franzap's avatar
franzap
fran@zapstore.dev
npub1wf4p...dgh9
Building nostr:npub10r8xl2njyepcw2zwv3a6dyufj4e4ajx86hz6v4ehu4gnpupxxp7stjt2p8 and #purplestack | BA ๐Ÿ‡ฆ๐Ÿ‡ท
franzap's avatar
franzap 22 hours ago
I don't share any personal details with big AI, I assume it's all being logged and could potentially be used against me. ("Claude has given your details to local authorities" lmao wot) For my public work (open source software), which is 95% of my AI usage, I want nothing less than the best โ€“ Opus 4.5 today. It's a good tradeoff. Some people don't realize you can't really get banned from using these models. Just use a proxy like @PayPerQ (there are many others), or Cursor. For private/personal queries, gpt-oss:20b via Ollama locally, and more powerful models via @Maple AI are great options. View quoted note โ†’
franzap's avatar
franzap 3 days ago
If you send an unsolicited PR with AI-generated text, especially for a previously undiscussed bug/feature: 99% chance I will close it. Code is cheap. Show proof of work! GM!
franzap's avatar
franzap 5 days ago
Non-interoperable freedom tech that works > interoperable freedom tech that doesn't
franzap's avatar
franzap 5 days ago
@Jumble heads up, Shift+Cmd+Z should redo but instead only does undo โ€“ Shift is ignored. This breaks user expectations, lost part of a note I was redacting
franzap's avatar
franzap 1 week ago
There's plenty of terrible things in the world. There's plenty of awesome things in the world. Be aware of the former, focus on the latter. GM
franzap's avatar
franzap 1 week ago
Obsidian and Keet are two closed source applications that typically deal with very private data. It's interesting how Keet gets 100x more hate.
franzap's avatar
franzap 2 weeks ago
User benefits are assets, code is a liability. Code is free to produce and hard to validate: technical debt can easily snowball. PRs are cheap, humans who understand them are not. Yet again, proof of work is the answer. GM
โ†‘