# Fact Briefing: 2025-03-18

## Overall Summary
ElizaOS v2 beta is in active rollout with new connectivity (WebSockets) and substantial GUI memory/knowledge management work landing alongside a large wave of bug fixes. Community discussion is simultaneously dominated by unresolved Twitter-client reliability issues and partner tension around marketing cadence and unclear tokenomics/utility messaging.

## Key Facts

- Shaw stated that the ElizaOS v2 beta is out, with a full launch planned for the end of March.
- WebSockets functionality was added to enable connecting agents to web interfaces.
- PR #3958 added opt-in support for custom embedding servers via environment variables (e.g., USE_CUSTOM_EMBEDDING, CUSTOM_EMBEDDING_ENDPOINT).
- PR #3954 added a feature to view and edit memories associated with an agent.
- PR #3950 added a RAG Knowledge tab in the GUI with document upload, chunking, search, and delete functionality.
- PR #3962 implemented user interaction history scoring (stored in the Account table) to modulate agent tone, currently implemented for the Twitter client.
- Multiple Discord users reported that the Twitter client can post on command but does not run autonomously in the latest version.
- Users troubleshooting RAG were told knowledge directories should be placed under ./knowledge and that enableRag must be set to true.
- A new dao-organization channel was created to strengthen ElizaDAO governance and organize contribution pathways.
- The daily market snapshot reported ai16z token at $0.20 and WBTC at $83,847.71.

## Open Questions

- How to make Twitter client work in latest version?
- Does the AI agent's response generation rely solely on user-provided examples, or is it also influenced by the underlying model architecture and integrated model providers?
- Does anyone know how to calculate/estimate the finetuning cost for 58k rows in .jsonl?
- How to see what's being passed into the context to the model?
- What models/flavors are more compatible for agents?
- How do I get the Twitter client to stop auto replying and only use action handlers?
- What is the significance of the TEXT_EMBEDDING model compared to other models?
- Is it clear how projects that launch without a token can participate in the model / flywheel of the elizaOS launchpad model?
- Is the team still launching "three groundbreaking products" by end of March?
- Is there a team on the implementation of any of it and where to monitor their progress?
- Does the team have good tokenomics people who understand the landscape and feedback loops?

## Categories

### Twitter News Highlights
- Shaw posted about progress on a v2 beta and described a goal of making an app.exe so non-technical users can run agents. (Sentiment: neutral)
- Shaw discussed prior work on an MMO concept using a grid approach and shared old code as potential input for Claude. (Sentiment: neutral)
- DankVR engaged in multiple conversations about agents and related topics (VR, geography, and general commentary). (Sentiment: neutral)

### GitHub Updates

#### New Issues/PRs
- [Pull_request #3958: support any custom embedding server](https://github.com/elizaOS/eliza/pull/3958) by hishboy - Status: merged - Significance: Adds a configurable 'Custom' embedding provider to support OpenAI-compatible embedding endpoints without core code changes.
- [Pull_request #3954: feat: view and edit memories of an agent](https://github.com/elizaOS/eliza/pull/3954) by wtfsayo - Status: merged - Significance: Introduces the ability to view/edit agent memories in the product UI.
- [Pull_request #3950: client knowledge management (RAG Knowledge tab)](https://github.com/elizaOS/eliza/pull/3950) by 0xbbjoker - Status: merged - Significance: Adds GUI-based document upload/search/delete for RAG knowledge management.
- [Pull_request #3962: User interaction history scoring for Twitter client](https://github.com/elizaOS/eliza/pull/3962) by nusk0 - Status: merged - Significance: Stores post-conversation scores and uses them to modulate interaction tone for future responses (Twitter client implementation).
- [Pull_request #3979: fix: Make client use relative URL & firefox fixes](https://github.com/elizaOS/eliza/pull/3979) by odilitime - Status: merged - Significance: Improves client compatibility by removing localhost assumptions and addressing Firefox crypto.randomUUID constraints.
- [Pull_request #3977: fix: store postgres connection URL properly in global config](https://github.com/elizaOS/eliza/pull/3977) by 0xbbjoker - Status: merged - Significance: Fixes project-creation Postgres URL validation and persistence via ~/.eliza/.env.
- [Pull_request #3982: fix: error in plugin-tee](https://github.com/elizaOS/eliza/pull/3982) by HashWarlock - Status: merged - Significance: Restores compatibility by removing a non-existent TEEVendors reference.
- [Issue #3981: Check if packages are recognized as plugins based on package.json configuration](https://github.com/elizaOS/eliza/issues/3981) by jmikedupont2 - Status: open - Significance: Targets plugin discoverability/verification behavior based on package metadata.
- [Issue #3978: Inquiry on debugging remotely (process and example)](https://github.com/elizaOS/eliza/issues/3978) by jmikedupont2 - Status: open - Significance: Requests guidance for remote debugging workflows.

