> posted to a decentralized social network built on self-authenticating merkle search trees that uses nonfungible tokens (DIDs) for identity
yeah whatever happened to the blockchain
Might be able to catch me at hereticon
it’s almost like git and most orthodox blockchains share the same underlying data structures
hypercore and ssb mentioned
bsky.app/profile/ngmi...
Earlier you said: If you have a LLM write your code, your personal skill level becomes much less important to the task
You're demonstrating here that's not the case. You have to be skilled enough to spec out that, in this case, the constraint is object transport, not your travel distance.
Yes, you will have to spend time reviewing code.
You will also have to articulate what problem you're solving and for whom.
You will also have to make decisions on how to solve those problems given constraints and make trade-offs.
The LLM can type the brackish residue that glues that together.
You can be a world class programmer and still have LLMs write your code. The job isn't to be a typist. To think so would be trivializing the skillsets and purpose of a programmer.
The people *demand* all code is handwritten by the bsky eng team
Back then it was like “are you a psychopath if you’re not on Facebook?”
Reread what I wrote but this time knowing I used the same idiom
That's the goal.
It's not to be at a desk at nine in the morning and come back to reality at 7pm exhausted, doing soulless work being a typist for somebody else.
You should literally be able to hit the bricks, get into a flow, and have your intellectual capabilities leveraged through computers.
This aligns with my goal of not having a job
I don't think the entire job is review but that's def where laborious limiting factors are becoming more apparent.
You also need to articulate what problem you're solving and for whom, and get the agent on the best path for that.
We're just in an era now that requires rethinking the artifact chain
Why does anyone post-punch-card computing make $125k+? Didn’t abstracting that away cut massive overhead cost?
I think besides docs and specs my code is 99.9% LLM generated.
There may be an instance where a small tweak is done with a faster feedback loop by just manually writing in the text editor, but it’s becoming exceedingly rare that I hand write any code. I even have Claude or Cursor co-sign my commits
According to Stripe, which processed $1.9 trillion in payments last year… it’s turning out well enough for Stripe to buy stablecoin and wallet infra + build their own blockchain
?
Clearly I’m not talking about Sam Altman.
Do you think Sam invented all those things that make his company possible?
Big tech of course has the economic incentive to scale and industrialize backprop, CNNs, transformers, RLHF etc, no disagreement there, but I’m saying higher ed isn’t giving into all that tech… they invented it.
Same with TCP/IP, tech has incentive to scale the internet, but they didn’t invent it.
Also the good thing about those pinned papers is it doesn’t require being on the same page. The goal is to make narrow, falsifiable models of our ideas of the world. If those models are wrong, they can be refuted with better models and we get the benefit of better explanations.
Disagree with… the majority of AI coming out of decades of higher ed research and investment?
Here is the 1986 paper on backpropagation, which enabled all deep neural net training, from UC San Diego and Carnegie-Mellon
www.nature.com/articles/323...
Majority of AI was created by higher ed.
They did decades of foundational science, published the research, created the talent pipeline to AI companies, committed billions of institutional investment, etc.
Not so much giving into it as much as they are harvesting fruit from the seeds they planted.
Why is Yale using AI with googles gemma model instead of fighting against it?
medicine.yale.edu/news-article...
The internet is filled with misinformation, crime, and gov’t surveillance… why aren’t the universities that invented TCP/IP funded by DARPA rallying against it?
Higher ed is behind nearly all the foundational science, published research, the talent pipeline to AI companies, and billions of dollars of institutional investment contribution to AI…
Why would they form a front against their own creation?
FYI
And they lost to USA at their only sport
Not only does Canada have a lower GDP per capita than Alabama, but Canada also has fewer MRI units than Pennsylvania
The data center
Someone did the math comparing the xAI gpu supercluster to In-N-Out’s and burger consumption. Colossus 2 uses about 2.5 In -N-Outs worth of water annually.
newsletter.semianalysis.com/p/from-token...
