You'd be hard pressed to do any really damage with numerical simulations outside of NaN propagations. C's problems come from the standard library

The argument is building anew is 90% cheaper.

I can’t see LLMs contributing to anything this size when they can’t remember what I said 4 messages ago

These tools do two things:

1. Rob one’s understanding of the design and implementation

2. Give the unqualified enough confidence to contribute

Sometimes the real deliverable is a happy team

Only to the extent that a happy team delivers something of value. Teams can be happy doing things that will drive the company bankrupt. there is only so much unhappiness they can stand

If the incentives of the team (team happiness) and the company (delivery of something of value) are misaligned, that's a much higher level failure. Either the engineering manager has been handed the wrong team or assigned the wrong task to accomplish. Setting high level goals for the engineering managers and allocating resources for them to achieve those goals is the entire purpose of senior leadership.

Not fun part is when team delivers good valuable stuff but market isn’t there.

What valuable stuff was delivered if the market isn’t there?

Whatever the sales department failed to create a market for.

It is always possible that dev team did all the code right, good product and still marketing failed or sales didn’t execute for plethora of reasons.

It may have been “good”, but I don’t see how it was “valuable”.

(I also think this is fairly common.)

Like OpenAI is losing money on each user - isn’t it valuable or is it?

If they close tomorrow because of lack of new funding would you say whatever they created wasn’t valuable?

Are you suggesting/implying that OpenAI lacks product-market fit?

I think they found a market for their product, making what they did valuable. That can still lose money for a time, and even maybe run out of money, but what they have is valuable to my mind.

Even outside the context of capitalism, society only functions when people perform unpleasant tasks which provide value. Nobody has fun collecting garbage, but it has to be done. Nobody finds happiness in digging graves, but it has to be done. Nobody is overjoyed at the prospect of telling a person they have a terminal illness, but it has to be done.

So no, I don't really buy the idea that a team which is "forced" to work on boring, but profitable tasks for a business instead of getting to rewrite core infrastructure in Rust as a fun and interesting intellectual exercise is equivalent to slavery.

> Nobody has fun collecting garbage, but it has to be done. Nobody finds happiness in digging graves, but it has to be done. Nobody is overjoyed at the prospect of telling a person they have a terminal illness, but it has to be done.

but we find ways to make the jobs of people who perform these tasks less horrific. We makes trucks that reduce the physical toll and increase the cleanliness of garbage pickup, we combine the digging of graves with the maintenance and beautification once they have been buried, sometimes you have to tell someone they have a terminal illness, but the majority of the time you are helping someone get their condition into remission.

I think the critical balance that management has to achieve between "having a happy team" and "having a productive team" is finding ways to keep morale up so that employees don't lose their minds and quit or reduce performance doing the miserable stuff.

I wasn't referring to slavery. There are a number of cases where people are unhappy with their job, but go down the road to a competitor and everyone is happy. The industry/product are essentially the same, yet the happiness level is very different.

Why misrepresent what someone else said to make your point?

So I'm guessing most people are downvoting this as a knee jerk reaction to the comparison with slavery, but I think the core point is quite valid.

At some point, if people are unhappy working towards some goal, you gotta re-evaluate if the goal is worthy. I consistently meet people in other industries who really enjoy their job, whereas in tech, most of the people I know consider their job to be one of the lowlights of their life. And I don't think it's a stretch to say many, many tech jobs are not serving a worthy goal.

So it's disappointing to see people who can't look past "but business value bro", as if we got where we are because capitalism is some holy, inevitable universal law.

Things really felt great 2023-2024

Django also missed a couple fingers ;)

I’m writing my own internal combustion engine simulator with some applied CFD. Nothing public yet but the sound is nice in my opinion:

https://glouw.com/2025/10/12/Ensim4.html

The demo with an inline 8 at 16000 RPM is hard to judge, because I've never heard such an engine IRL. Might I suggest adding demos of engines people know the sound of?

FWIW, that sounds exactly like how I'd imagine a 1.5L I-8 would sound on an equally imaginary test stand. Like a slightly throatier dentist drill.

What a neat tool!

I believe MIT Scheme compiled directly to C by intermingling switch statements and gotos to emulate its call stack. Problem was, as programs grew, compile times weren't linear. I gave it a shot once, it was a somewhat spiritual experience:

https://github.com/glouw/switch

Thanks! Do you mean MIT Scheme's C backend? I've used MIT Scheme on and off for a long time and have never touched the C backend & have no idea how it works, so this is interesting.

(MIT Scheme also has a native code compiler for Intel CPUs, which seems to be what most users of MIT Scheme (an admittedly small community) actually use.)

If you'd like to define the backend like this, it's the easiest way to compile to C; we're just repurposing C as a general-purpose assembler: https://glouw.com/2023/11/07/Switch.html

I believe MIT-scheme took it a step further with gnu extensions which allowed you to take the address of a label like &&label, which allowed for function pointers: https://gcc.gnu.org/onlinedocs/gcc/Labels-as-Values.html

I don't see anything in the manual about MIT scheme compiling to C, just to its own native code object files and some portable bytecode...

https://www.gnu.org/software/mit-scheme/documentation/stable...

It's there, but the documentation is weak.

The release notes from long ago when that back end was released are here, and give some detail:

https://share.google/xmUnDhC7lndujD7TI

I think a very rudimentary implementation (which is easy to understand) is in the SICP book

Even MIT Scheme was used with SICP, the Scheme implementation in the book is different from MIT Scheme.

MIT Scheme was for a long period one of the leading Scheme implementations. It lacks support for Apple Silicon, so it is not as popular now, as it once was.

https://www.gnu.org/software/mit-scheme/

Yes. In Apple the memory can be writable XOR executable. MIT scheme needs to write on pages that will execute, so it will trigger the MMU..

Install chicken scheme. Then:

       sudo chicken-install srfi-203

       sudo chicken-install srfi-216

At ~/.csirc put:

    (import scheme)
    (import (srfi 203))
    (import (srfi 216))

Now you can do nearly all the execises.

I mean, neither is gonna generate money

This is funny, because my personal view is that AI’s biggest pitfall is that it allows the unqualified to build what they think they’re qualified for

Yes, learning and creating should be reserved only for the licensed!

It is in many fields. They're Professional Engineers, and they build nearly everything that's important to a society. Software is overrun by literally anyone and everyone because it's not a protected field of engineering

Hang on, my understanding of the situation is that only PEs are allowed to sign off on the design of a thing. They don't have to design it, and they almost certainly don't build it. Am I wrong?

The stamp is legally binding. The company you work for isn't gonna protect you financially if the bridge you stamped falls. This is how they protect their profession

https://en.wikipedia.org/wiki/List_of_data_breaches