Krzysztof Szafranek

Krzysztof Szafranek

@szafranek.net

Software engineer & photographer at http://photos.szafranek.net

35 followers26 following46 posts

Today I switched to Brave as my new default browser. I started using Mozilla Suite in 2002, then moved to Firebird, now known as Firefox. I converted people to use it, blogged about it, and contributed to Bugzilla, so recent policy changes really feel like a heartbreak.

I've been using a mac for almost two decades and only now learned there a way to cut and paste files with a keyboard shortcut (i.e. what's Ctrl+X Ctrl+V does on Windows). The move command: Cmd+C Cmd+Opt+V More here: saurabhs.org/macos-tips

saurabhs.orgmacOS Tips & Tricks - saurabhs.orgA collection of 150+ nifty tips and tricks for macOS, for both system features like the Dock, Spotlight, and Mission Control and built-in apps like Safari, Mail, and Photos.

Krzysztof Szafranek reposted

Quick TIL: if you have uv installed and you want to try out ANY Python package in an ipython REPL, you can do so just by running uvx --with llm --with sqlite-utils ipython Pass as many --with package options as you like I wrapped it in a shell script, "itry llm" til.simonwillison.net/python/itry

til.simonwillison.netTrying out Python packages with ipython and uvxI figured out a really simple pattern for experimenting with new Python packages today:

This overnight success has been decades in the making, but the last few days feel like a watershed moment for the global balance of power. In one week, the US announced a half-trillion-dollar investment in AI, only for a tiny Chinese startup to erase more than that from NASDAQ.

Meanwhile, Europe is not even in the race. The EU AI Act will look even more hilarious when anyone will be able to download for free and run models much stronger than R1 in their pocket.

Krzysztof Szafranek reposted

Replying to @simonwillison.net

If LLM-assisted code IS uncopyrightable then most of the commercial software industry over the past 18 months is going to have one hell of a mess to clean up I'm personally comfortable with taking that risk

1. LLMs are just tools, like programming languages or frameworks. They can solve a wide array of problems but not all of them. It's more helpful to think about what I can use them for rather than complaining about things they can't do.

Simon Willison@simonwillison.net

What are some of the things you've learned about how LLMs (and LLM-powered systems like ChatGPT) work that were non-obvious but most helped you build a more effective mental model of how to use them?

"Don’t build castles in other people’s kingdoms" is a sage advice. My personal website (szafranek.net), turning 22 this year, is older than Twitter, Bluesky, Tumblr, and even MySpace. This particular castle needs some love in 2025. chrisholdgraf.com/blog/2024/bl...

chrisholdgraf.comHow I’m trying to use BlueSky without getting burned again - WelcomeSome quick thoughts on moving from Twitter/X to BlueSky and how I'll try to use social media after being burned once by Twitter.
View full thread
View full thread
View full thread

So I just did that and ChatGPT all of the sudden turned into a therapist with a set of thoughtful follow-up questions. Despite its warmth, I couldn't shake off the feeling that it's a commercial product eliciting personal responses for the benefit of its corporate overlords, but...

This post is unavailable.

...I do really want from Santa a local, powerful, highly personalized model that can play this role. A model that that becomes better with each conversation, knows me, my story and quirks, so it can provide personal insights.

Existing models, even the free ones, are already capable enough and the only thing I'm missing is the ability to run them *locally* with low latency on an off the shelf laptop or phone, and with *local history/weights* that I could sync myself across devices. Or maybe this already exists?

You are up to date.