jryio

Designing flow state is really interesting. I've never been lucky enough to build with the intention of having other people enter flow.

8000 hours of Dota 2 had many hours of flow state as the player however..

Thanks! Glad you enjoyed it ~ there's flow state in many things, music, crafts, climbing... But I miss the flow state in programming

I would do the same for my children ~ However children have a special ability to revolt against any arbitrary constraints provided by parents, community, society. It differs person to person of course.

This is a play on words from this (excellent) NYTimes Opinion piece by Jasmine Sun [1] titled "Silicon Valley is bracing for the permanent underclass"

[1]": https://www.nytimes.com/2026/04/30/opinion/ai-labor-work-for...

(Gift article)

Correct, even though Ruby the language exists and predates Rails. What -- 80%+ of all Ruby code is Rails? Effectively the largest consumer and producer of the language which I think is a net benefit to your point.

Correct in that Ruby never had a schism and is still massively productive and wideley deployed (e.g. Shopify + Stripe alone represent billions/trillions of dollars through Ruby hotpaths).

Python's general lack of success in this domain is telling and embodies whats I was trying to communicate in the article -- languages with low entropy in syntax, features, ecosystem, and toolchain compound slowly.

Author here, wasn't expecting this piece of writing to show up on HN.

The specifics of Python were chosen only due to the language ecosystem being fragmented and inconsistent while Python remains an essential learning, research, and now ML programming language (it was my first language and I still love it).

My thoughts on LLM generated code have changed immensely in the last 9 months as I've taken on teams and projects through my consulting work [1] as a fractional CTO. Python remains a difficult, flakey, and inconsistent programming language for complex production systems. Most other programming languages suffer from fragmented toolchains and ecosystems: JavaScript (famously), PHP, and even C/C++ to a degree.

Languages with a single way to do things benefit the most: Ruby, Rust, Swift (even). Low entropy is the way to go and convention > configuration seems to pay off with LLMs.

Mean cost of management is more important than specific edge examples "X company run on Y language". I think that 'boring' languages with rock-solid compilers, toolchains, testing frameworks, and package managers make for high return on engineering time and production maintenance.

[1]: sancho.studio

Send me an email I'd love to make one

If you haven't already, read the book "The Second Kind of Impossible" by Paul Steinhardt

https://www.goodreads.com/book/show/35297608-the-second-kind...

It's a riveting account of years of research to discover Quasicrytals from theory, to experiments, to literally hunting in a meteor field in eastern Russia!

I am also saying the same thing. They are commenting that the flight of human capital was coming from abroad and is no longer.

However that's not what brain drain means. You would say "Iran had a brain drain in the 70s" not "America was brain draining Iran" makes no sense.

If you "drain" something the subject of the verb is what is being drained not where it is draining to.

A brain drain means the intelligent population emigrates to other countries.

The narrative and data do not support Americans going abroad.

I think you're referring to a lack of competitive education for those coming outside of America and choosing Europe / China to study.

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

I run a s business (small if you compare it to tech companies).

I can tell you the drag is between your own tools and the real world (which is very messy and inconsistent): taxes, compliance, payroll, amendments, share structures, etc.

Within my island, my books are in order, invoices and time keeping is fully automated, calendars and sales pipelines are connected.

I'm sure there are many businesses whose inner islands are not as orderly. The zillion tools out there all try to bring equanimity to the chaos and yet here we still are with fresh books, quickbooks, and xero...

I'm glad we went to space, truly. Racing the USSR might have been the wrong reason but it got us there. We've benefited immensely as a species from exploring the solar system and looking deep into the universe.

I'm not certain that racing China in AI is the right reason but it might get us... somewhere.

Who remembers the Newgrounds games making fun of Bush during the Iraq war?

In higher dimensional vector space, yes it can.

Dimensionality gets bizarre in 1000-D space. Similarity and orthogonality express themselves in strange ways and each dimension codes different semantic meaning.

Therefore, if the training data is highly consistent you are by definition reducing some complexity and/or encoding better similarity.

In Go the statement

    result, err := Storage.write(...)

    if err != nil { ... }

   try { Storage.write() } catch (error) { ... }

Previously in my life as an IC, I wrote a lot of Golang. I worked on the larger end to end encrypted video calling service.

I hated it. I was dreaming of Rust the entire time to release me from the hell of if err != nil dozens of time per day.

After hours with LLMs I've changed my tune. There have been 5 clients of mine (who have excellent engineering teams) but cannot get coherent results out of LLMs using python or Typescript.

I arrived back at Golang being a frustratingly simple, consistent, and low-thrash programming language which inadvertently made itself well represented in the training corpus [1].

My concession is that if you are going to write a median program (reading/writing files, network, db, etc.)...

Pick Golang especially if you've never used it. LLMs are extremely good at it, frustratingly so.

[1] https://jry.io/writing/use-boring-languages-with-llms/

I wrote about the meta thesis of programming languages in the training data here

https://jry.io/writing/use-boring-languages-with-llms/

I wrote about this here [1]

The big idea with LLMs is consistent references in the training corpus produced cheddar output by the language model during inference.

Go is an amazing language for language models because it's actually quite boring predictable while packing a lot of powerful distractions with a world class tool chain supported by Google and strong std library as well.

As a programmer I actually hated writing Go... and wanted to write Rust; but using coding agents makes me appreciate writing Go more.

I can get consistent results out while having concurrency cross compilation and predictability.

https://jry.io/writing/ai-makes-golang-one-of-the-best-langu...

Don't mistake age for durability. A new cast iron pan is durable. The point isn't in providence but in practice. Durable tools (new or old) show themselves immediately.

Reminds me of Zed's setting { "disable_ai": true } [1]

Glad it's an option be it for regulatory compliance, security, privacy, or any combination of the three.

[1] https://zed.dev/blog/disable-ai-features

They've been in development longer than the product has been public.

1-3 in stealth if I remember correctly

Zed is a durable piece of software, rather than the current trend of cheap disposable software. Regardless of whether humans or agents use a tool like this, durability is a benefit for both.

Congrats to the team

Do the work for us...

If anyone was wondering ... it's racist

Unsurprisingly the texts written up until that time were dominated by such individuals which is tragic for LLM training if you think about it.

The voiceless groups or fringe opinions which we take as normative today do not appear.

Does this encourage us to write in the present such that we influence the models in perpetuity?

OpenAI has contracted to purchase an incremental $250B of Azure services, and Microsoft will no longer have a right of first refusal to be OpenAI’s compute provider.

Azure is effectively OpenAI's personal compute cluster at this scale.

I find it strange that you've anthropomorphized Claude but not ChatGPT seemingly based on one having a human name and the other not

Exactly - cooperation is not incentivized properly

Just another disposable piece of software maintained by a single person that does 80% of what other apps do but worse.

Max lifespan 2 years

This feels like an unethical release of a model. They've opened a can of worms without investing in defense first.

Anthropic announced their capabilities in advanced, issued a private release, then put up $100M in credits to Fortune 500 companies and OSS projects to secure themselves.

OpenAI sees that, makes a model equally capable at exploiting vulnerabilities, then released it to the pubic with no equivalent program [1]

[1] https://www.anthropic.com/glasswing

Model performance at inference in a data center v.s. stripping thinking tokens are effectively the same.

Sure they didn't change the GPUs their running, or the quantization, but if valuable information is removed leading to models performing worse, performance was degraded.

In the same way uptime doesn't care about the incident cause... if you're down you're down no one cares that it was 'technically DNS'.