Loic Profile picture
Sep 9 17 tweets 6 min read Read on X
You pay a 15% secret "Google tax" whenever you buy anything online.

Your pizza costs more because of Google.
Your insurance premium is higher because of Google.

And even the court admits this is illegal.

How Google rigged the system and got away with it: Image
In 2023, the DoJ filed a case against Google for their tech monopoly in advertising.

Google controls the entire "ad tech stack" - the tools publishers use to sell ads AND the tools advertisers use to buy them.

Like owning both sides of a marketplace.

Let me explain: Image
Look at the ads in this image 👇

When you visit any website, an instant auction happens for the ad space you see.

Publishers (websites) sell ad space, advertisers buy it, and exchanges run the auctions.

Everything happens in 100 milliseconds. Image
3 tools make this happen:

• Publisher ad servers (websites use these to sell their ad space)
• Ad exchanges (auction houses where bids happen)
• Advertiser tools (brands use these to buy ads)

Google owns the biggest player in all three categories.

And that's a problem... Image
It's like this: You want to sell your house.

Google owns your real estate agent.
Google owns the buyer's agent too.
And Google also owns the auction house where the sale happens.

Now imagine Google starts rigging the auctions...
Google built this control through 3 key acquisitions:

They bought DoubleClick for $3.1 billion (for publisher tools + ad exchange).

Then AdMob for $750 million (for mobile ads).

And then AdMeld for $400 million (to kill a competitor). They did all this by 2011. Image
Look at this internal email 👇

They knew it was wrong but they did it anyway.

But it gets worse when Google launched secret projects to rig the system... Image
1. Project Bernanke (2013)

Google identified publishers thinking about leaving their platform.

To keep them happy, Google overpaid for their premium ad space.

And then they secretly charged higher fees to publishers who couldn't leave to cover those losses. Image
2. "Project Bell" was pure extortion.

If publishers tried using Google's competitors, Google would automatically reduce their ad bids by 20%.

Publishers called it being "held hostage."

But they couldn't leave since Google controlled too much advertiser demand. Image
When publishers found a workaround called "header bidding," Google panicked.

Header bidding let publishers auction ad space to multiple exchanges simultaneously instead of giving Google first dibs.

So Google launched "Project Poirot" to kill it... Image
3. Project Poirot worked like this:

Google detected which publishers used header bidding, then intentionally submitted low bids on those auctions.

When publishers fed those low bids into Google's system, Google would swoop in with a bid just 1 penny higher to steal the sale. Image
By 2023, Google dominated every part of online advertising:

9 out of 10 websites used Google's tools to sell ads.

50% auction houses belonged to Google.

And 80% buyers had to use Google's purchasing system.

It was almost impossible to avoid them.
So what's the result of all this manipulation?

Google takes 35 cents of every dollar spent on digital advertising.

Competitors charge a fraction of that amount.

In 2024 alone, that's over $70 billion in revenue from controlling the auction house. Image
Every business you buy from pays these inflated advertising costs.

Your morning coffee, insurance premiums, pizza delivery - all include Google's hidden advertising tax.

They pass every penny to you through higher prices.

Ideally Google should be punished for this...
But ChatGPT's and other LLMs presence saved Google.

According to the ruling: Google might not be able to compete against AI companies in the future.

What do you think of this?

Let's talk in the comments. Image
Thanks for making it to the end!

I'm Loic - indie hacker, Slow-nomad, and a product tinkerer.

Currently building something exciting at ColdIQ (coldiq.com).

Previously built Podly.co & TubeRocket TubeRocket.com
Like and RT the first tweet if you found this thread useful.

Follow me @LoicReco for more threads on SaaS, Indie Hacking, AI, and all things technology.

More about me: loic.so x.com/13562517895202…

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Loic

Loic Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @LoicReco

Sep 2
100 lava lamps in San Francisco generate the encryption keys securing 48% of the world's top 10,000 websites.

The company that owns them blocks 190 billion cyber threats daily, controls $70+ billion in market cap, and when it goes down half the internet disappears.

