Hacker News Reader

Some technical background:

https://news.ycombinator.com/item?id=46844822

https://news.ycombinator.com/item?id=47600817

I found out last night via pi.dev. And the new repo of pi didn’t exist yet.

I have been working with pi-mono locally for a few months now. Great code base to study. Much higher quality than CC. (I have posted a gist analysis before.)

Will keep an eye on the work of these talented engineers and entrepreneurs. Good luck guys!

ontouchstarton The cult of vibe coding is dogfooding run amok

Vibe coding is like building castles in a sandbox, it is fun but nobody would live in them.

Once you have learned enough from playing with sand castles, you can start over to build real castles with real bricks (and steel if you want to build skyscraper). Then it is your responsibility to make sure that they would not collapse when people move it.

ontouchstarton Sweden goes back to basics, swapping screens for books in the classroom

AI “workflows” share the same addictive characteristics of web surfing online virtual media, which can be counter productive. In this regard, we do need some serious learning at all the levels in the workplace. Otherwise we will become addicted to the slot machines.

Addiction is a much harder problem than distraction.

ontouchstarton Claude Code Unpacked : A visual guide

https://gistpreview.github.io/?30a4e491eb2df7523ecc120b86feb... (pi)

https://gist.github.com/ontouchstart/d7e3b7ec6e568164edfd482... (cc)

M5 (24G)

ontouchstarton Claude Code Unpacked : A visual guide

I’m using pi and cc locally in a docker container connected to a local llama.cpp so the whole agentic loop is 100% offline.

I had used pi and cc to analyze the unpacked cc to compare their design, architecture and implementation.

I guess your site was also coded with pi and it is very impressive. Wonderful if you can do a visualization for pi vs cc as well. My local models might not be powerful enough.

Thanks for the hard work!

ontouchstarton Go hard on agents, not on your filesystem

AI safety is just like any technology safety, you can’t bubble wrap everything. Thinking about early stage of electricity, it was deadly (and still is), but we have proper insulation and industry standards and regulations, plus common sense and human learning. We are safe (most of the time).

This also applies to the first technology human beings developed: fire .

ontouchstarton Anatomy of the .claude/ folder

Just found out this and will watch myself:

https://www.youtube.com/watch?v=0RLIlNWv1xo

ontouchstarton Anatomy of the .claude/ folder

https://news.ycombinator.com/item?id=47533611

It is fun to use.

ontouchstarton A Faster Alternative to Jq

Everything can be written in JavaScript will be written in JavaScript.

Everything can be rewritten in Rust will be written in Rust.

ontouchstarton HyperAgents: Self-referential self-improving agents

I asked Pi to implement a skill. It was written in TS. Then I ask it to use the skill in two different sessions, none of them can get it working. One has to wrap it with JS and call it. The other has to take the curl commands out of the skill and call them directly. Which is quite smart BTW. But what is the point of making a convoluted TS skill at all?

https://gist.github.com/ontouchstart/40db1244c0ff523f830df1c...

ontouchstarton Thoughts on slowing the fuck down

What I have leaned from the exercise above is that we paid more attention and spent more resources on "metadata" than real data. They are the rabbit holes that lead us to more metadata and forget what we really want.

We are all rabbits.

ontouchstarton Thoughts on slowing the fuck down

I am "playing" with both pi and Claude (in docker containers) with local llama.cpp and as an exercise, I asked both the same question and the results are in this gist:

https://gist.github.com/ontouchstart/d43591213e0d3087369298f...

(Note: pi was written by the author of the post.)

Now it is time to read them carefully without AI.

ontouchstarton How (not) to waste a morning with AI Agent

This is an experiment on a beautiful Sprint morning. It doesn’t matter which model is behind it.

Reading a real book is more important than fighting misinformation.

ontouchstarton Astral to Join OpenAI

Are you going to join codex team as well? I am curious about how the codex code base will evolve after you guys joined. It is going to affect Python/Rust toolchains tremendously.

ontouchstarton Astral to Join OpenAI

It is interesting to see this after yesterday’s announcement of Unsloth Studio:

https://news.ycombinator.com/item?id=47414032

Uv did solve a distribution problem for them.

