Pinned post
Bruno Ahualli
Verified account This account is verified. Learn more@brunoahualli
Founder @HelloKilonova. Serial CMO to early-stage Web3 projects. Launched multiple 8 fig DTC brands. Rational Optimist
5.1K Following 2.6K Followers 3.9K Posts
Latest posts 5
@willdepue just pulled off some legendary GPU forensics.
His evidence says DeepSeek, Moonshot, MiniMax and Ant Group all trained their latest models on Nvidia Blackwells that are banned from export to China.
If it holds up, US chip export controls have basically failed.
The report was fascinating, so I made an explainer 👇"
His evidence says DeepSeek, Moonshot, MiniMax and Ant Group all trained their latest models on Nvidia Blackwells that are banned from export to China.
If it holds up, US chip export controls have basically failed.
The report was fascinating, so I made an explainer 👇"
4:09
A @garrytan soul md update sets the scene: For a long time, “good enough” was not a moral failure. It was adaptation.
Research was expensive.
Implementation was expensive.
Testing was expensive.
Documentation was expensive.
Cleanup was expensive.
So people learned to live with partialness.
After all, completeness had a real price.
But AI is collapsing the marginal cost of completeness across a growing share of knowledge work.
Not all of it.
But enough to matter.
Enough that unfinished work increasingly stops looking like necessity and starts looking like choice.
The old standard was built for expensive labor. The new one cannot be.
This does not mean “do everything.”
It means do the right things.
The bottleneck is moving upward.
From labor
to judgment.
The question becomes: What is worth finishing?
Many teams will use AI to generate more half-work at much greater speed.
More drafts.
More patches.
More features.
More movement.
Still incomplete.
Loosely reasoned.
The sharper founders will raise the standard on the things that matter:
Critical path work.
Customer-facing truth.
Core product behavior.
Canonical systems.
Structural fixes.
Institutional memory.
Anything that compounds.
Such things should stop receiving the “good enough for now” treatment when the real finish is within reach.
That is what boil the ocean should mean.
When something truly matters, stop normalizing the half-finished version of it.
A founder should still ask:
•Is this core or peripheral?
•Is this compounding or cosmetic?
•Is this the real fix or a distraction?
But once the answer is yes, the standard has to rise.
So do the real fix.
Close the loop.
Run the test.
Write the documentation.
Resolve the contradiction.
In an age of cheap generation, completeness becomes a strategic weapon.
It compounds trust.
Anyone can generate motion now.
Few can finish the thing that actually matters.
That is the new founder standard:
Finish the real thing.
Boil the ocean, but only the ocean that deserves boiling.
That is the hard part.
That is where judgment lives.
The machine can increasingly carry the labor. Which leaves the founder with a harder burden:
to decide what deserves completeness,
and to stop hiding bad prioritization inside unfinished work.
For a long time, “good enough” was the language of constraint.
In many domains, it is becoming the language of drift.
Stop leaking force through a thousand tolerated incompletions.
Boil the ocean.
Research was expensive.
Implementation was expensive.
Testing was expensive.
Documentation was expensive.
Cleanup was expensive.
So people learned to live with partialness.
After all, completeness had a real price.
But AI is collapsing the marginal cost of completeness across a growing share of knowledge work.
Not all of it.
But enough to matter.
Enough that unfinished work increasingly stops looking like necessity and starts looking like choice.
The old standard was built for expensive labor. The new one cannot be.
This does not mean “do everything.”
It means do the right things.
The bottleneck is moving upward.
From labor
to judgment.
The question becomes: What is worth finishing?
Many teams will use AI to generate more half-work at much greater speed.
More drafts.
More patches.
More features.
More movement.
Still incomplete.
Loosely reasoned.
The sharper founders will raise the standard on the things that matter:
Critical path work.
Customer-facing truth.
Core product behavior.
Canonical systems.
Structural fixes.
Institutional memory.
Anything that compounds.
Such things should stop receiving the “good enough for now” treatment when the real finish is within reach.
That is what boil the ocean should mean.
When something truly matters, stop normalizing the half-finished version of it.
A founder should still ask:
•Is this core or peripheral?
•Is this compounding or cosmetic?
•Is this the real fix or a distraction?
But once the answer is yes, the standard has to rise.
So do the real fix.
Close the loop.
Run the test.
Write the documentation.
Resolve the contradiction.
In an age of cheap generation, completeness becomes a strategic weapon.
It compounds trust.
Anyone can generate motion now.
Few can finish the thing that actually matters.
That is the new founder standard:
Finish the real thing.
Boil the ocean, but only the ocean that deserves boiling.
That is the hard part.
That is where judgment lives.
The machine can increasingly carry the labor. Which leaves the founder with a harder burden:
to decide what deserves completeness,
and to stop hiding bad prioritization inside unfinished work.
For a long time, “good enough” was the language of constraint.
In many domains, it is becoming the language of drift.
Stop leaking force through a thousand tolerated incompletions.
Boil the ocean.
New item in my SOUL md tonightView quoted post on X
The next Golden Age will be upon us when the means of growing, mining, manufacturing and moving physical things becomes fully divorced from human labor.
✊
✊
View quoted post on X