Colin Gaffney

Just launched my first open-source ML project to predict college football game winners! 🏈🤖 Perfect timing for Week 3 and my pick'em league. #MachineLearning #DataScience #CollegeFootball

openclaw is the best rts launch in over a decade

@HamelHusain @sh_reya one way to summarize is that LLMs are non-determinant black boxes, which is the bread and butter of data science

@thsottiaux You can ask Claude to remove it, but Cursor is even worse where it auto injects no matter the prompt

Do people like this? We don't do this for codex because it exists to help you and it's important that you remain the owner and accountable for your work without AI taking credit. At the same time it does mean that you can't trace how popular codex is among repos.

@thdxr i think a token budget could work in some cases, ramping up and down depending on project impact, calibrated to the right amount of friction for the builder to lock into their craft

we spoke to a company today who's security team is so concerned by ai code they're considering banning ai tools your first reaction might be "they're gonna get left behind" but if you are practical their concerns aren't invalid if you are a huge multi national org with tens of thousands of employees and they just got a button that appears to do their work, it's gonna get pushed a lot and the process around knowing what is making it to production is totally melting being honest we're all getting a bit lazier see that kiro related aws outage as a real life example so they're genuinely arguing over how much this is going to be allowed esp since the net productivity gains for the average dev seem to be pretty low

@bentossell what is exe?

my current setup; - droid/pi/codex - 5.4 default, opus for frontend + non code - cmux, multiple panes - telegram for my claw - droid missions for big work my “build loop” from my agents .md - create a /spec/ folder - numbered 00_spec1 - progress .md for progress thru implementation - test gate - use agent-browser with dogfood before sending me a url to test - debug until green exe for VMs trigger for automations react-doctor for react best practices here now for quick deployable urls vercel for any full stack deployments (and now domains)

Introducing my American friends to Sam Robert’s Band, uncut Canadian culture

your prompts are weak, plans cheap, context is heavy hallucinating in prod already, claude’s spaghetti

@WHOOP can we get an MCP or something so I can automate the “add exercises” part of weightlifting? Maybe just add that tool to your chat bot?

“I keep it 300 like the Romans” CCC = Creatine, Claude, Celsius

@kepano Not a specific skill, but sometimes pointing to a sub agent that specifically uses the Haiku model is helpful for speed. Opus can be overkill and break flow for simple tasks

are there any Claude Skills you have found useful in your Obsidian vault?

@kepano Dump ideas, links, todos into a scratch inbox file, then have Claude Code parse and route them to the right home

if you're using Obsidian with Claude Code, tell me about your workflow, and what you've used it for

The only thing standing between Google and AGI is fixing Gemini’s voice to text

@APompliano Monthly budget forecasts. Use an expense tracker like YNAB to tag expenses, pull together current accounts, get Claude Code to dump it all into that month’s JSON plus scenarios. Get it to run all the forecast, summarize in graphs and a report

What is the best implementation of artificial intelligence you have made in your business or in your personal life? Extra points for being specific, including how you did it and what the outcome has been.

@LeafsPapi_ Also, most of us in 2023

@lisabdunlap Just ask an LLM to get rounded bar tops by using matplotlib.path.Path with CURVE4 instructions to draw Bézier curves at the corners. Define a custom vertex list for the bar’s outline, wrap it in a PathPatch, and scale the radii by the axis aspect ratio for perfect circles

With all these releases, my biggest question is what package are these companies using to get these rounded bar plots

We have not figured out memories for LLM chat. It still feels like talking to the guy who read the networking book and isn’t discrete enough about his personal CRM

@mikeystephens81 Most unlikely Gemini 3 announcer

Brad Treliving is speaking with the media tomorrow

My wife requesting coffee through the smart light like it’s Season 1 of Stranger Things

@cursor_ai Composer 1 absolutely rips on simple tasks. Definitely part of the toolkit now

@benln LA 🌅

Where would you like to see Cafe Cursor next?