moves to a model built to decide. real performance data calibrates the baseline. you make the final call. your move this week: pick one repetitive judgment you make by hand every week. lead scoring, ad fatigue checks, brief scoring. write out the possible answers once, then let a decision model handle every instance after that. the full write-up with the scoring prompts we use is here:
if you want to set one of these up for your business, book a call see you in the next one,
joon
|
Joon Ahn's newsletter
hey JEV is insane. it scores your video before you film it, so you know if it's worth making. let me explain. most AI tools write. ChatGPT and Claude generate scripts, hooks, titles, whatever you ask for. JEV is different. it's a new model from TypeSafe AI, and it doesn't write anything. it judges. you give it information plus specific questions. it returns decisions, scores, and confidence your software can use right away. LLMs write. JEV decides. why this matters AI made content cheap....
hey. new long-form video is up. this one solves the most annoying problem in ai video: your character looks different in every shot. the short version: stop generating videos from one random portrait. build two reference files first. one defines who the character is. one defines how the shot looks. then every scene you generate inherits both. i just published the full build: How to Make Consistent AI Characters with GPT Image 2.5 + Seedance. same character across three separate videos....
hey. new long-form video is up. i think short-form editing just changed and most people haven't noticed yet. the short version: you describe the edit in plain english. astra analyzes the footage, marks where to cut, and builds the motion graphics inside ae. i never open the timeline. i just published the 12-minute full build: GPT-6 Astra Changes Short-Form Editing Forever. same avatar video on the left, the astra + after effects edit on the right. same script, same voice, same avatar. only...