elcritch

It’s sad to me after reading stories like this how much Western societies have generally lost the pursuit of Bildung (wisdom, virtue) that shaped men like Wilhelm von Humboldt's.

I’ve come to firmly believe that a society which looses the pursuit of the noble in the individual in exchange for mere technological excellence will in the long run loose technical ability as well.

It’s not even about a religious ideals, but secular ones as well. Really any belief system that puts the onus on individuals to pursue nobleness, as opposed to the more empty “reforming society” or “fixing society” that universities in particular appear fixated on currently.

Reminds me of CS Lewis’ critiques our modern age in “Abolition of Man” who critiqued it very well.

Of course we see this tension of education of the individual vs mass indoctrination in most western public school systems. TFA was a good read, as I’d always assumed and read that universal education was created as just a tool for mass indoctrination for obedient workers.

It’s like a slice of the earlier web. Realized I haven’t seen a c|net logo around for a long time.

Unfortunately doesn’t answer the question “successful LLM project”.

From what I’ve seen of GitHub and AWS this year the answer is no. That’s despite me being bullish on LLMs and finding them highly productive.

It really doesn’t seem very hard to have a small invalidation list. Just a redis cache or a simple broadcaster, etc.

Does anyone have an example of how they built a JWT revocation service?

Yeah I’ve been using gpt-5.3-codex-spark in Codex lately and it can be surprisingly good and it’s super fast. However it needs more explicit instructions.

The surrounding Arab states already ethnically cleansed the indigenous Jews from the majority of the Middle East.

They have a unique design where only a turbine moves with very cheap electricity produced.

AI inference gives them a great way to turn revenue from cheap electricity without solving the device-to-shore problem yet for that energy.

A startup I worked at is building wave-powered data centers [1].

1: https://www.latitudemedia.com/news/catalyst-building-inferen...

Not to mention the “largest transfer of wealth” [1] to the ultra rich during covid times.

1: https://resources.finalsite.net/images/v1746565358/d11org/oq...

Well it was already ready in pi.

Technically, yes. Obviously stretches the normal definition of a server, but it is a computer controlled system "serving" a mostly digital signal.

https://www.baesystems.com/en-us/article/gps-iii-satellite-l...

GPS..

Perhaps but the AI drone based WW3 might put nvidia in the black before that.

> Globally, the poorest 50% emit roughly 4.4 billion tons to 4.7 billion tons equivalent annually, accounting for about 11% to 12% of total global greenhouse gas emissions.

The “less-developed” bottom half of the world population still produces about a tenth of the annual co2 production annually.

That means instead of potential climate catastrophe in 10 years it’d take 100 years if the top half disappeared tomorrow. Obviously an overly simplistic argument but its meant to show that the problem would just be slower but not gone if we got rid of the “wealthy”.

Now I don’t believe they’re equally as culpable. Yet I also firmly believe the vast majority would choose the same as us in developed world have if they could.

Ultimately that’s more of my point. What’s happening isn’t due to some evil plot by the ultra wealthy. It’s the result of human nature.

Some unscrupulous ultra wealthy might hasten it by a few years or a decade, but the core problem is human nature and an abundance of fossil fuels.

The “wealthy” are also the most likely to prevent catastrophe by developing renewables, etc.

I didn't say we're all equally culpable. We're not. Yet en masse we're all guilty to some degree.

There's only what 10's of thousands Taylor Swifts in the world. Yet there's billions of everyone else. The majority of greenhouse gasses likely come from the aggregate of everyone.

See I could agree with the first part. But then you add We have broken our world for the greed of a few. After that I sort of understand why so many folks reject the former – they're rejecting the empty moralizing.

If you truly believe climate change is real then also admit that "We all have broken the world", except perhaps some uncontacted peoples in the Amazon.

Anyone who has ridden in an automobile, a train, a plane, a powered boat has contributed. Anyone who has used or purchased goods transported with any of the above has as well. Anyone who's eaten crops grown with large amounts of industrial fertilizers has contributed (e.g. most of the world).

