
Michael Russo
@mjrusso
Followers
765
Following
2K
Media
148
Statuses
1K
Bug fixes and performance improvements.
Toronto, Ontario
Joined January 2008
I called it "Wingman" (as there's already a bunch of Emacs packages with "llama" in the name). Source available here: https://t.co/aNTtGsNI4b It's rough, and I haven't put it through its paces yet, but I'm still blown away that you can get good completion running entirely
github.com
Emacs package for LLM-assisted code/text completion - mjrusso/wingman
0
0
3
I think I found the best way to prep for @TOtechweek
Bringing the heat for @TOtechweek. “On Tuesday, conditions are expected to peak at 33 C and feel more like 41 C.” ☀️
0
0
4
Some terms that pair well here: "context rot" (to describe quality drop-off as context fills with junk) and "context engineering" (an alternative framing of prompt engineering; more accurately describes what we're doing as we solve problems with LLMs). https://t.co/GhJ07vJTFR
I really like the term “context engineering” over prompt engineering. It describes the core skill better: the art of providing all the context for the task to be plausibly solvable by the LLM.
0
2
2
This is a great tutorial on using Claude Code. My process is slightly different and probably useful to document. But first: - Tokens: more != better. Anthropic's business model is selling tokens; if you don't babysit Claude Code it will happily slurp up more and more input
Claude Code has completely changed the way we develop our app, Chorus, so much so that we now consider it one of our employees. We now put an asterisk on the whiteboard next to any task that Claude will mostly handle. And the asterisks are multiplying. Here’s what we’ve
2
2
14
This is an excellent interview, packed with a lot of great advice that's broadly applicable to all facets of building products, leading teams, and honing your craft.
I recently had the privilege of chatting with @ridd_design who runs the highly respected @joindiveclub series — truly an exceptional interviewer and storyteller. We (had!) to start with @tobi's note on AI, but we ventured through many of my favourite topics: critique, quality, a
0
0
2
For the past few months, I've been using my Moonlander clamped to my desk (with tripod mount). Actually a pretty nice setup, but I couldn't quite get it positioned exactly where I wanted. (Also annoying to not be able to easily move the keyboard out of the way to make more desk
1
0
3
If you happen to find yourself at a party surrounded by other like-minded LLMs, a fun question to ask might be "is ISO 8601 a superset of RFC 3339".
1
1
3
Incredible snowfall, perfectly timed for Family Day. Let's do this again next year
0
0
3
Synthesis is such a great feature. Works amazingly well for technical questions (quickly getting to "there's consensus that x is the best approach", "some disagreement about y, consider z" for any query is huge signal). Even works well for squishier questions, like this one:
1
1
5
Sidenote: this particular example references the newly open sourced Oban Web package. (Massive thank you to @sorentwo... a truly incredible gift.)
📯 Oban Web is open Source 🐬 Oban v2.19 includes MySQL support 🧭 Web v2.11 is released 📊 Met v1.0 is released (and open source) https://t.co/PUV8DXgmPL
0
0
2
Nice, the latest version of Chorus automatically extracts the contents of any URLs you paste into the chat. Here's an example of using this feature with hex2txt (hex package docs → llms.txt file):
We made a Mac app that lets you chat with Claude, 4o, 01, and Gemini all at once. Introducing Chorus!
1
2
9
@mitchellh Update: I've deleted *all* of these keybinds (and the horror of custom escape sequences) by adopting Kitty Keyboard Protocol (and migrating tmux → zellij). I'm just about ready to move to the terminal as my primary way of interfacing with Emacs. Absolutely love Ghostty 🤩
0
0
1
In case @mitchellh asks why my Ghostty config is so long, it's because running Emacs in the terminal is the hardest problem in computer science
2
0
8
I was about to give my 7yo a stern talking-to for sorting Lego by colour. Then I saw the instructions :-/
0
0
2
As with hex2txt, there are tons of opportunities for improvement. Source here:
github.com
hex.pm package → llms.txt. Contribute to mjrusso/hex2txt development by creating an account on GitHub.
0
0
1
hex2context is packaged as a self-contained Livebook. It does Retrieval-Augmented Generation, with embeddings generated and stored locally. The notebook source is part of the hex2txt repository, but you can easily run it here: https://t.co/ii8jvvbEAT
1
0
0