DavidCanHelp

I built a self replicating forth bot that speaks S-Expressions over the wire. https://github.com/DavidCanHelp/unit

I built ChatDelta.com to give developers this power hands-on instead of outsourcing it to a company.

unofficial I just made it because I wanted to read headlines in Claude Desktop.

`cargo install hackernews-mcp`

Related: https://github.com/cloudstreet-dev/What-we-lost-when-Git-won

Nice find!

Companion repo: https://github.com/cloudstreet-dev/AI-Red-Teaming-Harness

Open Source: https://github.com/devrelopers/clock-mcp

`cargo install gitwell`

The speed is set to all the way slow when it loads to avoid headaches. Use with caution.

After you try the WASM, don't forget to try the native version: `cargo install unit`

Thanks! It's been a fun project to build. The live demo only scratches the surface — the native CLI (cargo install unit) does real UDP mesh networking, persistence, and cross-machine coordination. More coming soon.

There's a live demo in the browser: https://davidliedle.github.io/WISP/

It's open source! https://github.com/Flutterby-App/Flutterby

There's a companion project (that actually inspired this) that implements it so you can see it in action: https://github.com/devrelopers/repo-mood

Built this yesterday as a quick tool to solve a problem I've seen since 2017 — Slack exports are essentially unreadable JSON blobs with no good way to browse them.

SlArchive processes everything client-side in the browser using JSZip. Your export never touches a server. Drop the ZIP, browse channels, DMs, and threads, search across everything, then export to HTML, Markdown, or JSON if you need to migrate to another platform.

It's a single HTML file, MIT licensed, source at https://github.com/devrelopers/slarchive. Contributions welcome.

Full title: Your Tools Are Lying to You

https://crates.io/crates/ratatui-form is the crate. The docs are at https://docs.rs/ratatui-form

Source: https://github.com/DavidLiedle/ratatui-form

I originally named this project tform, but that was taken so the URL has been updated.

I'm a senior developer on medical leave who wanted to contribute to the community. I used Claude to help create the programming guides I wished existed - carefully prompting for content about languages I've worked with extensively. People were engaging positively with the Perl book before it was flagged. If there are specific technical issues with the content, I welcome that feedback via issues or PRs. But dismissing all AI-assisted content as 'slop' regardless of quality or utility seems shortsighted, especially as these tools become part of how we create and share knowledge.

I'm sharing a work product, just one that used a power tool. You didn't use that power tool to make my work product, I did. That's why people are curious - they want to see what was made. I'm an experienced Elisp programmer, among many other things. I'm curious what Claude Code Opus 4.1 has to say on the topic, aren't you? Why the hell anyone would try to get rich off a book on Perl or Lisp shared under CC0 in an open community is beyond me, but maybe -- just maybe -- somebody wants to read it? It's free. And open source. If that matters.

More context: haters are saying LLM generated content is bad. As a Senior Full Stack Developer with over 26 years of pro experience, I'm having the time of my life with these new AI powers and the doors and discussions they open. People are upvoting. I'm personally not getting anything from this open source sharing. You're the one calling spam. When you pay 200 a month to max your claude code output and really hunker down, think twice before you share its work. Not everyone understands.

My first one even captured the banter with the LLM: https://github.com/cloudstreet-dev/React-is-Awful/blob/main/...

These are all open source github repos. Claude code came up with the commit messages; I'm guiding the tone and content. I read it as it is generated, and review it. I recommend getting a folder with claude code, codex cli, or gemini cli, and tell it it's an author writing a book. Once you've done the work, you'll know how much is human and how much is machine. Moot point amongst all the haters. I'm hoping some more PRs will come in and this content will evolve into a new form of media we collaborate on together.

I often tell Claude Code: "You're an author writing about..." and really dial in the content. When it says it was written by an author who _blank_, that author is a combination of my prompts and Opus 4.1's generated output. I would argue that there is a place for guides, opinions, and more from our LLM friends, and I don't make a penny telling you that or sharing what I've generated. I create these kinds of works because I value them, and the stars on the github repo tell me other people value them too. Take, for example "React is Awful" with over 300 stars. It fills a gap in the learning space, and is CC0. You can OPEN A PR if it's wrong about something, or contribute more. Seemed pretty chill to me.

Thank you for that bit of clarity.

Just trying to be a good community member here. What are your thoughts?

The book is in Markdown on GitHub, CC0 license: https://github.com/cloudstreet-dev/Because-Ruby/blob/main/00...

I went with Sinatra over Rails for complexity and to avoid encouraging cult-joining. :P

Fixes Applied:

  1. AI Self-Crashing Bug - Increased collision buffer zones and improved trail detection
  2. Linux Dependencies - Added comprehensive installation instructions for Debian/Ubuntu, Fedora, and Arch

  Improvements Made:

  - AI now skips 30 recent trail points (up from 10) to avoid self-collision
  - Increased look-ahead distances for all difficulty levels
  - Fixed trail collision detection in pathfinding
  - Added full Linux dependency documentation

  The AI should now survive much longer and provide a better challenge. And thank you for the two PR's!