low_tech_punk

If you click on the "create link" button multiple times, the menu would pile up vertically.

I believe Wardgard has zero external dependencies. The three packages it depends on are all utils written by the author himself.

https://code.haverbeke.berlin/wordgard/wordgard/src/branch/m...

all the emphasis on cyber security. feels like a reaction to anthropic, not a real next generation.

When mixing languages, why does the English have Chinese accent and Chinese have English accent? Is it a feature or bug?

I hope this puts more pressure to extend the software life-span.

Saw the DOOM raycast demo at bottom of page.

Can't wait for someone to build a DOOM that runs entirely on GPU!

When I visit https://www.molt.bot/ with Edge browser, there is a bloody red screen screaming malware. What's wrong with the name?

User: Hey AI, turn off my Mac

AI: (after 10 minutes of deep reasoning) Ah! I found the solution. I need to cause a power surge in the grid to power off your device.

Google's ML team really nailed it. There are lots hand/head/eye tracking demos made with mediapipe.

I recommend checking out The Poet Engineer: https://x.com/poetengineer__

No AI is appealing but there is the cliff problem. If there is one small thing the mini language can't handle, the user would have no chance solving it themselves. They might as well start with an LLM solution first.

One workaround is that when there is syntax error, let user optionally switch to LLM?

Is it possible that the coffee drinkers have more social interaction with the barista and others? It's unclear from the paper if they eliminated the confounding factors around coffee drinking.

I believe the entire web app was a mod on top of monaco (vscode)? Really great hack!

A slingshot, A blind fold, A pocket -- Steve Jobs never said that.

Everyone asks, what does the fox say, no one cares how the fox feels. Thank you Mozilla, now we know the fox feels like open source.

IMHO, parallel tasks across multiple repos is not as useful as parallel tasks in one repo.

Not all is lost: AI can still be acronym for Apple Intelligence.

I had a similar thought, feature not bug.

The nature of programming might have to shift to embrace the material property of LLM. It could become a more interpretative, social, and discovery-based activity. Maybe that's what "vibe coding" would eventually become.

Most programmers don't understand the low level assembly or machine code. High level language becomes the layer where human comprehension and collaboration happens.

LLM is pushing that layer towards natural language and spec-driven development. The only *big* difference is that high level programming languages are still deterministic but natural language is not.

I'm guessing we've reached an irreducible point where the amount of information needed specify the behavior of a program is nearly optimally represented in programming languages after decades of evolution. More abstraction into the natural language realm would make it lossy. And less abstraction down to the low level code would make it verbose.

This piece reminds me of another article musing on the necessity of manual memory: https://numinous.productions/ttft/#how-important-is-memory.

That article articulated the reason slightly differently, arguing you need to hold multiple concepts in your head at the same time in order to develop original ideas.

Still, I'm not sure you have to remember everything, but I agree you have to remember the foundational things at the right abstraction layer, upon which you are trying to synthesize something new.

maxwell's demon peeking from the dark side

Nice write-up!

I think the router applies to chatgpt app. The developer APIs expose manual control to select the specific model and level of reasoning.

Thanks for calling this out. Yes, agree. I should revise my understanding of plagiarism.

To be fair, the blog has a "Source: NASA" link near the beginning.

Can confirm that they are rolling out. It's working for me.

That was debunked by Veritasium 13 years ago: https://www.youtube.com/watch?v=aFO4PBolwFg

ah, thanks!

I find it strange that GPT-5 is cheaper than GPT-4.1 in input token and is only slightly more expensive in output token. Is it marketing or actually reflecting the underlying compute resources?

Tried using gpt-5 family with response API and got error "gpt-5 does not exist or you don't have access to it". I guess they are not rolling out in lock step with the live stream and blog article?

The ability to specify a context-free grammar as output constraint? This blows my mind. How do you control the auto regressive sampling to guarantee the correct syntax?