Bernie

For the love of the game | Founding Engineer at @alloyrobotics

[Bernie @berniecroll]

You can navigate the hyperdimensional space of life to get exactly what you want.

The chances of your success is 0.

But that’s the same as any model with randomised initial weights.

The secret is non stop iteration/feedback and you can effectively backpropagate yourself.

But the crazy thing is you don’t have to start at randomised weights. We can leverage off the shoulders of giants who have already done much of this learning for us.

[Founder Mode @Founder_Mode_]

Elon Musk was asked why his companies move faster than anyone else.

His answer:

"I'm constantly addressing the limiting factor. Whatever the limiting factor is on speed, I'm going to tackle that. If capital is the limiting factor, I'll solve for capital. If it's not the limiting factor, I'll solve for something else."

He then said something most managers never figure out:

"If something is going really well and making good progress, there's no point in me spending time on it."

"The irony is if something's going really well, they don't see much of me. But if something is the limiting factor, they'll see a lot of me."

He spends his time entirely on whatever is blocking the next step.

Not on what's interesting. Not on what he's best at. But on whatever is the bottleneck right now.

Most leaders do the opposite... They gravitate toward what they're comfortable with and away from the hard problem.

From: @dwarkesh_sp and @collision

[Bernie @berniecroll]

codex plugins is going the way of the app store

[Bernie @berniecroll]

@LiebermanAustin Investing in yourself as a young person has the highest return rate.

[Bernie @berniecroll]

@LiebermanAustin but doesn't consider a person's increased productivity from coffee over 20 years. what's that worth $$$

[Bernie @berniecroll]

@dara_venture the game is feedback and loving feedback.

why do you need a high pain tolerance when you already enjoy it

[Bernie @berniecroll]

@dara_venture if you have silence from customers and investors (even day 1) then you have a real problem

the best founders seek feedback AS FAST AS POSSIBLE

[Bernie @berniecroll]

if you're not wrong at least 30% of the time you're not making enough decisions

[Race 🕊️ @multiplanet1]

Elon Musk was asked by an interviewer what he does when he makes a wrong decision.

He said he doesn't make decisions based on being right. He makes decisions based on expected value. Which means he expects to be wrong frequently. He said if you're not wrong at least 30% of the time you're not making enough decisions.

He described SpaceX's approach to rocket design. Instead of spending years perfecting a design on paper and then building it, they build it fast, launch it, watch it explode, and use the explosion as data for the next version. The explosions aren't failures. They're the fastest way to generate the information needed for success.

He said most companies fail not because they make bad decisions but because they make too few decisions. They spend so long trying to be right that the opportunity passes. A wrong decision made quickly generates data. A delayed decision generates nothing.

Then he said the part that made the interviewer go quiet. He said the biggest mistakes of his career weren't the times he was wrong. They were the times he was right but too slow. The times he knew the answer and waited for more confirmation before acting. The delay cost more than any wrong answer ever did.

The world sees Elon as a risk taker. He sees himself as someone who moves fast enough that the cost of being wrong is always lower than the cost of being slow. That's not the same thing as courage. It's math.

[Bernie @berniecroll]

- cut requirements like its a disease
- shorten deadlines to 1 day
- cultivate anti big corp thinking

[Codie Sanchez @Codie_Sanchez]

I’m convinced 90% of a founder's job is to just do 3 things:

- be delusional in your optimism
- push everyone to move faster
- make it crystal clear what to work on

[Bernie @berniecroll]

you're getting dumber because you aren't working at a higher abstraction
- you are the bottleneck, engineer yourself out
- go into yolo mode and build guardrails if you dont trust
- plan meticulously and give the agent goals, directions, north stars so not every decision goes back to you
- plan with UI not text, plan.md is too much mental load, build decisions, mermaid diagrams, sequencing charts
- plan so agents can parallelise work, a migration should not be sequential
- if an agent makes a mistake its because it doesn't have enough context, think senior with no direction
- think how can I scale to 10 or 20 agents
- less touchpoints, have 1 or 2 agents surfacing crucial decisions back to you and propagate down
- don't be the approval monkey

[Maximilian @MaximilianMoj]

I'm 33 and I think Claude Code is melting my brain.

For 6 months straight I've had 5-6 terminals open at once, waiting on responses just to smash "enter" 90% of the time. That's the whole job now.

And it's doing something to me. A few friends and I keep circling back to the same thing in conversations: none of us feel as sharp as we used to.

Maybe it's just us. But I keep wondering how many other people in their 30s feel it too.

(And yeah: this is a me problem, how I lean on the tool, not the tool itself. Doesn't make the effect any less real.)

[Joe Harris @_joe_harris_]

ridiculous unfair advantage in robotics if you have made your robot+company data accessible to agents

[Bernie @berniecroll]

Frontier models aren’t getting that much better at decision making. I feel it’s a gap in data that’s inaccessible. Either will need to let it “play” at life like AlphaGo or inject in real world situations to RL like Claude Tag.

Making good decisions is the hardest skill to learn

[Bernie @berniecroll]

Intelligence = effectively search through high dimensional space

[Bernie @berniecroll]

you don't listen to your customers you understand them

[Bernie @berniecroll]

people talking aloud to themselves is just LLM thinking

[Bernie @berniecroll]

I dont want to learn your UI, where is the AI tab?

[Bernie @berniecroll]

people aren't stupid with math

it's because the language of math is taught terribly
or not taught at all

[Bernie @berniecroll]

Not going to be easy when they out number us 10-1

[Bernie @berniecroll]

@chesterzelaya 🙌

[Bernie @berniecroll]

A tale of two internets

[Zeno Rocha @zenorocha]

We started with the docs. Then pricing.

Now, the whole site renders as Markdown.

Blogs, changelogs, handbooks, customer stories, security, and even legal pages.

• Humans get HTML.
• Agents get Markdown.

Example: https://t.co/7wlqFtTSAc

[Bernie @berniecroll]

it takes hours to debug your robot, what if it could take minutes?

now extrapolate that to a fleet of hundreds of robots

you see the value? https://t.co/TIrAMS4x4r

[Marwa ElDiwiny @MarwaEldiwiny]

I'm curious what happened here..