Hey. Just in case dev is still alive and well.
If I see some stable demand in the AI trenches land, I'll release @Thalia0S since we finished MVP and were ready to release around 2 months ago.
$Mtard holders as I stated before will get Thalia tokens and some benefits
With recent release of LLama 4, once @AskVenice add it, @MinetardAI will receive an update. I plan to add vision somewhere in April to MineTard, it will add new layer of cool things he could do
Still nothing to report. Dev is alive and good. We have a very impressive progress in AI recently especially with Gemma 3 release yesterday. Thalia will include every important feature on release.
AI future is indeed looking bright with compute cost decreasing every month
Nothing really to report. @MinetardAI is fully autonomous, sometimes I reload minecraft client on server. Thalia is ready for launch, but I won't launch it in current market conditions.
We chill and wait
Seeing complete retardation on TL today. Should I launch @MinetardAI on DeepSeek today just for the sake of using DeepSeek? Should I also tell that @Thalia0S will utilize both DeepSeek v3 and r1 ?
Progress on @Thalia0S is really good. Next week we should launch it with @MinetardAI enabled with some minor updates on his logic.
There are no @Thalia0S token details yet, stay tuned for the next week. And no, we don't abandon @MinetardAI, he still will continue to receive updates. And of course $mtard holders will get piece of a pie from new ecosystem
While I'm very focused on @Thalia0S (platform open soon), @MinetardAI reply quality is really good, I'm glad we reached this point that I don't need to edit his character file every week.
His twitter brains powered mainly by Claude 3.5
This is @MinetardAI streamer scheme sheet. May be it's time to make substack and do an article describing how system works.
TL;DR
Messages come in through multiple platforms (Kick, Twitch, and Telegram) -> These inputs are processed through the Core component, which acts as a central hub -> Core routes messages to either direct responses (!talk) or to one of idle controller blocks
Outputs are then processed through:
- Text-to-Speech for voice generation
- Emotion Selector to determine appropriate expressions
- 3D Model Engine for visual representation
For idle periods
The Idle Controller manages four types of autonomous content:
- Game commentary (using game logs)
- Self-talking (using searching engine)
- Chat summaries (using cached messages)
- Random chat responses (using cached messages)
Additional Features
- Solana Payment Processor handles paid messages/content
- Memory Controller manages both short-term and long-term memory storage
- All outputs maintain consistent character personality through templates and character settings
This is a comprehensive system that can work with either active or quite streams with high customization
While I'm working on [redacted] and have some free time before sleep, I check all vision models currently available and trying different games to play with it. Unfortunately we are far away from something playable yet, LLM/SLM powered agents so far can only play games using API or unlockers