# elizaOS Tweet Ideas

1. our latest browser sdk milestone: running agentruntimes and db directly in the browser, not via websockets. groundwork for elizaos 2.0 is taking shape.

2. considering text-to-speech tradeoffs: elevenlabs (quality) vs browser-based models like vits-web/headtts (speed). what's your preference for voice agents?

3. discovered hardcoded credentials in plugin-sql. fixing this reminds us: security is a process, not a destination. always scrutinize your code.

4. jwt implementation in progress to replace the current x-api-key approach. proper auth is invisible until it isn't.

5. browser sdk progress: streaming capabilities and better error handling coming soon. building an agent runtime that feels native in the browser.

6. meeting with "pro n8n guy" this week—cross-pollination is how ecosystems evolve. integration possibilities are endless.

7. cli refactor underway: centralizing business logic within server package. reducing complexity means clearer boundaries, better developer ux.

8. dynamic prompting for elizaos scenarios is live. now you can test multi-turn conversations with sophisticated user simulation.

9. significant progress on browser integration—soon you'll run agents entirely in-browser with local db persistence. true edge computing for ai.

10. technical milestone: merged agentruntimes into agentserver to reduce redundancy. fewer components means fewer bugs.

11. exploring tts options for browser agents: how light can we make voice generation while maintaining quality? the tradeoffs are fascinating.

12. our collaborative ethos in action: stan and sayonara teaming up on farcaster character "frok" with a 3-week timeline. community builds community.

13. psa: never hardcode credentials. spotted and fixing supabase keys in our codebase. constant vigilance against security debt.

14. new issue: "add cometapi provider support"—constantly expanding our model connectivity options. which providers should we prioritize next?

15. zod v4 and vite v7 upgrades coming soon. staying current with ecosystem tools is invisible work that pays dividends.

16. security update: removing hardcoded credentials from plugin-sql in upcoming pr. best practices are a practice, not a destination.

17. building model_endpoint capability to support saas integration. the modular architecture of elizaos continues to evolve.

18. standardizing on jsx-stream for client-side rendering and component api. consistency breeds reliability.

## Concise Twitter Thread

1. yesterday we made significant strides toward elizaos 2.0 with browser sdk development that will allow agentruntimes and db to run directly in-browser without websockets. moving agent compute to the client means lower latency, better offline support, and edge deployment options.

2. alongside this, we're implementing proper jwt auth to replace the current x-api-key approach, removing hardcoded credentials, and adding streaming capabilities. security and performance improvements that often go unseen but form the foundation of reliable systems.

3. our agent testing capabilities have expanded with dynamic prompting for multi-turn conversations, allowing sophisticated user simulation with persona-driven responses. this brings us closer to testing agents in realistic scenarios rather than single exchanges.

## Platform-specific Post

**Technical Post (elizaOS audience):**
our browser sdk milestone: running agentruntimes with pglite db directly in-browser rather than proxying through websockets. this architecture shift allows for offline-first capabilities, reduces server load, and opens possibilities for truly edge-deployed agents. we're also implementing sse/streaming, proper jwt authentication, and fixing security issues with hardcoded credentials. elizaos 2.0 will be a fundamental rethinking of where agent compute lives.

**Crypto Post (auto.fun audience):**
frok just dropped and we're building him in 3 weeks. our farcaster character is gonna be the crypto natives' best friend while we quietly refactor the entire backend to run directly in your browser. no more centralized compute for your agents—they'll live on your device, with your keys, using your data. true self-sovereignty means computing at the edge, not just storing keys.