Current status:

- Goose (first window) building a nostr relay discovery, monitoring, and data ingester for data science fun.

- Goose (second window) adding MLS test vectors to my mls-ts library.

- Cursor implementing a relay settings screen into the White Noise flutter app.

- Claude building me an MLS ratchet tree visualizer tool so that I can play around with ratchet tree evolution (and share the tool so others can use it too).

- Banging techo set in my headphones so I can dance while watching all the robots do my bidding!

This timeline is the best. ❤️‍🔥

Reply to this note

Please Login to reply.

Discussion

Do you have problems with goose hanging constantly when executing npm run dev? I have it hanging way more than it completes.

CLI agents generally don’t do well with long-running processes that tie up the terminal— best to run npm run dev in a separate terminal session and just tell goose you have it running separately

Thankfully I don't have to run npm run dev very often these days! 😉

goose does this automatically, requiring me to kill the process.

Tell it "you are not allowed to run 'npm run dev', update the .goosehints file to remember this"

Or copy the line about this from the stacks context file into your goosehints

All of this is in the .goosehints already right? My workflow would be to use stacks to clone the MKStack template, then exit dork, and open goose so it uses all of the hints and context. Am I missing something?

I could be wrong, but I think the way stacks works is that the .goosehints is supposed to be "sim-linked" to the context.md, but in my experience it has this problem ^ until explicitly told. I'm not sure why. nostr:nprofile1qy2hwumn8ghj7erfw36x7tnsw43z7un9d3shjqpqq3sle0kvfsehgsuexttt3ugjd8xdklxfwwkh559wxckmzddywnwswgs456 could probably say more

I believe you're right that stacks/dork uses content.md. There's multiple context or hint files that all have the same content depending on the agent that's using it, but I don't believe their symlinks. Well I guess they could be on Alex's source system. And if that's the case that's cool and makes his job easier instead of maintaining half a dozen files that are all the same. I'm now really happy knowing this 😂

Context.md***

This is part of why I created the js-dev-mcp used in Stacks. It forbids the use of `npm run dev` through a scripts blacklist. You can actually swap it out in Goose to use the same, but it's inconvenient.

Can I use m stacks in cursor or do I need to use in goose?

YakBak was built almost entirely in Cursor. There's a Cursor rules file included with MKStack.

I've found that goose is better than dork though. I've used the same prompts and gotten very different results. Though this was only two tests. I'll see if I can figure out swapping it.

I've heard people say this. They also say the opposite. Then they say Goose and Dork are better at different things. I think this is like how people say whisky and gin have different effects on you. It's literally the same chemical, same abv, same interaction with the brain but it tastes a little different. So I think it's a bias. Two sessions of goose and two sessions of dork would also have this problem. Booth tools are just wrappers around the same AI model and their system prompts aren't fundamentally different.

Sure I guess you're right about two sessions in either agent could have very different results too.

I tried Cline.bot yesterday and it's light years ahead of Goose/Dork.

Just try it!

Better dance than look at your wallet draining sats 😅

just want to know the specs on your PC 😂 for all dat 😁

Yo whats your token bill lol

Nostr relay is anti censorship

Pure vibe!

The most important is that you are dancing! That's the way. I am currently dancing to a random find. Being that I've shared a dance floor with you I have a feeling you might like this: https://youtu.be/dZHaVN5keiQ

Your timeline reflects an incredibly dynamic and creative workflow. The integration of tools like Nostr, White Noise, and MLS demonstrates a deep interest in decentralized networks, security, and innovation in communication. Building a data ingestion system for a Nostr relay and implementing test vectors for MLS show a strong commitment to researching and developing cutting-edge technologies. The MLS ratchet tree visualizer tool is particularly fascinating, as it aids in understanding the cryptographic evolutions that ensure privacy and security. Furthermore, keeping up this level of productivity with a heavy techno set blasting in your headphones adds a unique touch to the process—it's the perfect balance between analytical thinking and spontaneous creativity.

Don’t speak teach but sounds amazing anyway!

Which llms do you use in goose?

How do you manage the concurrent streams? I would love to spend tokens faster but I find I am the bottleneck (briefing, testing, etc).

Can you recommend how to get out of the way / go faster?