Eric Jackson Profile picture
Dec 26 11 tweets 2 min read Read on X
In a power-law world, a small number of outcomes drive most results.

Upside is the visible part.
Survivability is the hidden one.
Power laws don’t reward the smartest ideas.

They reward the ideas and the people
who can stay alive long enough
for the distribution to reveal itself.

That’s a very different skill.
Being early and right often looks identical
to being wrong.

Same drawdowns.
Same ridicule.
Same uncertainty.

Only time separates the two.
Take Masayoshi Son and Alibaba.

The insight wasn’t the hard part.
The hard part was surviving years of doubt and volatility
long enough for the bet to matter.
Or look at NVIDIA.

For years it was “just a gaming GPU company.”
The thesis was early.
The patience was expensive.

Power laws never feel obvious while you’re inside them.
Or Carvana.

The idea didn’t change much.
The structure almost broke.

What mattered wasn’t optimism.
It was whether the container could survive long enough.
This is why so many “correct” theses fail.

Not because the idea was wrong.
But because the structure holding it
couldn’t tolerate variance.

The math didn’t fail.
The container did.
People chase power laws by focusing on upside.

They ignore the harder part:
•drawdown tolerance
•pacing
•optionality
•the ability to do nothing under pressure

That’s where outcomes usually collapse.
Power laws punish fragility more than error.

A fragile structure with a great idea still dies.
A durable structure with a decent idea can compound.

The power law never apologizes.
Power laws aren’t about predicting the big winner.

They’re about not getting eliminated
before the big winner can matter.

Survival comes first.
Asymmetry comes second.
If you want to understand power laws, ask this:

“What breaks first when I’m wrong for longer than expected?”

That answer matters more
than any upside projection ever will.

• • •

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

Keep Current with Eric Jackson

Eric Jackson 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 @ericjackson

Dec 26
Most financial advice isn’t designed to make you rich.

It’s designed to make the advisor unfireable.

That distinction explains almost everything.
Index hugging isn’t a strategy.
It’s career insurance.

If everyone owns the same thing, no one gets blamed.

That’s great for institutions.
Terrible for individuals.
This is why Wall Street loves the “Mag 7” narrative.

Big. Liquid. Consensus.
Easy to explain. Easy to defend.

You don’t need to think.
You just need to conform.
Read 6 tweets
Dec 25
A lot of people are asking the wrong question about NVIDIA buying Groq.

The question isn’t “why now?”
It’s what weakness this quietly covers — and why Groq agreed to it.

Here’s the real explanation.
Start with NVIDIA’s uncomfortable truth.

GPUs dominate training.
Inference is different.

Inference is:
• margin-sensitive
• power-constrained
• long-duration
• optimized ruthlessly by buyers

That’s exactly where GPUs are least defensible.
Most production inference does not need a $40k GPU.

It needs:
• determinism
• low latency
• throughput per watt
• predictable cost

That’s why LPUs / ASIC-style inference architectures keep resurfacing.

Not because GPUs are bad —
but because they’re overkill.
Read 11 tweets
Dec 25
Credit to @FransBakker9812 — his February “fake data center” take aged well.

But NVIDIA buying Groq adds an important new dimension that changes how I think about IREN, CIFR, and HUT.

Here’s the distinction most people are missing.
The NVIDIA–Groq deal doesn’t reward “more megawatts.”

It rewards optionality across compute regimes.

Training.
Inference.
GPU.
LPU / ASIC.
Hybrid.

The winners are the operators who can reconfigure, not just host.
Groq’s critique in February was valid — but it was narrow.

Groq runs inference-first, small-facility, tightly coupled deployments.

That is not the same problem hyperscalers are solving at 100MW+ scale.
Read 8 tweets
Dec 25
Most people think conviction means confidence.

It doesn’t.

Confidence is how you feel when you’re being proven right.
Conviction is what you rely on when you’re being proven wrong.
I learned this the hard way in 2021 and 2022.

I was wrong.
Early.
Publicly.
And for longer than felt reasonable.

Those years didn’t test my ideas.
They tested my stamina.
Markets don’t break people with losses.
They break people with time.

Time spent underwater.
Time spent doubting yourself.
Time spent watching consensus mock you.

That’s where most “conviction” quietly dies.
Read 10 tweets
Dec 25
Most people think Stoicism is passive.
Emotionless. Detached. Conservative.

That is a misunderstanding.

Stoicism is one of the most aggressive forms of capital preservation ever invented.
Stoicism is not about avoiding action.
It is about controlling inputs so action is never forced.

Markets punish forced decisions more than wrong ideas.
Volatility does not destroy capital.
Reactivity does.

Stoicism is a system designed to survive variance without breaking posture.

That is not philosophy.
That is risk management.
Read 8 tweets
Dec 25
Most “crypto treasury” strategies fail for reasons that have nothing to do with price.

People argue BTC up or down.
That debate misses the real failure mode entirely.
The mistake is treating volatile assets like static holdings.

Volatility is not just a return profile.
It is a decision generator.

If you don’t design for that, price eventually forces your hand.
Treasuries don’t blow up because assets go down.
They blow up because organizations are forced to decide at the wrong time.

Liquidity stress.
Optics.
Governance panic.

Price is just the trigger.
Read 8 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!

:(