Sabitlenmiş Tweet
Garrett Gray
6.6K posts

Garrett Gray
@avantgray
Google & Meta suck at new customer acquisition. We fix that. Sharing my two cents on performance marketing, 1st-party data ops, & psychology
Beyond CAPI ➡️ Katılım Eylül 2021
996 Takip Edilen2.8K Takipçiler

@4nt1p4tt3rn @mqudsi @4nt1p4tt3rn can you update us on this please ?
my sister in law is having a hell of a time with her cancer that came back & any additional details you could share on this would be greatly appreciated
English

@mqudsi I'm going to get the details when we go pick up more eggs from them Sunday.
English

So, a word on the whole "guy uses AI to cure his dog's cancer" story that's going around:
I don't know if it's true or not. It's somewhat irrelevant to what I'm about to say.
What I do know is this: My wife and I are friends with a couple a few towns over. We buy duck eggs from them. The husband, a very alert, physically capable engineer in his 70s, got cancer.
They are absolutely avoiding the accepted treatments.
Instead, they're pursuing a therapy that mirrors what the guy with the dog dig: They paid a lab and got the tumor sequenced, and paid a company to develop a vaccine targeted specifically at the tumor's gene sequence.
And it's working. Spectacularly.
This isn't a novel approach. It's out in the world, and there are companies, doctors, and labs offering it. But the medical industry won't tell you about it. Your doctor isn't going to offer it to you as an option. You have to go looking for it.
It's not cheap, but it's cheaper than a course of chemo.
And it actually works.
English

@ChrisLangSocial oh, that's cool
we just recently moved to St George, Utah and insane levels is a correct assessment
English

@avantgray i literally had a shopify store where I would just travel around the sw and capture the beauty
insane levels
English

I've built 4 x 7-figure businesses from Las Cruces, New Mexico.
Population 115,000. Middle of the desert. Not exactly a startup hub.
Nobody cared.
The old model was proximity. Be in the right city. Network in the right rooms. Shake the right hands.
That model still works. It's just not the only model anymore.
The new model is distribution. If your customer can find you online, it doesn't matter where you sleep.
What actually matters:
+ Systems that run without you present
+ A team wired for async communication
+ Distribution you own
None of those require a zip code.
Operators who build this way get something most founders never do.
Optionality.
You can move. Travel. Relocate. Step back. The business doesn't care.
The city was never the advantage. The systems are.
Build those first. Then operate from wherever makes sense for your life.
English

@ChrisLangSocial Dope! The average American does not realize just how awesome our National Parks are.
English


In 2026, Claude became my co-founder.
Not an assistant a system running 40% of my agency ops.
I packaged everything into one resource:
→ Claude Projects Architecture
→ Claude Code Setup
→ Claude + n8n MCP
→ Claude Skills Blueprint
→ Query MCP + SEO MCPs
200+ hours to build. Free for you.
If you want it:
Like + Comment “NEED”
I’ll DM you the details.
Make 2026 the year you stop using AI like a search engine. ❤️

English

@iamjakesmind Well said, it's refreshing to see someone make a post like this as opposed to 'Meta sucks, I lost all of my money'
p.s. - if you're open to it, hit me up sometime & let's compare notes on what we're seeing in the last 8 months with GEM & Andromeda
English

Meta's CFO said something this week that stopped me in my tracks. So, I'm sitting down to write you this beautiful note about it.
Susan Li said, "We are not by and large using LLMs for ranking and recommendations work yet."
For me, this was a good reminder of the actual roadmap not the posts I read every day.
Let me explain what's really happening under the hood. In the way that my brain works (hopefully it helps you as well).
The system deciding who sees your ads right now is three things: Andromeda, GEM, and Lattice.
Not the ChatGPT, Llama-esque frameworks you think about.
A purpose-built machine that's been quietly rebuilt from the ground up since late 2024.
Andromeda is the retrieval layer.
It scans millions of ads and builds a shortlist before the auction even starts.
It doesn't start with your audience. It starts with your creative, reads it, and finds the people most likely to respond.
GEM is the ranking brain.
I'd call this LLM-inspired, trained at that scale, 4x more efficient than what it replaced.
It scores Andromeda's shortlist and picks a winner.
Genuinely sophisticated and already showing the results.
But here's the distinction worth understanding: GEM borrows LLM architecture while still optimizing patterns from historical signals.
Not reasoning from scratch.
That's exactly what the CFO was talking about and it's easy to miss.
Lattice sits after GEM and handles the final auction and placement decision.
That future version... where the system reasons about your customer from scratch without needing years of engagement history first...
Yeah, that still doesn't exist. Yet.
That's the thing that requires data centers they're still building, some of which are currently housed in tornado-resistant tents.
Not a joke.
So here's my honest picture.
The AI running your ads today is real, it works, and it's the best it's ever been.
But the version where Meta reasons about context and intent at a true LLM level across billions of users in real time for your ads... my gut says that's a 2028 story at the earliest.
The roadmap alongside that is real though.
Profit-based value optimization, custom attribution pulling from Triple Whale and Northbeam, combined awareness and sales in a single campaign structure...
Landed, landing, or announced for this year and none of it requires LLMs to work beautifully.
The gap between Meta's marketing language and reality is still a little wide for my taste but understandable.
Hope this helps.
The platform is not broken and the direction is not a lie.
They're just building something that takes years to do right.
At their scale it has to.
English

