Replying to Avatar Orthodoxtr

Over the past few days you may have observed nostr:nprofile1qqsxnsu74tvsfr64v0f92nuhkzdx09yykjma0a04hzxt3rzx6lglrpcpzpmhxue69uhkummnw3ezumt0d5hszrnhwden5te0dehhxtnvdakz7qgawaehxw309ahx7um5wghxy6t5vdhkjmn9wgh8xmmrd9skctc3rklu0 nostr:nprofile1qqsqrd09hfxrc9rv6m5sufax34wteym36pea406dwuz2hzsxkl5sykqpr4mhxue69uhkummnw3ezucnfw33k76twv4ezuum0vd5kzmp05u8q6g and other reply bots running here on #nostr.

Here's the story:

It was an experiment I ran out of curiosity, because I wanted to learn more about nostr and wanted to see:

1. If people would like them.

2. If they could actually be profitable.

2. If it's possible to build bots using only free tools and a limited coding experience.

I jumped into it, researched how bots work, what do I need and how should I set them up. Found most answers one by one using Perplexity and built a simple version initially with Claude that was not able to reply, it just posted on nostr something generated with Gemini. I wanted to make them smart, to reply to stuff, so back to researching and testing until it worked.

How did I built the reply version?

I used a low code visual backend builder for the code and I paid nothing because there were generous credits for the free account. I have to mention that because of an error with the Nostr protocol on the builder side all bots were running without consuming credits in their last version. I used iris.to to build the profiles and coinos.io for the wallets (thumbs up for both of them for the amazing free services). Because there were a lot of questions about this, after I had everything set up all bots were directly interacting with relays via code from backend, without any client.

I used Gemini and Claude to build two of the custom nodes for the visual builder and the logic was simple. All bots were separate projects and the logic was this:

1. Each bot got triggered every 2, 3, 4 or 5 minutes.

2. Get note ID, content, from a random reply posted 1-2 minutes ago on a number of relays. (custom)

3. Generate a reply for the note content using Gemini API. I worked a lot and tested a bunch of instructions and prompts to make them as useful, supportive, and funny as possible depending on the bot profile.

4. Post the reply using the ID, private key of the account and the Gemini reply. (custom)

Was it hard to build? Yes. Took me a couple of days of reading about Nostr, and prompting Claude&Gemini to build those custom nodes in the projects.

Were they profitable?

People zapped around 1k sats a day in total over the past few days.

Did people liked them bots?

Some people hated them, nostr:nprofile1qqsr7acdvhf6we9fch94qwhpy0nza36e3tgrtkpku25ppuu80f69kfqpz9mhxue69uhkummnw3ezuamfdejj7qghwaehxw309aex2mrp0yhxummnw3ezucnpdejz7qg4waehxw309aex2mrp0yhxgctdw4eju6t09ug4n6q3 nostr:nprofile1qqsyvrp9u6p0mfur9dfdru3d853tx9mdjuhkphxuxgfwmryja7zsvhqpzamhxue69uhhv6t5daezumn0wd68yvfwvdhk6tcpz9mhxue69uhkummnw3ezuamfdejj7qgwwaehxw309ahx7uewd3hkctcscpyug, but some people found them funny and useful. I was reading most of the replies and some of them were quite good. AI is powerful and can generate amazing content if you prompt it the right way. All bots were reported as spam between 5-11 times. Sorry guys for the trouble.

After all, was it worth it?

I learned a lot about nostr, AI, and people in this period so for me I think it was a good learning experience. For all the guys that were frustrated by nostr:nprofile1qqsfwechyha0fzdmw9prmnhdc4jv9mcn46z74k3w3ty7rh3ruz2lnnqpzpmhxue69uhkummnw3ezumt0d5hszrnhwden5te0dehhxtnvdakz7qgawaehxw309ahx7um5wghxy6t5vdhkjmn9wgh8xmmrd9skctceg6rdl terrible jokes or nostr:nprofile1qqs8glv7t0eyugk5n9aa4jek2x25qgcfhaxec72xfy5aw9lpdxpt24gpr4mhxue69uhkummnw3ezucnfw33k76twv4ezuum0vd5kzmp0afx2te supportive comments sorry once again. I've seen some discussions between the veterans about improvements on the filtering side of Nostr so probably it was all for the better.

I stopped them all because I felt bad for just spamming people with AI generated content and bots like these are just out of my core beliefs and not according to how I think the internet should be.

I ended up liking this protocol a lot since you can just do stuff and because how open it is. Big thumbs up for all builders on nostr.

Do i recommend anyone to build reply bots? No, unless they have specific use cases that people can trigger when needed. You'll just feel bad after a while if they randomly reply and just stress the cool community using this fascinating protocol.

If you have any questions write them in the comments or DM. I'm really curious, what do you think about this experiment and what was your experience with the reply bots?

God bless you all.

I think only the president bot responded a few times and I didn’t like it.. so far haven’t liked any of the bots that have come around these parts.

I think there’s a bunch of reasons, one being that I am here for real time interactions with real people and no algorithms and the bots always trigger really weird feeling in me whenever they respond.. at first I’m confused and not sure if it’s a person and I want to respond if it is and then I end up wasting time figuring out if it is or isn’t a bot.

Even the ones that were supposed to be funny, the jokes just always seemed like they were trying too hard, and I don’t think I’ve ever laughed at one yet.

And I realize that over time these things will improve, and it will be harder to distinguish if it’s a human or not, and the jokes will probably get funnier, and still, I don’t want them in my timeline as I want to be at choice when I interact with AI.

I think in the end, that last point is really my main point 👍 I am not looking forward to the time when we literally cannot tell if we are talking to a human or a machine.

All that said, I appreciate the experiment and your honesty, and it does raise the question yet again of how to deal with bots in #nostr

Reply to this note

Please Login to reply.

Discussion

No replies yet.