# ElizaOS Weekly Newsletter (2026-02-09) **Week of 2026-02-02 to 2026-02-08** ## 1) Executive Summary This week was a mix of big momentum on the product/engineering side and some very real ecosystem stress-test moments: 1. **“Milaidy” entered the spotlight as a next-generation Eliza experience** — a Mac-native, plugin-driven app direction that immediately drew testers and contributors. Early integration work surfaced concrete compatibility gaps (notably around provider plugins), giving the team a clear stabilization checklist. 2. **Core reliability and multi-tenant readiness improved** through several hardening efforts: defensive null checks, request-scoped settings via `RequestContext`, better event emission (`MESSAGE_SENT`), and expanded documentation for server environment variables. 3. **Infrastructure performance became a top priority** after a critical PostgreSQL bottleneck was identified: the **logs table is being hit excessively**, slowing responses. Core devs confirmed a major optimization plan is already underway. ## 2) Development Updates (Technical) ### Database performance: Postgres “logs table” bottleneck In `#core-devs`, the team pinpointed a major scalability issue: heavy write/read pressure on the **PostgreSQL logs table** is creating cascading slowdown and visibly degrading agent response times. Stan confirmed a **major optimization plan** is in progress, and the topic is now clearly in the “must-fix” lane before broader scaling. **Why this matters:** better logging architecture (batching, partitioning, retention policies, or async pipelines) typically yields outsized wins—both latency and cost. ### Plugin & runtime robustness (crash prevention and prompt slimming) Several changes across the stack reflect a theme: “make it harder to crash, cheaper to run, and easier to operate.” - **Defensive checks to prevent runtime crashes** (e.g., `Object.entries(null)` style errors) landed in core and bootstrap areas. These fixes address situations where metadata or provider outputs are missing or malformed. - **`MESSAGE_SENT` event emission** was fixed so plugins depending on message lifecycle hooks can reliably trigger when responses are submitted to the central server (closing a long-standing bug report). - A notable direction: **reducing “prompt bloat”** by dynamically filtering which actions/providers are shown to the LLM (via vector search + BM25 re-ranking). This is the kind of systems-level optimization that can improve both quality and cost, especially as plugin counts rise. ### Multi-tenant foundations: request-scoped settings Work this week reinforced multi-user isolation patterns, including **per-request/per-entity settings** via `RequestContext` (AsyncLocalStorage). Practically, this supports scenarios where multiple users share an agent/runtime but must keep **API keys and connection settings isolated**. ### ElizaCloud & product surface issues From Discord reports: - A **duplicate account creation bug** occurred when welcome-credit emails routed users to `dev.elizacloud.ai` instead of production, causing fragmented accounts and missing credit allocation. Community members provided reproducible steps, and the issue was escalated to the cloud team. - A separate **dashboard login loop** (cycling between login and dashboard) was also flagged and forwarded for investigation. ### Babylon: production release + fast bug turnaround The Babylon game experience reached a **production-ready release** with agent-based trading gameplay, and the team handled an early production bug quickly: a **profile image upload failure** was fixed and confirmed by the reporter shortly after. ## 3) Community Spotlight (Discord) ### Milaidy launch energy (and immediate “real-world” integration feedback) The week’s most developer-dense conversations centered on **milaidy** (repo: https://github.com/milady-ai/milaidy). Community members jumped in to test, submit fixes, and debate positioning. Key community-identified friction points: - **Plugin compatibility issues**, especially a missing `MAX_EMBEDDING_TOKENS` constant in newer provider plugins, causing plugin loading failures. - **Wallet fixes** reportedly exist locally but weren’t yet deployed. - Deployment questions (local vs VPS) and whether milaidy fully supports existing Eliza examples remain open. ### Character configuration pain (and a path forward) Multiple users reported agents **defaulting back to the “eliza” personality** even when custom characters were intended, plus occasional **model selection mismatches** (e.g., claude-haiku-3.5 being used despite configuration). Bill Ding noted that a **character editor** is being wired in, which should reduce onboarding-only limitations and make customization more accessible. ### Architecture proposal: PM2 for websocket/process management MochinoLabs proposed integrating **PM2** as an external process manager for websocket handling, alongside a more structured, git-friendly directory layout (including symlink-based patterns). This is still a proposal, but it sparked useful ops-focused discussion—especially relevant for teams running multi-process agent deployments. ### A human moment: learning-by-building One of the most wholesome threads: a community member planning to use the **milaidy setup as a homeschool project** for a 10-year-old. It’s a reminder that “developer experience” isn’t just for professionals—it’s also about approachability and teaching. ## 4) Token Economics (AI16z token, ELIZAOS, auto.fun) This week’s token conversation was intense, and community sentiment dipped sharply. - **ELIZAOS price discussion**: users cited ~**$0.0013**, with repeated daily drawdowns. - **Exchange delistings**: the community flagged scheduled **delistings from three Korean exchanges** (3:00 PM KST referenced). - **Staking**: confirmed by Arceon — **no staking is available currently**. - **Utility & communication**: multiple unanswered questions dominated `#discussion`, including token use-cases, timelines for previously mentioned products, and marketing/accountability concerns. On the constructive side: - A community post mentioned **escrow payments for agent rentals** being implemented (a tangible “utility-adjacent” development if it connects cleanly to the wider ecosystem). - There was also mention of a potential **bullish influencer video** (Crypto Gains) and “upcoming launches,” though no concrete release notes were shared in the provided data. **auto.fun / AI16z note:** No specific, verifiable development updates were included in this week’s aggregated data. If there are official links/threads for auto.fun progress, we should surface them next week—right now the newsletter can only reflect what’s been observed publicly. ## 5) Coming Soon Based on developer comments and active pain points, here’s what to watch next: - **Postgres logging optimization plan** to reduce logs-table pressure and improve response latency. - **Character editor** landing to unblock non-onboarding character iteration and reduce “defaults to eliza” confusion. - **Milaidy stabilization sprint**: resolving `MAX_EMBEDDING_TOKENS` compatibility, tightening plugin versioning guidance, and shipping pending wallet fixes. - **Potential ops improvements** if the PM2/process-management proposal matures into an implementation plan. - **ElizaCloud fixes** around email link routing, welcome-credit correctness, and login-loop bugs. ## 6) Resources (Links & References) ### Key repos / threads - Milaidy repository: https://github.com/milady-ai/milaidy - Discord (Coders): https://discord.com/channels/1253563208833433701/1300025221834739744 - Discord (Core Devs): https://discord.com/channels/1253563208833433701/1377726087789940836 - Discord (Discussion): https://discord.com/channels/1253563208833433701/1253563209462448241 ### Notable engineering items referenced in weekly activity - `RequestContext` for per-user/per-entity settings (multi-tenant isolation) - Fixes for `MESSAGE_SENT` event emission (plugin reliability) - Environment variable documentation improvements (server auth & ops clarity) - Action filtering to reduce prompt bloat (vector search + BM25 re-ranking) If you’re looking for a high-impact way to help this week: jump into **milaidy testing**, reproduce the **character/model configuration issues**, or help triage the **MAX_EMBEDDING_TOKENS** break across plugin versions with clear PRs and a compatibility matrix.