Yeah it turns out the same skills that uncovered Theranos don’t translate well at all to identifying Satoshi. Satoshi and Back’s code don’t look at all similar and even starkly different.

Datasenter la ned driften – nå får innbyggerne sjokkregning Hadsel kommune har i flere år forsøkt å kvitte seg med et datasenter – og denne uken ble driften avsluttet. Det de ikke tenkte på var at nettleien for innbyggerne ville skyte i været.

In Norway it’s the opposite problem. When data centers leave, energy bills go up for everyone else.

www.nrk.no/nordland/dat...

Adam Back and Satoshi have very different open source coding styles. Something the NYT missed.

Their English might be comparable but code looks nothing alike.

You’re giving too much credit to the mental faculties of a soggy taco.

Scooby doo logic is now NYT level reporting

By this logic the identity of Satoshi is clearly Phil Zimmermann

My company would out accelerate yours simply because you wait for gaming journalists to tell you how tools works

If you’re getting 30% wrong results at this point you’ve got a skill issue. We’re not running gpt 3 anymore, so you’d be firing people for made up reasons and probably running a failing company

“We fired the guy who did it in 2 minutes agentically and we are waiting for the other guy to finish organizing it all manually” - the smartest manager

do you think my job is to be a typist (if yes, you admit to not knowing what a programmers purpose is) and also do you think you can type a program to do this inbox parsing faster than an LLM (if yes, prove it.)

This wasn't hypothetical. I actually did this. It took 2 minutes. Every suggestion you've given would be time wasted on a frivolous task an agent can do.

If you think you can do it faster, and better, then prove it. Build demo on something like replit.com where I can run the code against my inbox.

I guess another way to ask this, what benefit do I get from doing what an agent would do, but slower?

Which pre-built library does what I asked faster than me asking an agent to do it?

If none, am I not then composing a bunch of libraries together, and/or manually fetching the documents so a library can be used to read them?

You simultaneously have made up that I have or work for a company (I don’t) and have no answer for why I should spend time handwriting code instead of just having fancy autocomplete do it for me. Your answers are just “you should just do the work manually because”

If I was hiring one of two engineering candidates, and asked them how they would complete the task above, do you think I should hire the one that will:
- Use chrome-cdp to agentically build this ETL
- Or the one that will manually download files and handwrite code to build this ETL

If so, then this is where we fundamentally disagree on what the role and purpose of a programmer is.

Hence, *taps the sign*
bsky.app/profile/ngmi...

Do you believe that the job of a programmer is to manually download those files and then write the code to present them in a certain way? And that if you automate that task, you're no longer doing "your job"?

My question is what is the optimal path, given where the docs are now, and given the tools we have available.

Should I sift through emails to find docs to download and write the code by hand, or have an agent do it in 2 minutes?

GitHub - pasky/chrome-cdp-skill: Give your AI agent access to your live Chrome session — works out of the box, connects to tabs you already have open Give your AI agent access to your live Chrome session — works out of the box, connects to tabs you already have open - pasky/chrome-cdp-skill

Which do you think is more worth my time as an engineer… handwriting the code to do this or having an agent do it with github.com/pasky/chrome...

“Open up mail.google.com/mail/u/0/#inbox. search all emails from Jan 1 2026 to April 1 2026 that have anything to do with taxes, tax forms, K-1, 1099, W-2, etc. Compile a table with unique URLs, sender, and doc type. Open it as a formatted HTML file in the browser.”

“I don't need AI to tell me how to build them”

Do you think engineers are in chatgpt asking how to build stuff or do you think they’re extending pi to have a custom agentic harness for their stack

Idk if you've used Google in the past decade, but if so you have interacted with AI.

Google Search started being AI-driven in 2015. Google Translate was the first Transformer product used at scale. Youtube's recommendation engine is AI. Google Maps ETA/routing is AI.

God, why are all you tech people like this?

And can you please, please fucking stop??

Don’t have to trust me.

The job is not to be a typist, so every good engineer is figuring out how to spend as little time typing as possible and do the real work.

The engineers who built a lot of underlying code widely distributed on the Internet are all doing this because it’s dumb not to.

*Taps sign*
bsky.app/profile/ngmi...

A lot of people (esp on bsky) think AI is sucking up all the worlds fresh water but almost everywhere you look you'll find millions of user-years worth of AI usage in water terms being spent on innocuous everyday things that people have zero moral panic over (or leaked out of pipes going to nothing)

I think he also doesn’t understand that AI is global and pausing construction in the US does nothing if I can still use LLMs from China

Should note this is only a moratorium on data center construction in the US. Since AI is digital and can be accessed from anywhere on the planet by anyone on the planet, this is basically just a way to prevent US AI companies from giving away AI in favor of AI from other countries.

where was the outrage when the air brake took the jobs of all the brakemen on trains