jurraca's avatar
jurraca
julien@krewe-of-keys.fly.dev/
npub19w6s...syzj
☯️ | working on asmap.org
jurraca's avatar
jurraca 2 days ago
jurraca's avatar
jurraca 2 weeks ago
mistral finally released a really solid model with medium-3.5. Deepseek 4. Kimi 2.5. Maple AI excellent product update. PPQ crushing it. Routstr. open source AI lookin very bright these days...
jurraca's avatar
jurraca 1 month ago
Milla Jovovich is dropping LLM memory systems and you're bearish?
jurraca's avatar
jurraca 2 months ago
it finally happened... today my father looked at me and said: "OK son, I want you to show me Linux" 😤😤😤
jurraca's avatar
jurraca 2 months ago
openclaw is the first implementation of a new category of software. It's fairly clear (read the codebase) that it's not the best and not the last implementation of this idea. First mover advantage doesn't get you much here.
jurraca's avatar
jurraca 2 months ago
anybody want to send replit credits for sats?
jurraca's avatar
jurraca 2 months ago
Some ASmap developments: an embedded ASmap will be shipped with Bitcoin Core starting in v31. Fabian went on the Optech pod to discuss the PR and the ASmap initiative in general. It's a good overview, check it out! (starts at 15:30 mark)
jurraca's avatar
jurraca 3 months ago
doing the mistral hackathon this weekend, throw me ideas for 402 payments or bitcoin-native agents.
jurraca's avatar
jurraca 3 months ago
a "conversation" with an LLM is an anthropomorphism. a chat should re-render the context at every turn, so the LLM only ever sees a system prompt, a context message, and your message.
jurraca's avatar
jurraca 3 months ago
very good and directionally correct, but some fundamental errors imo: - 402 was revived by lightning labs over 5 years ago, not by coinbase's openx402 - unclear that agents would *prefer* stablecoins when bitcoin's right there. i bet cashu will win. If the agents are truly autonomous, setting up their own bank in cyberspace is ideal. - they won't use proprietary models. The self-improvement loop also means a capability loop, so they may organize to train new models. Either way the idea that their lifeblood and intelligence will depend on API calls to third-party providers is a non-starter for autonomous intelligence. - they still won't know _what_ to build. Agents will have to develop taste, perspective, immense context, etc. View quoted note →