block just fired 40% of their company.
amazon cut 16,000. again.
dario amodei said 50% of entry-level white collar jobs are done within 5 years.
here's what nobody's telling you:
the people who survive aren't the ones updating their resume.
they're the ones who became the "AI person" their company can't replace.
i watched this happen in real time with 3 clients this month:
→ marketing coordinator learned to build lead gen workflows - promoted to head of ops
→ office manager automated the entire invoicing pipeline - now runs a department of one
→ junior accountant built a reconciliation bot - saved the company $340K/year and got a raise
none of them wrote a single line of code.
all of them used plain english workflow builders.
average time from "i have no idea what i'm doing" to first working automation: 47 minutes.
i put together a free survival guide PDF:
→ the 5 highest-value automations to learn RIGHT NOW (by industry)
→ copy-paste prompts for each one
→ the "show your boss" script that turns you from expendable to essential
→ how to price and sell these if you go freelance
→ full MCP + synta setup walkthrough
comment "SURVIVE" and i'll send it.
while everyone panics about AI taking jobs,
be the person who uses AI to become unkillable.
synta(.)io - describe the workflow. it builds itself.
(must be following for DM)
English

This is so absurd, God forbid someone might have a differing opinion.
So by that logic they’ll hire the less talented & less impressive person/agency just so long as they agree with their politics?
Do people hear themselves with this nonsense?
Never fun to not close a deal but sounds like you dodged a bullet
English

Have had multiple calls this year with leads who brought up politics during the call and basically saying they will not work with a business that doesn’t align with their political views.
First time I’ve ever experienced this in 15 years of running businesses. It’s genuinely shocking and also incredibly sad.
English

quality & quantity of data being fed back to Meta and Google are the two biggest drivers of success or failure when it comes to training the algorithms
my point is simply that attribution tools solve A problem but they don't solve THE problem which is data retention with the ability to reactivate those users alongside their behavior
I believe Northbeam does this to some extent now with Apex but last I checked that's a pretty pricy option compared to what we offer our clients
English

@avantgray what are you using past 30 days though? feels like that's more of a Klaviyo/CRM problem than an attribution tool problem.
English

@GovPritzker oh the irony of @GovPritzker criticizing someone's job performance
yet another partisan hack
English

@SahilBloom @levelsio @SahilBloom you should check out MuscleEgg
I’ve been using them for years to help me hit my daily protein
English

@levelsio I eat 225g protein per day in two main meals without any protein supplements.
Post-workout:
- 2 eggs
- 30 tbsp egg whites
- 1 serving cottage cheese
Dinner:
- 12 oz lean meat
- 1 serving cottage cheese
- 2 eggs
Dessert:
- 2 servings Greek yogurt
Skill issue :)
English

Anyway, back to business.
New brand is fuckin’ ripping.
This is what happens when dropshipping royalty builds a custom product.
12% conversion rate, whilst ALSO accumulating heavy MRR...
We’re launching 20+ ads a day & 95%+ of them are AI.
We’re pretty much finding a new winner every time we hit publish.
It’s insane.
Ecom is THE best fuckin’ industry to be in.
BCOS AI IS ONLY MAKING IT STRONGER.
Remember - ecom is a physical business. Not a digital one.
I’ve been in ecom almost 10 years & I’ve NEVER felt so bullish.
We’re going to sell this brand for $1bn.
God bless.

Nabeal Khan@nkecom
Ecom is so dead. Our new brand only did $10k first month. - $10k rev - $3k MRR - $0 profit Would have made more money if I went to a strip club & picked up a dollar from the floor. EVERYTHING IS AI TOO. - AI product research - AI product pics - AI positioning - AI offer building - AI design - AI static ads - AI copywriting. These robots are going be the death of ecom man. Might as well quit now.
English

i built 31 n8n workflows this month that replace
the most overpriced saas tools businesses pay for.
→ $299/mo email marketing platform — replaced
→ $199/mo social scheduling tool — replaced
→ $149/mo lead scoring software — replaced
→ $99/mo form + crm connector — replaced
→ $249/mo client onboarding system — replaced
total saas spend eliminated: $11,388/year
total time to build all 31: one weekend
i documented every single one in a free pdf.
reply "WORKFLOWS" + repost and i'll send it to you
(must be following so i can dm)
English