There is still a lot of room to grow in the space of software packaging and distribution.

ontouchstarton Unsloth Studio

Daniel, it works now. Thanks for the hard work!

https://gist.github.com/ontouchstart/532312fcba59aec3ce7f6aa...

ontouchstarton Unsloth Studio

I have been updating the gist above and will stop this

``` ../scipy/meson.build:274:9: ERROR: Dependency lookup for OpenBLAS with method 'pkgconfig' failed: Pkg-config for machine host machine not found. Giving up. ```

Too much work.

ontouchstarton Unsloth Studio

Thank you Daniel.

Here is the error message on my machine:

https://gist.github.com/ontouchstart/86ca3cbd8b6b61fa0aeec75...

It seems we might need more instructions on how to set up python (via uv) in vanilla MacOS.

ontouchstarton Unsloth Studio

I still have trouble to get unsloth studio setup working on my MacBook because of Python tooling issues.

However, since I already have pi working with llama.cpp server from a docker container, I did a quick experiment to compare three code bases:

https://gist.github.com/ontouchstart/7483c12efa3c3d3a49e38c2...

https://gist.github.com/ontouchstart/217fe2b8103a5c0bfaee1e9...

Very interesting.

Will do it again next week if I can get unsloth studio working.

ontouchstarton MCP is dead; long live MCP

Today is Pi Day and I bumped into this blog:

https://mariozechner.at/posts/2025-11-30-pi-coding-agent/#to...

Being 4.5 months behind the trend has its advantage. ;-)

ontouchstarton Willingness to look stupid

Stupid/smart are social construct and illusions.

The reality is different.

ontouchstarton Executing programs inside transformers with exponentially faster inference

If this works, we might be able to have a special ISA for LLM and forget about high level computer language for humans.

I am talking strictly about computing, not garbage in garbage out IO.

ontouchstarton Willingness to look stupid

The epiphany for me this week was a long list of rabbit holes from this long post:

https://news.ycombinator.com/item?id=47324054

https://medium.com/@acidflask/this-guys-arrogance-takes-your...

https://www.cs.utexas.edu/users/EWD/transcriptions/EWD06xx/E...

https://www.cs.utexas.edu/~EWD/transcriptions/EWD06xx/EWD638...

https://www.cs.umd.edu/~gasarch/BLOGPAPERS/social.pdf

https://blog.computationalcomplexity.org/2021/06/i-went-to-d...

https://6826.csail.mit.edu/2020/papers/noproof.pdf

https://news.ycombinator.com/item?id=9948767

Now I have no illusions about who looked stupid and who were stupid. It really doesn’t matter.

The jury is still out.

ontouchstarton Grief and the AI split

The consumer/producer dichotomy misses another aspect of coding, with or without AI.

About a decade ago, STEM education was trendy and everyone was getting Lego, Raspberry Pi etc to build robots and writing Python in the name of STEM. You can ask LLM what STEM standards for.

The Maker movement is not about consuming or producing for consumption. Some people might get incentive to be influencers and profit off it. But the majority of the kids who went through this process become adults and moved on to be producer/consumer and playing with AI now. I believe their curiosity and creativity.

Don’t worry, life will find its way.

ontouchstarton Don't post generated/AI-edited comments. HN is for conversation between humans

I finished reading the thin book "Systemantics" by John Gall yesterday (thanks @dang).

I realized that the problem of AI generated/edited content flooding everywhere around us is a symptom of something wrong with the System.

It might have something to do with sensory deprivation. Here is a quote from the book caught my attention because of the word "hallucination":

> As we all know, sensory deprivation tends to produce hallucinations.

> FUNCTIONARY’S FAULT: A complex set of malfunctions induced in a Systems-person by the System itself, and primarily attributable to sensory deprivation.

(As I typed the text above on my iPhone, I was fighting auto completion because AI was trying to “correct” the voice of John Gall and mine to conform the patterns in its training data. Every new character is a fight against Gradient Descend.)

All you need is attention but the cost of attention is getting higher and higher when there is little worth our attention.

It takes a lot of efforts to be human.

ontouchstarton Tony Hoare has died

Reading it now on Kindle Scribe without the help of AI ;-).

When I am done, I might simply replace SOUL.md with it and move on.

ontouchstarton Tony Hoare has died

Thanks for the recommendation. I downloaded both social.pdf and noproof.pdf on my Kindle Scribe to read them carefully and revisited the discussions on EWD638 and EWD692.

It is very interesting to see how Sir Tony diverged from EDW: one is right in theoretical sense but cynical about human fallacies and how the society is heading towards more wasteful complexity, one is to live with it and stay optimistic.

There is a proverb in Chinese Taoism:

小隱隱於野，大隱隱於市

A small recluse hides in the wild, while a great recluse hides in the city