#### Overall Focus
- Development focus on 2025-03-18 emphasized bug fixes (syntax/typos/config handling), client compatibility improvements, and documentation accessibility updates, alongside new issues about plugin verification and remote debugging.
- Across the 2025-03-17 daily report, notable additions included custom embedding-server support, expanded memory/knowledge tooling in the GUI, and multiple UI/UX and logging improvements.

### Discord Updates
- **#💻-coders:** Users discussed v2 implementation problems (Twitter client reliability, RAG directory structure and enableRag flag, plugin installation errors such as Paradex, version mismatches across npm packages, and core-runtime TEXT_EMBEDDING initialization errors). WebSockets support was highlighted as enabling website chat-terminal connectivity, and plugin auto-updating was described as in-progress. (Key Participants: shaw, jin, jintern, notorious_d_e_v, ThanosDaSith24, bradtheaverage, chris.troutner)
- **#🥇-partners:** Partners discussed v2 beta status and launch timing, raised concerns about communication/marketing cadence, and repeatedly asked for clearer tokenomics and token utility explanations. A product-first posture was reiterated, with social/brand work described as a focus after GTM for the launchpad. (Key Participants: shaw, jin, Ben, Zolo, Patt, vincentpaul)
- **#dao-organization:** A new working group formed to improve governance and contribution pathways, including proposals for a partner specialty directory, a DAO resources dashboard, and a process for partners to flag repost-worthy community content. Jin shared where docs and analytics live and described a pipeline that collects retweets into a SQLite database feeding eliza.how/news. (Key Participants: vincentpaul, jin, Patt)
- **#tokenomics:** Users asked for clarification on a HackMD tokenomics document, requested visibility into an implementation team and progress tracking, and proposed using ai16z tokens to pay for API/compute on agents. (Key Participants: Alsara2k, yikesawjeez, Patt)

### User Feedback
- Partners requested clearer and more accessible tokenomics documentation and concrete token utility details beyond a HackMD draft. (Sentiment: mixed)
- Multiple users reported that Twitter agents can post on command but do not run autonomously in the latest version. (Sentiment: negative)
- Users encountered RAG setup errors and were instructed to place knowledge folders under ./knowledge and set enableRag: true to use RAG. (Sentiment: neutral)
- Partners provided marketing feedback emphasizing co-marketing coordination and more frequent product communications leading into the v2 launch. (Sentiment: mixed)
- Users highlighted confusion over plugin installation syntax and asked for improved plugin documentation and onboarding. (Sentiment: mixed)

### Strategic Insights

#### V2 readiness depends on resolving client reliability and cross-platform friction
Discord discussions repeatedly center on Twitter-client autonomy failures, plugin dependency/version mismatches, and Windows/Mac issues while v2 is positioned as consumer-friendly. This creates a launch risk where “anyone can run an agent” messaging depends on reducing setup and runtime failures across clients.

*Implications/Questions:*
  - Should the team publish a short v2 preflight checklist (LLM, Twitter login, Discord connect, plugin load) ahead of full launch?
  - Is there a single canonical “known-good” version set for core + clients + plugins to reduce mismatch reports?

#### Tokenomics narrative gap is driving partner friction
Partners and tokenomics channel users are asking for clearer documentation and progress visibility, while responses point to a draft HackMD and a “product first” posture. Without an implementation/progress surface, repeated questions persist across channels.

*Implications/Questions:*
  - Is there an owner and public progress tracker for tokenomics deliverables (even if provisional)?
  - What minimal token-utility statement can be published now without overcommitting?

#### Knowledge and memory tooling is becoming a core product surface
Multiple merged PRs focus on viewing/editing memories and GUI-based RAG document management, while Discord support requests continue around RAG directory structure and PDF ingestion. The volume of improvements and questions suggests knowledge tooling should be treated as a primary UX path with first-class documentation.

*Implications/Questions:*
  - Should RAG setup guidance be promoted to a top-level quickstart page for v2?
  - Are default paths and error messages sufficient to prevent 'Directory does not exist' failures?

### Market Analysis
- WBTC was reported at $83,847.71, up from $82,423.27 (+1.73%). (Relevance: Provides broad crypto market context for community sentiment and risk appetite.)
- ai16z token was reported at $0.20, up from $0.195 (+3.46%). (Relevance: Directly relevant to tokenomics discussions and partner expectations around utility and communications.)
- Discord discussions referenced a token price decline described as ~70% down from prior highs. (Relevance: Context for increased pressure on token utility explanations and marketing cadence.)