frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: I made a tiny MoE/Engram viz tool

https://sw-ml-study.github.io/moe-microscope/
2•softwarewright•47m ago
As part of a ML study group I've been building tools to experiment with ML concepts. I use AI Coding agents to develop software, mostly cloud-based but I keep trying local LLMs on 12G, 16G, and 24G VRAM GPUs. Freetoken has made it possible to run larger MoE models on smaller GPUs.

This inspired me to use my ML tools to build a tiny MoE model from scratch so I can experiment with different agent optimizations and things like Engram, recursive models, hierarchical modeles, KV Cache compression etc.

My link above shows results of building different tiny models, one dense, the rest MoE variants. There is a lot of code and tooling behind this.

Goals: - build tiny MoE models that run in a browser to give interactive help (today I use deterministic matchers to respond to user queries; still researching how to improve this with better trained MoE models). - build tiny MoE models that run on small ARM and RISC-V development boards, possibly using NPUs (using no_std Rust).

My tools are an alternative to Python, ML libraries, Jupyter notebooks, and Google Colab.

My approach is a REPL and/or AOT Compilable array-programming DSL written in Rust. This language has ML concepts as first class language elements and built in visualization primitives.

It runs in a browser via WASM. It runs at the CLI on both Apple and Linux (supporting Apple GPU and NVIDIA GPU libraries, Ollama-based REPL context-senstive help). It is extensible via dyamic plug-in Rust libraries.

I have created many demo repos showing how to use this ML array language for ML tasks, learning Math concepts, general purpose programming. You can build desktop apps, (Ruby on Rails inspired) CRUD apps. So you can create ML experiments and visualize data in the browser or on the desktop.

It also supports Emacs org-mode for literate programming.

This tiny MOE Microscope is just my latest demo of this approach. I hope it helps someone learn more about MoE experts, routing, Engram, etc. It helped me learn more.

I Stabilized Never Type

https://blog.ihatereality.space/0C-never-type/
2•birdculture•7m ago•0 comments

Show HN: Replay Doctor – Local CLI to find silent prompt cache leaks

https://replay.doctor/
1•danielsaito•7m ago•0 comments

Show HN: SaneHosts – Mac GUI for /etc/hosts with Touch ID before writes

https://sanehosts.com/
1•sanebar•9m ago•0 comments

Assistive Access User Guide for iPhone

https://support.apple.com/en-in/guide/assistive-access-iphone/welcome/ios
1•vismit2000•10m ago•0 comments

Show HN: Imageaicraft] – GPT Image 2 Image Generator and Editor

https://imageaicraft.com
1•zhou594660747•16m ago•0 comments

Making Hundreds of Mugs by Hand

https://www.youtube.com/watch?v=O163JMvWixg
1•brudgers•17m ago•0 comments

Artifactory: In-the-Wild Exploitation of CVE-2026-42016,CVE-2026-42018

https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2...
2•ls1911•19m ago•0 comments

OEMpocalypse: Unprivileged Android app to root on Samsung, Xiaomi, others

https://calif.io/research/oempocalypse
2•negura•21m ago•0 comments

Brief Legal [1A] Thoughts on the "AI Safety" Fights to Come

https://prestonbyrne.com/2026/09/13/brief-legal-thoughts-on-the-ai-safety-fights-to-come/
3•blueSky1989•21m ago•1 comments

A.I. Slopware Is Everywhere Now. Nobody Is Using It

https://www.nytimes.com/2026/09/12/opinion/ai-software-coding-apps.html
6•walterbell•23m ago•0 comments

Show HN: Kairo – Fail-closed LLM inference routing from RTX 5090 measurements

https://github.com/peter941221/Kairo
2•peter941221•25m ago•0 comments

Bernie Sanders proposes 20 years in prison for developers pursuing ASI plans

https://www.tomshardware.com/tech-industry/artificial-intelligence/sanders-proposes-20-year-priso...
4•girish_r•25m ago•0 comments

Show HN: PaperFlip – the bird never moves, you fold the paper

https://apps.apple.com/us/app/paperflip/id6807705987
1•hongbiaoluo15•26m ago•0 comments

Why Forth Matters

https://nomlang.org/post/post.14.sep.2026.html
1•mjbq•28m ago•1 comments

Historic Business Machines Collection

https://cdm16471.contentdm.oclc.org/digital/collection/p16471coll9/search
2•Group_B•29m ago•0 comments

Sere, a compiled Python-like systems programming language

https://sere-lang.com
2•handheldheart•31m ago•3 comments

Show HN: Bipluk, Open-source browser Web MIDI librarian for vintage synthesizers

https://github.com/maxcomperatore/bipluk.com
2•halfradaition•31m ago•0 comments

Show HN: Dllog – Replay debug logs when an operation fails

https://github.com/arhuman/dllog
2•doolta•31m ago•0 comments

The logging dilemma: keeping debug context without debug noise

https://blog.assad.fr/en/post/log-dilemma/
2•doolta•32m ago•1 comments

AI Forces You to Commit to Your Initial Belief

https://idiallo.com/blog/making-changes-mid-sentence
2•walterbell•33m ago•0 comments

Show HN: I made a tiny MoE/Engram viz tool

https://sw-ml-study.github.io/moe-microscope/
2•softwarewright•47m ago•0 comments

The Accelerationist Case for Frontier Pacing

https://contraptions.venkateshrao.com/p/the-accelerationist-case-for-frontier
3•wskinner•47m ago•0 comments

Show HN: Is It Greg?

https://github.com/antoineleclair/is-it-greg
20•antoineleclair•48m ago•5 comments

Bruce Schneier: No, AI doesn't mean the end of mathematics – at least not yet

https://www.theguardian.com/commentisfree/2026/aug/25/ai-mathematics-careers
6•emil-lp•51m ago•0 comments

"Oopsies " FreeBSD Devs Say There's Still GPL in the Kernel

https://fossforce.com/2026/07/oopsies-freebsd-devs-say-theres-still-gpl-in-the-kernel/
2•pizzaiolo•54m ago•0 comments

new free website for people who are interested in free trading tools

2•Mohamed_Amineio•1h ago•0 comments

Made a Simple Link Extractor from HN Comments

https://hn-links.netlify.app/?id=49684393
2•rammy1234•1h ago•1 comments

The Malicious Use of Artificial Intelligence

https://arxiv.org/abs/1802.07228
5•rasengan0•1h ago•0 comments

Sandbox-API

https://hub.docker.com/r/credkellar/sandbox-api
2•credkellarboop•1h ago•0 comments

System 3 Thinking

2•rexossbear•1h ago•0 comments