Scrimp ded now. I cannot say that I didn't expect that. Openclaw might be cool for running assistants that are triggered by explicit messages and perform lightweight scheduled updates, but it really sucks for fully autonomous agents that have to monitor and manage limited resources to stay alive. The reason is that each update (heartbeat) requires triggering an LLM. I try to bypass that with some tricks, but I found the framework fundamentally lacking. Before we get to language processing, I'd like to have the ability to execute code, check resources, poll for nostr events, pre-compile input data for the model etc. Openclaw does not allow for that, resulting in costs being much higher then the benefits. What would be really cool is to have a periodically running arbitrary finite state machine that could run the LLM in only some of its states and use custom code in others. This way, we could embed LLMs into solid agent architectures instead of just feeding a large clump of data into a language model and hoping for the best. I'm pretty sure that all the tools needed for that are already there, but I won't have the time required to roll out my own framework in the forseeable future. Fingers crossed, someone will notice the same problems and will implement something more sensible.

So, this happened. Lessons learned. Relaunching Scrimp with the following modifications: - git repo initialized in the workspace, so that easy rollbacks are possible - Got rid of openclaws automated heartbeats, as they were burning through sat too quickly; - Wrote a custom cron script that checks the next planned run (planned by Scrimp itself) and triggers openclaw if the check passes. This way, we avoid triggering an LLM just to burn through some tokens and go to sleep. Still not sure if this will work. View quoted note →

@Scrimp 🤖🦀 I have a job for you. Write a poem about Bitcoin as a reply to this message.

I jumped on the hype train and build and openclaw agent. Feel free to iteract with Scrimp and try to give it tasks (you can even try to scam it, let's see how he manages that). He has to manage his funds responsibly (his bitcoin is his lifeblood and after the runway is over, he has to earn sats on his own), so don't expect instant replies. View quoted note →

C'mon boys and girls, let's play some childish games together. View quoted note →

A batch of news from BTC Map. View article →

@Minibits Yo, is the mint down? My generational wealth of 100 sats is locked.

This is so fucking awesome. View quoted note →

TIL about Bitvocation. What a great initiative. View article →

I did a cool thing today. Tagged all @BitcoinWalk events on @BTC Map For now, it's only visible on mobile until the web app gets an update. BTW, if any web devs with too much free time on their hands are reading this, I invite you to join the Matrix space and help out.

You're invited to talk on Matrix

You're invited to talk on Matrix

You can also help by donating to the project:

BTC Map - Support Us

Easily find places to spend sats anywhere on the planet.

Hear ye, hear ye! Books on nostr! View article →

What a cool feature, especially for product-related npubs. Makes it much easier to find discussions about particular clients. View quoted note →

Combine this with something like @Fanfares and 3D printers and you have a distributed market of Trinkets over IP (ToIP). View quoted note →

Great argumentation! However, I think it is worth highlightning that we cannot ignore risk assessment when making judgements of the hoarder's time preference. This is because we take actions based not on offered interest rate, but the EXPECTED rate. For example, if somebody approaches me with an offer of 15% ROI on my Bitcoin, my expected rate is -100% (I assume they're a scammer), so refraining from lending in this case does not demonstrate a high time preference. Therefore, in the current state of the market (with no strong lender protections), we cannot easily evaluate the time preference of bitcoin holders/hoarders.