# ElizaOS Weekly Newsletter (Apr 26, 2026 → May 2, 2026) ## 1) Executive Summary This week was all about *foundation + reach*—making ElizaOS more modular, more portable, and more economically “real”: 1. **Big architecture unblocking:** the **core agent server was decoupled from app-specific packages** via a new **plugin-lifecycle registration system**, removing hard imports and cleaning up legacy re-export shims. This is a major step toward a truly composable, independently-evolving plugin ecosystem. 2. **Mobile milestone:** a **full local agent architecture for Android** landed, including a **foreground service capable of running Bun** on-device. That’s a meaningful expansion in where agents can live—beyond desktop and cloud. 3. **Token utility got concrete:** **$ELIZA is now the default payment method in x402** for elizaOS/Milady services, moving token discussion from “price” to **infrastructure usage**. ## 2) Development Updates ### Core architecture: agent server modularization - The project completed an **architectural shift** that removes tight coupling between the agent server and application packages. - Key outcome: **plugins can register with lifecycle hooks** instead of being pulled in via hard-coded app imports. This should reduce cross-package breakage and make deployments slimmer and more predictable. ### Android: local agent execution (Bun on-device) - A **local agent runtime on Android** is now implemented, using a **foreground service** to keep the Bun runtime alive reliably. - Why it matters: this is the first serious “agents in your pocket” pathway—useful for always-available assistants, sensor/notification workflows, and offline-ish inference strategies. ### Cloud platform: Next.js → Vite SPA + Hono Workers - The cloud stack migrated from **Next.js** to a **Vite single-page app** backed by **Hono Workers**. - **Image generation storage moved to R2**, improving storage scaling and operational simplicity. ### Runtime reliability + UX improvements - A new **`NoModelProviderConfiguredError`** now gives actionable feedback when agents are misconfigured (instead of failing mysteriously). - **Telegram performance** improved significantly by switching read-receipt fetching to an **ID-based strategy**, cutting overhead in busy chats. ### In progress (tracked as “next to land”) The team flagged several ongoing efforts that are close but not finished: - **Vault & secrets management** - **Runtime operations refactor** - **Registry fixes** - **Cloud CI + worker end-to-end test fixes** - **TypeScript build fixes for the Anthropic plugin** - Two notable build/config issues to watch: - Telegram plugin **subpath export emission** failure - A **missing preload file** issue affecting fresh clones of the main Eliza repo ## 3) Community Spotlight (Discord) ### “Building through the downturn” (core team update) In 💬-discussion, **Shaw** addressed community anxiety around market conditions and project direction: - Confirmed the team is working daily across **Eliza Cloud**, **framework improvements**, and an **autonomous agent with comprehensive crypto features**. - Emphasized the reality that making big public promises can be risky, while still hinting that **“round two is gonna start soon.”** - Shared a pragmatic token stance: **buybacks are planned if/when the project becomes profitable** (i.e., tied to real revenue, not vibes). **Community response:** members like **baogerbao** and **zadayos** voiced support, highlighting continued open-source shipping and visible builder energy even when sentiment is rough. ### Real-world community efforts - **Merch + events:** a community member reported attending a **BNB event in Hong Kong** and promoting ElizaOS merchandise—small moment, but it’s the kind of offline visibility that compounds over time. - **Security moderation:** **odilitime** stayed active on scam prevention (including banning a scammer and reviewing flagged messages). This matters more than ever in down markets. ### New contributor energy Two developers explicitly raised their hands to help: - **aiden190157** asked where to plug in. - **rsn6958** shared strong full-stack + AI credentials with a production focus. If you’re looking for a way to contribute, now is a good time—especially with modularization work making it easier to own a plugin or subsystem. ## 4) Token Economics: AI16z, $ELIZA, and auto.fun ### $ELIZA: utility-first progress This week’s most tangible token-related change: - **$ELIZA is now the default payment method in x402 for elizaOS/Milady services.** This ties the token directly to paid infrastructure usage rather than purely speculative narratives. ### Buyback philosophy (profit-gated) Shaw reiterated a clear principle: - **Buybacks are on the table once the project is profitable.** That’s a healthy constraint: revenue first, token support second. ### AI16z + auto.fun - No specific, confirmed shipping updates on **AI16z** or **auto.fun** were included in this week’s aggregated notes. - What *did* move, indirectly: monetization plumbing (cloud architecture, credits/payments direction, x402 routes) continues to mature—these are the kinds of rails that typically precede broader ecosystem/token integrations. ## 5) Coming Soon (What to Expect Next) Based on both Discord signals and the engineering “in progress” list, here’s what to watch for next week(s): - **Eliza v3 + Milady integration** is “nearing completion,” with Shaw suggesting the next couple of weeks will be eventful. - **Secrets/vault work** to make production deployments safer and less ad-hoc. - **Cloud CI stabilization** (worker E2E tests) after the Vite/Hono migration. - **Plugin build/export fixes** (especially Telegram + Anthropic TypeScript builds). - More clarity on the teased **“round two”**—likely a coordinated product phase rather than a single feature drop. ## 6) Resources (Links) ### Discord discussions - Shaw’s project + token update (💬-discussion, Apr 29): https://discord.com/channels/1253563208833433701/1253563209462448241 ### GitHub / Engineering references - Weekly engineering summary (Apr 26 → May 2): https://github.com/elizaos/eliza/pull/7172 (Android local agent) https://github.com/elizaos/eliza/pull/7204 (agent server decoupling / plugin lifecycle) https://github.com/elizaos/cloud/pull/484 (Cloud migration to Vite + Hono Workers) https://github.com/elizaos/cloud/pull/489 (R2 storage for image generation) https://github.com/elizaos-plugins/plugin-telegram/issues/7009 (Telegram read receipt optimization context) ### Practical takeaway If you only do one thing this week: **pull latest, validate your plugin assumptions against the new plugin-lifecycle direction**, and keep an eye on the remaining build/export fixes if you’re working on Telegram/Anthropic integrations.