Micheal Mar 7
Sneak peak on what I'm working on for Marchintosh this year. A modern telnet client for System 6 classic 68k macs called Flynn.

Originally started building it with Claude Code since NCSA Telnet doesn't work well with ctrl, esc, tmux and other utilities I use on unix based systems.

Started it last night and now has VT100, VT220, xterm and font support. Plan for 1.0 release will have a dark mode and a few other tweaks.

https://github.com/ecliptik/flynn

#marchintosh #macintosh #retrocomputing #telnet
Show threadMicheal 
Flynn 1.0.1 is released and usable for day-to-day telnet things.

I dropped it into a few shared folders on #globaltalk shares and added it to both Macintosh Garden and Macintosh Repository.

Feel free to try it out, test it, break it and let me know of any issues. I already have a list of features to add to 2.0 and plan to start working on those later this week.

https://github.com/ecliptik/flynn/releases/tag/v1.0.1
https://www.macintoshrepository.org/87841-flynn
https://macintoshgarden.org/apps/flynn

#marchintosh #retrocomputing #classicmacintosh #telnet
Mar 7 at 11:11pmWeb
Show threadnoodleMar 7
@micheal Claudecode on the Mac Classic makes me uneasy.
Show threadMicheal Mar 8
@noodle it is a bit strange, but also surprising usable. It also gave me the idea to add color and emojii support on System 7 for Flynn 2.0 to make Claude look as you would expect.
Show threadnoodleMar 8
@micheal That's worse though.
Show threadJon SchwennMar 8
@micheal I’ve been wanting to try to build something with Codex or Claude to run on system 7 - any tips or recommendations to get started?
Show threadMicheal Mar 8
@jonschwenn

tldr; start small, have Claude make a plan, execute and repeat.

I started with a markdown document that had a list of features and requirements (telnet support, target is System 6 in a Mac Plus, use C). I also put some reference code and pdf documents on Macintosh programming in the repo (but not committed), told it to use retro68 for cross compile and basilisk for testing. Basically something to start with.

I then used Claude Team Orchestration [1] with Opus 4.6 High Thinking, and asked it to review the docs, reference material, ask me questions and make a plan.

Once the plan was done I asked to create a dev team and start implementing the plan, steering things as necessary with everything committed to git.

Biggest hurdles were it tried to use docker containers for retro68 (eventually I told it to build from source) and switching to Snow for emulation since it had issues with BasiliskII.

I tried to be clever and have Claude test things too, taking screenshots, and having it write python scripts to use the keyboard/mouse on X (this was all done on a Linux system). It worked, but was slow and prone to error. Now I just have Claude build and start Snow with the new version and test it myself, providing feedback to Claude.

I think Claude worked well for this app because it was relatively simple and I added features incrementally. I also heavily used Claudes memory (remember this) which helped it avoid doing the same thing twice.

Hopefully this helps. I'm planning to write a blog post on this as well as I took notes (ironically in a physical notebook) on the process.

1. https://code.claude.com/docs/en/agent-teams
Orchestrate teams of Claude Code sessions - Claude Code Docs

Coordinate multiple Claude Code instances working together as a team, with shared tasks, inter-agent messaging, and centralized management.

Claude Code Docs
Show threadMicheal Mar 8
@jonschwenn oh almost forgot. Use --dangerously-skip-permissions and run it on a system dedicated to Claude. I use a 4 cpu 16GiB Debian VM on my home server.

If you don't do this you'll forever be hitting enter whenever it wants to do something. Isolating it on a VM gives me peace of mind it won't totally destroy something. It has its own ssh key for git and only git.