Hacker News Reader

mark_l_watsonon I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

Yes! My main use of very strong models is in writing my own coding harnesses for small local models, tailored for my needs. I also use very strong models to get much smaller skill files and also writing tools for my harnesses.

re: data centers: pump and dump. Wealthy investors will have made their money and walked away, and the corrupt democrat and republican politicians in Washington will, as usual, protect the interests of the ultra wealthy and leave the general public to pay for poor decisions. There will be a government bailout.

Anyway, on a positive note, I am all in for small local models that are augmented by strong hosted models for specific tasks. Use technology to help people, not make billionaires even more money.

mark_l_watsonon A Friendly Introduction to Racket

I wrote my own AI coding harness in Racket, it works fine, meets my own needs. I did earlier versions in Common Lisp and Python but the Racket version is the most full featured.

mark_l_watsonon Ornith-1.5: From Self-Scaffolding to Self-Improvement

What you say is in agreement with someone I follow on X. My family has been traveling for two weeks and I have been looking forward to trying qwen3.8-37b when I get home.

Now I might try Ornith-1.5-35B first even though I don’t see an official MLX version.

mark_l_watsonon How to disable or avoid intrusive AI

You make a reasonable argument. I will take my chances and more people I talk with now seem more concerned about privacy.

mark_l_watsonon Rhombus 1.1 is now available

That is a really good question. My personal intuition is that LLMs do well with any language as long as you set up for fast syntax checking tools, access to either a REPL or very fast to run tests, etc.

I usually use small local models, and the work to set up very concise skills and efficient tooling is a big part of the fun. I have also adopted the practice of writing my own custom coding harnesses (these can be less than 2000 lines of code, not the huge project you might expect.)

mark_l_watsonon Rhombus 1.1 is now available

Since 1982 I have spent much of the time being paid to work in Lisp languages.

A weird thing: whichever Lisp language I am currently using for work (or a side project) is my favorite.

mark_l_watsonon Rhombus 1.1 is now available

I love Racket, and have a ton of Racket code I have written over the years. When I get home from a vacation I will have an AI translate a small sample of my Racket code to Rhombus: easiest way for me to play with the language.

mark_l_watsonon How to disable or avoid intrusive AI

Good list! Removing as much as possible all AI and surveillance features, then carefully adding back the very few I want is the way I roll.

mark_l_watsonon How to disable or avoid intrusive AI

On our car I can hook up my iPhone via Bluetooth to listen to audio books and music, bypassing carplay.

mark_l_watsonon Health benefits of Tai Chi

There are a near infinite number of Qi Gong and Tai Chi videos on YouTube. I prefer to following new videos, and where my wife and I live they have Tai Chi and Yoga classes. A great use of time, and the feeling good benefits are manefest.

mark_l_watsonon Why Target Common Lisp for Code Generation?

In the mid 1980s I wrote a commercial neural network product SAIC Ansim and I usually got things working in Common Lisp first, then manually translated to C++. Now we can write in CL and use coding agents to translate to other languages... progress!!

mark_l_watsonon Hax – a minimalist, terminal-native coding agent written in C

I didn't try Hax but I hope someone who has can answer a question: does it support suspending current operation if user hits ESC key?

EDIT: I just looked at interrupt.c in the github repo: handles user interrupts. I trashed my Common Lisp harness because I had problems handling interrupts in a portable way. I did better with Racket. I am traveling on vacation without a computer, will try Hax when I get home.

mark_l_watsonon Pixel 11 Pro Fold

I am only 75, but in a few years I have thought of downsizing to one device (and a VPS for a little coding). A future version of this phone that is triple fold would work for me.

mark_l_watsonon Jolt: Clojure compiler implemented with Chez Scheme

Thanks for the kind words.

I just built the Chess example in my Clojure book and the executable was a tiny 18M.

mark_l_watsonon Jolt: Clojure compiler implemented with Chez Scheme

Great work. It just simply worked for me.

mark_l_watsonon Nvidia Nemotron 3.5 Lightning and NeMo Switchyard

I love the wave of new small model releases. Pleasantly surprising that an NVIDIA model runs so well on Apple Silicon using MLX! I was using nemotron-3.5-lightning:30b-mlx with OpenCode on my old (cheap) Mac this morning and no bad experiences except for running slowly.

mark_l_watsonon Squeak 6.1

I used to love using Squeak but moved to Pharo to get better high resolution display support. I should try the Squeak 6.1 release.

mark_l_watsonon Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

From the Ollama docs for this new model:

Ollama's MLX engine provides state-of-the-art performance on Apple Silicon, with support for DFlash and image input:

ollama run muse-glimmer:30b-mlx

mark_l_watsonon Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

To be honest, I never give benchmarks a look. I just use the models for whatever I need to work on, so I can't really make comparisons that are useful for other people.

mark_l_watsonon Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

Meta is rocking AI. As of last week I have been using their excellent muse coding harness with their model Muse Spark 1.2.

