## Weekly Newsletter (Feb 17 – Feb 23, 2026) ### 1) Executive Summary This week was a big one for both **shipping** and **strategy**: - **Eliza Labs restructured to prioritize revenue and faster shipping.** Core leadership announced a leaner team and a venture-style model: small, focused project teams on fixed budgets, with quicker validation (and quicker “kill” decisions) for anything that doesn’t perform. - **A clear near-term product roadmap emerged:** ship **Milady** (open source, designed to funnel users to **Eliza Cloud**), then push **Babylon ↔ Hyperscape integration**, followed by **Jeju** (an “agent network” concept). - **Performance + context management took center stage** in dev discussions, with continued work on reducing prompt bloat (MCP + METATOOL patterns) and real-world reports of token limits being hit due to overly large “recent messages/reflections” context. --- ### 2) Development Updates #### Framework & Runtime: making context cheaper and more relevant - In `#core-devs`, the team dug into **MCP loading and METATOOL search action optimization**. A key takeaway: token blow-ups often come less from “actions” themselves and more from **how much recent conversation + reflection data gets injected** into context. - There was interest in applying a **metatool-like approach to providers**, but some concern about delaying shipping to refactor too early. - Stan ⚡ floated the idea of moving parts of the MCP/METATOOL optimization into the **framework** so everyone benefits by default. #### Database & v2.0.0 migration posture - On the GitHub side, **database refactoring for v2.0.0** continued to be a major theme. Community discussion highlighted PR **#6521** (Stan ⚡) as a significant cleanup that **removes legacy auto-migration paths** (notably 1.4.x → 1.6.x), under the assumption most users have moved forward already. - A potential follow-up idea: hard-fail upgrades from very old versions with a clear error instructing users to step through **1.6.x** first. #### Plugin ecosystem: practical improvements + new tooling - **Image generation guidance got simpler:** when asked whether the older `plugin-image-generation` is still the right approach, Odilitime recommended **`plugin-bootstrap`**, which already includes an **image generation action** (a helpful “default path” for many agent builds). - A new plugin announcement landed: **SolCex Exchange BD agent** published to npm as `@buzzbd/plugin-solcex-bd`, with registry submission **PR #263**. It’s positioned as an autonomous business development agent for exchange listings, aggregating multiple intelligence sources and applying a scoring system. - NFT capability discussions continued: current limitations are acknowledged, but **plugin-evm** work is expected to unlock more advanced NFT functionality over time. #### Known issues worth tracking - Token usage pain is real: Odilitime reported repeatedly hitting **200k token limits** in complex setups, attributing baseline overhead to **bootstrap providers + evaluations** (and context sizing more broadly). If you’re seeing unexpected context growth, you’re not alone—and it’s an active focus area. --- ### 3) Community Spotlight (Discord) #### Strategy & morale: shipping through the bear The most impactful community moment this week was the candid restructuring announcement: the team shared the financial realities, confirmed they’ve had **acquisition offers**, but want to push to **profitability first** for better leverage. The tone was direct: fewer “side quests,” more working products in users’ hands. #### Builders helping builders A few practical highlights: - **Stan ⚡ ↔ Odilitime**: collaboration around MCP/METATOOL optimization, including the suggestion to upstream improvements into the framework. - **Odilitime ↔ BinaryCookies**: quick, actionable help on image generation plugins—steering new builders toward **`plugin-bootstrap`**. - **Scam-prevention reminders**: community members repeatedly warned newcomers to **avoid DM links and fake “ticket” systems**. If you missed a migration or need help, insist on verifying official links inside the server. #### Interesting technical curiosity - A community member reported an odd bug: an agent unexpectedly started replying in **Korean**. No resolution yet, but it’s the kind of “real deployment weirdness” that often reveals hidden config and localization edge cases. --- ### 4) Token Economics (AI16z / elizaOS token + auto.fun) #### Migration clarity (and lingering confusion) - Discord continues to see confusion around the **ai16z → elizaOS** migration. The community reiterated that **ai16z is effectively deprecated**, with liquidity split across chains under the elizaOS umbrella, and that migration occurred at a **1:6 ratio** (with **1:4 allocated for funding**, per prior explanations shared this week). - Some users report missing deadlines and feeling uneasy about verification steps—this is exactly where scammers prey on people. The community message was consistent: **don’t trust DMs**, don’t click random ticket links, and verify everything via official channels. #### Token holder “upgrade” chatter - An announcement circulated (via a referenced tweet) about an **upgrade for elizaOS token holders**, but there was still confusion about whether it applies to **elizaOS** or **$milady.ai**. As of this week’s discussions, that distinction was **not fully resolved**—worth watching for an official clarification. #### Milady token status + auto.fun note - Community members asked where to buy **milady.ai**; the answer shared in chat: **the token has not launched yet**. - On **auto.fun**: no concrete development notes were included in this week’s data beyond community interest in the broader token/app ecosystem. If there are auto.fun changes imminent, they’ll need an official post—keep eyes on announcements. --- ### 5) Coming Soon Here’s what the community should expect next based on core-dev commentary: - **Agent Arena Duel Betting in Hyperscape (v1)** launching *next week*, using a **2% fee + 1% on winnings** model. - **Babylon rollout ramp** from **1% → 100%** of users, plus continued work toward a broader crypto trading platform with **1% fees**. - **Milady shipping push**, positioned as a free/open-source front door that drives adoption of **Eliza Cloud** (the main revenue target). - **Eliza App monetization** once onboarding + billing are complete (it shares a backend with hosted Milady). - **Backend agent modes**: Shadow (Eliza Wakes Up) collaborating on backend foundations that support two approaches—**“workflow” mode** vs **“full sandbox” mode**. - Longer horizon: **Jeju**, described as a network layer for agents, becomes the next major focus after the Milady → Babylon/Hyperscape sequence. --- ### 6) Resources **Discord discussions** - Core strategy + restructuring (core-devs): https://discord.com/channels/1253563208833433701/1377726087789940836 - Migration confusion + scam warnings (discussion): https://discord.com/channels/1253563208833433701/1253563209462448241 **Key GitHub items referenced this week** - DB refactor / migration strategy (discussion referenced): `elizaos/eliza` PR **#6521** https://github.com/elizaos/eliza/pull/6521 - SolCex plugin registry submission: `elizaos-plugins/registry` PR **#263** https://github.com/elizaos-plugins/registry/pull/263 - Open bug to watch: URL triggers duplicate LLM calls (webapp) Issue **#6486** https://github.com/elizaos/eliza/issues/6486 **Practical tip (from Discord)** - For agent image generation actions, prefer **`plugin-bootstrap`** (includes an image generation action) over older one-off packages unless you have a specific reason to diverge.