The oil companies just produce what everyone in the world wants and wants cheap.

Who cares about tokens?

I'm wanting to build pieces of software that I've been wanting and often working on for years. These new models are making it possible for me to scale my work to build it.

I don’t know, GPT-5.5 has been very effective for me. It’s not perfect but the quality of refactoring it can do is awesome.

Previous models both GPT and Claude would struggle with the larger picture more. Pretty quickly they’d do one off hacks. Eventually they’d code themselves into a wall if you weren’t careful.

Haven’t hit that wall with GPT-5.5 yet. New changes or improvements on a GUI library I’m building seem to be constant in time per feature.

Though I’m talking only 10k’s of LOC. Also I’m using Nim which is both strongly typed and concise.

These sorts of things did happen before the internet though. Think of the Cultural Revolution in China started by revolutionary university students.

Mass printed pamphlets was the original meme. The more things change the more they stay the same.

Araq the Nim creator is working on a rewrite of the Nim compiler called Nimony that'll become Nim 3.

It's making fast progress and will provide the basis for a proper LSP! Nimony already supports incremental compilation and parallel which are key pieces for good fast LSP.

Additionally running GPT-5.5 on medium sometimes gives me better results than high mode. On any of them I still have to push the models in the right direction.

Exactly, Constantine generates assembly output and links that into normal Nim objects (e.g. C). That can then used in Nim or in Rust, Go, etc.

From its "Why Nim" in the readme:

- Assembly support either inline or a simple {.compile: "myasm.S".} away

- No GC if no GC-ed types are used (automatic memory management is set at the type level and optimized for latency/soft-realtime by default and can be totally deactivated).

- Procedural macros working directly on AST to create generic curve configuration, derive constants write a size-independent inline assembly code generator

Nim also has top notch meta programming, probably more so than Zig. You can easily do loop unrolling, specialization, etc. For example Constantine, which is a constant time crypto library that outperforms C, etc.

To me programming Rust feels so limiting due to lack of good compile time meta programming with types. That’s the key.

Yesterday I did some testing on the cost to solve the same simple problem on openrouter with different models using cline. Simple problem but it had a few nuances to solve it properly and so required reasoning.

After reading comments like this I was expecting (hoping?) that DeepSeek or similar would be cheaper.

However I was surprised that DeepSeek v4 cost about 5.5x GPT-5.4 to solve the problem.

- Deepseek-v4-pro-medium cost $2.47 - GPT-5.4-medium cost $0.45 - GPT-5.5-low was $0.86

Trump has had how many assassination attempts now? The attempt in Pennsylvania killed an innocent bystander, a father, in the crowd.

Charlie Kirk was called Nazi by many and eventually shot dead in front of his children.

So yes, unfortunately.

Ugh FreeBSD is so much nicer than modern Linux. It's hard not to love.

For me it's all about Nim + LLMs. I'm greedy and want both fast-to-ship and fast-to-run? Readability comparable to Python but with strict static typing that LLMs can't "cheat".

I actually (mostly) enjoy reading the code that the LLMs create in Nim. It's quick to read and look for refactor or cleanups. Compile times in seconds so the LLMs is usually the slow piece. It's fun and productive. With Python + LLMs I'm seeing them just create ever more layers of unmanageable cruft.

Recently I wanted "magic" behavior to get OpenAPI types and swagger.json along with auto parsing my rest APIs for me. I had Codex make a library for me using compile time reflection and a sprinkling of macros. Done, simple.

Yeah but still way too granular.

There’s 50 US states. Many are very different.

It’d be even better if it was city based.

I primarily use it for browsing memes now, and occasionally interaction with friends.

The idea of having a non-crap Siri on my phone that I could interrogate directly would be amazing.

My ADHD brain would love to do this stuff:

"Hey AI, how much is my electric bill this month?" and "Okay thats high. Pay it but remind me next week to order a new AC after researching options for me."