Starting this morning I am running their new local 30B model muse-glimmer on my old MacMini 32G using Ollama (remember to increase the context size!) and pi coding harness. I am getting good results with muse-glimmer running locally, with the caveat that everything runs slowly (e.g., give it a task and then go walk outside or do Qi Gong exercises for a while).

mark_l_watsonon Mea Culpa – Dark Hours

He did the right thing. The very idea of Claude copying an existing project to answer one of my prompts or a whole project in Claude Code is a nightmare scenario.

I have all but given up using the multiple trillion weight models in favor of smaller open weight models that have hopefully abstracted away copyrightable details. Right now deepseek-v4-flash-0731 seems usually perfect for my needs.

I saw Cees de Groot post about LLM vibe coded projects containing all sorts of licensed material, and I think he is correct.

Perhaps another good reason to just use the AGPL for new projects? I decided to use the AGPL for may latest fun project https://github.com/mark-watson/racket-coding-agent (personal AI coding harness written for my own use in Racket Scheme). Everything I do is open source or open content books, so I am happy to have everything I do to be used by humans and also for AI training, however: it is scary to by mistake be in the position of not acknowledging other work when appropriate.

mark_l_watsonon Ask HN: What are you working on? (August 2026)

I am retired, so this may not be of general interest: I am working on my fourth AI coding agent, this one written in Racket Scheme: https://github.com/mark-watson/racket-coding-agent

I don't necessarily suggest that anyone besides myself use it. I open-sourced it to perhaps give other people ideas, but it is tailored to just my own needs (e.g., hardwired to use deepseek-v4-flash-0731 via FireWorks.ai; supports just the tools I need).

mark_l_watsonon Jolt: A Clojure compiler implemented on top of Chez Scheme

Are you one of the authors?

Great job! It took me just a few minutes to install Jolt and build a native executable for the Chess program in my Clojure book. Jolt rocks!!!

mark_l_watsonon Jolt: Clojure compiler implemented with Chez Scheme

Clojure hosted on Chez Scheme (same as Racket). So cool. I tried building a standalone executable for the Chess game in my Clojure book using "jolt build -m chess-game.cli -o chess_jolt" and it simply worked great.

I tried again with a datomic demo and it needed the source code to the open source version of datomic, so I need to do some work on that.

mark_l_watsonon Jolt: Clojure compiler implemented with Chez Scheme

I just saw this - works great to build small self contained apps. (I just tested on macOS, also supports Linux).

mark_l_watsonon Pi's Minimalism Is Its Advantage

I find pi-sandbox works well for me. Did you evaluate pi-sandbox? If not, give it a try.

mark_l_watsonon Pi's Minimalism Is Its Advantage

Same experience, sometimes using Little Coder helps make things faster because it is built on pi and strips out some functionality. Probably best to stick with pi however, adding pi-sandbox for some security. I will only run Little Coder in an Apple Container.

mark_l_watsonon Pi's Minimalism Is Its Advantage

I am starting to share the opinion in the article. I used to use Claude Code with strong open models, and sometimes pay for Opus.

Now one of my main drivers is out of the box pi’s 4 built in tools, and I add just two extra tools: pi-sandbox and a paid for search service tool. This setup works great with the latest deepseek v4 flash, switching to more powerful open models occasionally.

I wrote my own coding harness in Common Lisp that is almost free or 3rd party libraries and I basically copied pi + the 6 tools I use for pi (except I have two search tools using different vendors in my Common Lisp code).

Everyone (and every company?) should run their own tests and experiments. I find it sad when I talk with people who default to the most complex and the most expensive tools without even trying to evaluate alternatives.

mark_l_watsonon Harness Engineering for Self-Improvement

No, but a good idea. I want something simple, practical, understandable. I think I will write a short book on this general topic.

mark_l_watsonon Harness Engineering for Self-Improvement

Great article! I am currently writing my second harness (first was in Emacs Lisp, using Emacs as UI; second is a command line coding agent written in Common Lisp).

If anyone wants to argue that it is inefficient writing your own AI coding harness, I wouldn’t disagree. That said it is satisfying to have long coding sessions using my own tools.

The article is a valuable resource, thanks to the author.