Thread
These lava lamps belong to Cloudflare.

Cloudflare needs truly random numbers to generate encryption keys that can't be predicted or hacked.

So they photograph 100 lava lamps continuously because the wax patterns never repeat exactly.

But there's more... Image
Cloudflare's London office uses double pendulums.
Austin has suspended rainbows.
And Lisbon uses ocean waves.

The camera captures images every millisecond, converting each pixel's RGB values into unpredictable data streams.

But why don't they rely on computers for this? Image
Read 15 tweets
Aug 19
The average American spends 5.16 hours daily (13 years in a lifetime) on their phone.

Scientists wanted to see what would happen if they blocked their internet access.

What happened next completely baffled the scientists...
Scientists approached 467 people and asked them to block their mobile internet for 2 weeks.

But only 25% actually kept it blocked the full time.

75% "failed" and cheated.

But here's what broke the researchers' minds... Image
Even the people who failed saw massive improvements.

91% of participants improved on at least one major life measure.

Most antidepressants have a 40-60% success rate. But a phone restriction had a 91% success rate.

Read this again! Image
Read 11 tweets
Aug 17
You’ve been lied to.

The healthcare system doesn’t want you healthy: they want you sick, tired, and obedient.

Debunking the 11 biggest health myths they planted in your head:

1/ Myth: Eggs are bad for your heart
Truth: Eggs don't cause high cholesterol for most people.

They're loaded with vitamins, amino acids, healthy fats, and are a great source of protein.

Nature's multivitamin got demonized, so you'd buy processed cereal.
2/ Eating Fat makes you Fat

Truth: Healthy fats support hormones, brain health, vitamin absorption, and keep you full.

Science shows that high-fat diets like keto are better for losing fat.

Sugar and refined carbs are to blame for the obesity epidemic.
Read 16 tweets
Aug 11
How does a checkbox know I’m not a robot… when I didn’t click anything?
Because reCAPTCHA isn’t about bots, it’s about you.

It fingerprints your device across millions of sites, scores your “risk,” and learns from your clicks.

Why that’s spyware dressed up as security🧵 Image
The moment that checkbox appears, Google's already been watching.

Before you click, it recorded your mouse movements, your typing rhythm, your browser fingerprint, and whether you're logged into Gmail.

The checkbox is theater. The decision was made.
And it gets worse.. Image
Same person, different browsers:

Chrome + Google account = instant pass
Firefox + privacy settings = 5 rounds of puzzles
Tor Browser = locked out completely

The "bot detector" punishes privacy and rewards data surrender.
Then a developer exposed what's really happening.
Read 18 tweets
Aug 4
I noticed everyone arguing about XML vs JSON prompts, which made me curious. So I tested both thoroughly.

5 problems, 2 LLMs, 20 tests total. XML vs JSON.

Here's what actually works when prompting LLMs (it's not what anyone expects):
The first test result made me run it again because I couldn't believe it.

Claude and GPT don't just prefer certain formats. They become different programmers based on how you format your prompt.

Let me show you what 20 experiments revealed...

(Go to the last tweet for results)
My test methodology was rigorous but simple:

- 5 classic coding problems
- Each problem tested on both Claude and GPT
- Identical requirements, just different formats (JSON vs XML)

Analyzed not just if code worked, but HOW they solved it Image
Read 18 tweets
Jul 28
How to use ChatGPT's Agent Mode to automate literally anything on the web.

No blocks. No visual builders. Just natural language.

Here's exactly how I automated 7 real tasks (with prompts you can steal):
I'm a SaaS builder who's tried every automation tool out there.

Lovable, , Zapier - they all require learning their interface, building flows, debugging connections.

ChatGPT Agent Mode? I just describe what I want in plain English and it executes.

Real examples:Make.com
Example 1: API Documentation Explorer

I needed to integrate Notion's API. Instead of reading 50 pages of docs,

Agent Mode did this:

Read the entire API documentation
Identified authentication methods
Generated working Python code with comments
Created 3 ready-to-use examples

One prompt. 5 minutes.
Read 19 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(