Manski

[Manski @crypticmanski]

@jack_9947 DOTS

[Manski @crypticmanski]

@cleeeeeeeeement PRODUCT

[Paul Gotea @paulpagepilot]

going to regret sharing this for free...

$467K/month selling boring consumable products to boring demographics in boring countries. dog supplements. cleaning refills. teeth whitening strips. things that run out and need replacing. 20 products at $3K average. one dies = 5% dip. not a crisis. subscription checkbox pre-checked on everything

I put the full catalog system into a 36-page guide:

> why consumable products are worth 7-36x more per customer than one-time products (the empty container is the retargeting, $0 in remarketing)
> why running 20 boring products beats running 1 viral product (1 of 20 dies = 5% dip, 1 of 1 dies = 100% revenue gone)
> why women 45-65 subscribe for 8-12 months while gen z returns 7.7x per year (she holds 53% of US wealth, he holds 5%)
> how to test 10 new consumable products per weekend for $50 each (the catalog cleans itself, only winners survive)
> how to sell the same boring products in 5 boring countries at $2-4 ads (knee pain in poland is identical to knee pain in ohio)
> the pre-checked subscription box that converts 35% of buyers into recurring revenue (biology handles retention, not your email sequence)

→ the 36-page $0-$10K/mo ecom playbook
→ the 6-point product page checklist to go from 1.2% → 3.4% CVR
→ the full video walkthrough - finding products, building pages, launching ads, going live

Follow me, RT this post, Comment "BORING" and I'll send it over.

[Manski @crypticmanski]

@paulpagepilot BORING

[Henry @HenryCrochemore]

the fastest way to find a winning angle is to steal the map from a competitor who already paid for it

targeting didn't die, it moved into the creative. if all your ads are one angle with different first lines, facebook reads them as one ad and you hit a ceiling no matter the budget

the big spenders don't run campaigns anymore, they run portfolios. resilia runs ads through at least 4 different pages and one of them has 1,658 ads live

so i x-ray competitors instead of guessing. 12 prompts you paste into claude with the gethookd mcp, one product url in, their whole angle map out, then rebuilt for your product

rt + comment "INTEL" and i'll send it
(follow for dm)

[Manski @crypticmanski]

@HenryCrochemore INTEL

[Manski @crypticmanski]

@spwfeijen JSON

[John @johnvirality]

i've been running Sonnet 5.5 + Jev scoring + Higgsfield + Capcut MCP as ONE closed loop instead of a pile of tools...

and it has been ABSOLUTELY cooking

most people open a chat window, ask for 10 tiktok ideas, make one, do 400 views, then decide AI content doesn't work

generation was never the bottleneck. it's the cheapest part of the process and everybody optimised it first

so i documented the ENTIRE machine...

covering research, ICP excavation, idea scoring, image gen, video animation, the edit layer, distribution, and the feedback loop that makes next week's batch smarter than this week's

here's what's inside the guide:

→ the context engine that makes every downstream prompt actually work

(sales calls, DMs, support tickets, closed-lost notes, 90 days of analytics. the system tells you what context it's MISSING, which is usually how you find out your positioning is vague)

→ ICP excavation that builds your hook library from your customers' mouths

(verbatim quotes mapped to pain, failed solutions, objection, awareness stage. any phrase appearing in 3+ separate conversations gets flagged as a probable hook)

→ the research layer: platform scrape + adjacent market mining + competitor gap teardown

(your competitors are all watching each other. the formats that feel fresh in your category were invented in a different one. this prompt steals the mechanics, not the topics)

→ the scoring layer that decides what deserves production

(50 ideas scored across viral, business and execution dimensions. you produce the best 5 of 50 instead of the first 5 you thought of, for the exact same production cost. the part everyone skips)

→ the sonnet 5.5 x gpt images 2.0 x seedance 2.5 x higgsfield production stack

(consistency lock built once and pasted into every prompt. character reference locked across every shot AND every video in the batch. broken renders flagged and regenerated automatically)

→ the capcut MCP edit layer, because editing is retention not polish

(cut on action never after it, no shot over 2.5s in the first 6 seconds, word-by-word captions, hook in frame 1, CTA in the final 2 seconds)

→ the closed-loop feedback system that compounds

(post-mortem names the exact second retention dropped, winners get doubled down on with one variable held constant, context file updates every week. week 12 you're generating from your own performance history instead of guessing)

19 prompts. written out in full. copy-paste and run.

all backed by what i've seen running distribution for 200k+ creators and watching which clips actually convert to installs

like + comment "LOOP" and i'll send the whole protocol over

(must be following + RT for priority access)

will stop sending in 24hrs...

[Manski @crypticmanski]

@johnvirality LOOP

[Manski @crypticmanski]

@realworlddanny GUIDE

[Manski @crypticmanski]

@georgesttock MAKEUGC

[Manski @crypticmanski]

@freddiexpott Data

[Manski @crypticmanski]

@davidfigeira WANT

[Manski @crypticmanski]

@AIAdsApps MRR

[Manski @crypticmanski]

@loisaios SLIDESHOW

[Manski @crypticmanski]

@consumerxai IDEAS

[Amir D @starks_arq]

REALITY cost us $2,000 in Seedance 2.0 credits and did 5M views on Instagram.

I decided to open EVERYTHING behind it:

- 45 prompts, word for word

- the 21 reference images

RT + comment "REALITY" and GROK DMs you an access code.

[Manski @crypticmanski]

@starks_arq REALITY

[Manski @crypticmanski]

@cleeeeeeeeement @theo_louro CARROUSEL

[Manski @crypticmanski]

@adswithcami AI UGC