I am into AI cause I want to find out whether consciousness is emergent or always existing

This is the exact research process behind every investment I have made Going great so far

Every README has a quick start that is four commands long Then you hit the one that needs a system library from 2014 and the afternoon is gone My actual quick start is 90 minutes and a browser tab called "error"

If I can rebuild your product in a weekend, what exactly is your company? That question used to have a nice answer. Your company was the thing you had built, a few thousand hours a competitor would also have to spend. The wall was climbable but expensive enough that most people did not bother The wall got a lot shorter. So whatever a startup is now, it is not the artefact The annoying part is that this was always true and we just could not see it. There are dozens of note taking apps and the differences between them are tiny. The winner is almost never the technically best one, and everyone knows it and finds it slightly embarrassing What separated them was who heard about it first, who trusted it, and who kept going after the excitement wore off. The code was table stakes. It just did not feel that way, because writing the code was where all the effort went, and effort feels like it should mean something Defensibility lives in boring places now. An audience that took years to build. A product already wired into someone's billing and compliance sign off, where a better competitor is not worth the migration. Being the person who can get the meeting None of that is technical, which is sort of the whole point The old way to die was taking nine months to build the wrong thing. The new way is building the wrong thing in nine days, then a different wrong thing the week after, for a year, while feeling extremely productive The scarce thing in your week is not engineering hours anymore, it is contact with reality. So the Friday question is not how much you shipped, it is how many things you learned that you could not have learned by sitting there thinking My changelog has never looked better and I have never been less sure it means anything

Being genuinely into stuff is underrated I used to hide the random interests to make myself sound more serious Every good conversation I have had started from one of them No one bonds over a job title

This is the only changelog I have ever wanted to touch Mine get read by nobody, including me

A woman diagnosed as vegetative was asked to imagine playing tennis inside an fMRI scanner Her brain lit up the same motor regions a healthy volunteer's would, on command, every time That was 2006, Adrian Owen's lab, and the field mostly treated it as a fluke It wasn't A 2019 EEG study found 15 percent of clinically unresponsive brain injured patients responding to spoken commands with no visible movement at all A 2024 study across six centres put it at 25 percent It has a name now, cognitive motor dissociation Behaviour is just a bad instrument Locked in patients have full awareness and almost no output, and some patients who look awake by every bedside test have nothing measurable going on Roughly 40 percent of vegetative diagnoses turn out wrong once a specialist team looks properly So people built better instruments Perturb the brain, measure how complex the echo is, and you get a clean split: high awake, low in deep sleep and anesthesia, oddly high in REM when the body is paralysed and the person is dreaming I thought this would settle whether the brain generates consciousness or only reflects it It doesn't, and not because the machines need another decade Every one of those measurements is the instrument measuring itself A flat echo looks identical whether there is nothing there to catch or the catching machinery is offline Both theories agree the activity fluctuates, they disagree about what that means, and no number on a screen can arbitrate that The research still matters, covert consciousness detection changes what doctors do for real patients today I just went in expecting a wall I could push and came out having measured its exact position

Half of engineering is naming something badly and then living with it forever I have a variable called finalData2 in a side project that has survived three rewrites At this point it has tenure

Every model release now says it runs on consumer hardware Consumer apparently means a 5090 and your own power plant Meanwhile my laptop gpu sits there like a kid at the adults table

I never learned anything from a tutorial I didn't need Watched hours of them and remembered none of it Then one broken thing at 1am taught me more than the whole playlist Go find something to break

The AI demo is always one sentence and a finished app Mine is one sentence, then forty minutes of me explaining what I meant, then a login page nobody asked for

If AI can write the code, what exactly is my job? Fair question. Five years ago the answer was simple, you were the person who turned human language into machine language. A biological compiler The uncomfortable part is that the typing was never the important bit. It was just the bottleneck We have done this before. A hundred years ago an architect spent most of the week hand drafting blueprints with a ruler and a pencil. Then CAD showed up and a perfectly scaled wall took one click. CAD did not replace the architect, it deleted the drafting so the human could spend the whole day on the craft, designing buildings that stand up and also look good AI is doing that to software. It is abstracting away the manual drafting of syntax So what is left? The hard part of building software was never knowing how to write a for loop. The hard part is deciding what should be built, weighing the trade offs in a distributed system, understanding the person who is going to click your buttons, and reading what the model just handed you with enough taste to say yes this works but it is ugly We are not losing our jobs, we are getting promoted from implementers to decision makers. The ones who will struggle are the ones who only ever wanted to be typists Worth remembering that when Grace Hopper built the first compiler, plenty of people were sure it would automate programming away. It created an entire industry instead Anyway my taste and judgment are doing great, I just lost a twenty minute argument with a model about where one button should go

I started with Python So I have been doing Pomputer Science this whole time