cushychicken

I wouldn’t describe the people this blog talks about as “normal people”, or “average people”.

Most people don’t behave this way or have these kinds of problems.

A reminder: tech jobs are their own kind of filter bubble. It’s worth taking a step out of this space every now and again to see what other people are thinking and worrying about.

It’s one of the things I love about my rec hockey games. People there have much more tangible worries: when they’re gonna fix the dock at their vacation homes, if the girl they texted will text back, when will their kid grow out of their threenager phase, etc.

Most of the world is not nearly as preoccupied with AI as this community is. It’s refreshing to see when you can access it.

This is a real problem.

I suspect that part of it is that people don’t have enough time to mentally incorporate the fix.

Is it weird to submit the MR later, after people have had a chance to digest the issues?

Yes, if they are "responsible" for the code delivered, where responsible means they understand the code, the architecture, the decisions made, etc.

This is a good principle to maintain, I think.

I'm not a professor, but I manage a team of about a dozen people. The maxim I have is: "You're responsible for anything that hits git."

Don't care if the LLM generated it, or the LLM told you if it's a good idea. If you commit it, you are endorsing it as a good idea - so you're the one I'm going to ask about it. I see the same principle at work in your pedagogy.

I'm not anti-AI (and really, what could I do if I were?) since I use it myself, I'm just anti-slop, especially from my students.

This hits. Especially this part:

and really, what could I do if I were?

My completely unsolicited opinion: you're doing a responsible thing by teaching these students how to use AI as a reference, and keeping them honest about not using it as a substitute for their own critical thinking.

Uber was too expensive once, too, and yet they seem to be currently functioning as a profitable public company.

“AI is too expensive right now” is an accurate title. Plenty of things could change in the future to change that. Off the top of my head:

* end user pricing

* breakthroughs in model efficiency

* better chips to run models

Any one of these things is easily possible in the next three years. Probably sooner.

Personally I hate how badly internal users are served by the majority of their systems and am willing to take some calculated long-term governance risks

This, I think, is the LLM/vibe coded app’s current place to shine.

Most internal systems don’t need massive concurrency or redundancy. It’s a webapp that reduces coordination cost between 20ish people. That’s something you can typically vibe code and deploy for ten bucks a month, and create real value.

Zero of these instances were due to shipboard reactor failure.

The engineering side of running reactors safely is a solved problem, the US navy has > 7500 reactor-years with a perfect safety record.

It’s also worth noting that the US Navy is the only organization with a perfect nuclear safety record.

My point being: by god, let the Navy nukes train everyone else!

I was one of probably eight people who played the Emperor: Battle for Dune RTS game, and I always think of the Fremen character sound bite whenever I see the Old Man of the Desert’s true name invoked:

”…for Shai-Hulud!!!”

geohot definitely ticks the box for “so ambitious he occasionally sounds unhinged”.

Don't forget Sauron.

https://www.sauron.systems/

One other commenter asked a decent question - does going lighter (Zig) or harder on memory safety (Rust) confer any meaningful advantages against the phenomenon you describe?

Repeatability and/or an actual negative effect.

POC generally means “you can demonstrate unintentional behavior”.

“Exploit” means you can gain access or do something malicious.

It’s a fine line. Author’s point is that the LLM was able to demonstrate some malfeasance, not just unintended consequence. That’s a big deal considering that actual malicious intent generally requires more knowhow than raw POC.

Fair critique. Mueller was a pretty upstanding example of how to run the FBI, however.

It's equal opportunity corruption.

...and look how nice it sounds it live in Russia.

If they can build a peaceful relationship with Taiwan without military involvement where both countries can continue to prosper we really will have a new super power

Ah, if only.

Those damn intransigent Taiwanese!

It’s almost as if they don’t want to join the PRC.

…like most other independent nations.

The last 15 years has significantly changed peoples' opinions on that matter.

I’m gonna need to see some immigration statistics on influx of foreigners into the PRC to believe that claim.

A few weeks ago my friend Aadil and I were at Whole Foods buying a birthday cake for a friend. We wanted to write something clever on the cake but couldn’t really think of anything. We stood around thinking for a few minutes before Aadil said "Let's just say a bunch of bad ideas out loud so we can get to the good ones." And it worked!

It's a well known creative / brainstorming trick that the best way to have a lot of good ideas is to have a lot of ideas.

Focus on genesis decoupled from critique, then critique later.

Not sure what you mean, but I’d never heard of Sarah Paine before that. I thought she gave a very concise yet nuanced view of the modern world order in her lectures for Dwarkesh.

Love the interviews Dwarkesh sponsored with Sarah Paine from the Naval War College.

Also, somewhat spitefully, find it funny that he has multiple roommates.

it's usually coding something more precise in a sort of plausible deniability.

Yep. I'm a director now. This is exactly how it is. A big part of being effective in this role is understanding how direct you can be in a given scenario.

A senior manager on reviewing a proposal asks them to synergize with existing efforts: Your work is redundant you're wasting your time.

Option 1 is how I'd say it to a peer whose org is duplicating effort. You can give your advice, but at the end of the day: not my circus, not my clowns.

Option 2 is a more-direct way of how I'd say it to someone in my own org. I'd rephrase to: "Someone else is already doing this; focus your efforts on something more impactful."

Mickens is a rare combination of bright, engaging, and absolutely hysterical.

I hope I get to meet him someday.

They're good advise, but can be hard to execute on for most people.

OP gave the thread a very good and valid suggestion. Treating this as a societal problem - for "society" to solve - is lazy thinking.

If you want something you've never had, you have to do something you've never done.

Sure. http://cushychicken.github.io

It's only dumb if it doesn't work.

If it works... well, congratulations. You now have an edge that no one else knows about.

I think there’s a lot of people out there who don’t want to believe written communication skills like these are as important as raw technical skill.

It’s the “embedded” part that people struggle with.

I can find systemd script gremlins all the live long day.

I can’t find anyone who can write device drivers for custom peripherals, then hook them to user space utilities in a sane way.

And embedded Linux!

I’ve had a hell of a time finding good embedded Linux devs.

I got insanely lucky to hire two this year.

Didn’t everyone kind of migrate to more capable chips like ESP32 and STM32 in the intervening decade since Arduino got big and commercial?

I for one really like developing and co-managing specs with an LLM.

I think it’s a great conversational tool for evaluating and shaking out weak points in a design at an early phase.

The difference is purely in how much spec you write before you start implementing.

Ah, and therein lies the problem.

I’ve seen companies frequently elect “none at all” as the right amount of spec to write.

I’d rather have far too many specs than none.