## Weekly Newsletter (ElizaOS) — Week of **2025-12-29** to **2026-01-04**

### 1) Executive Summary

**1) Faster, more resilient multi-step agent execution landed in core.**  
Multi-provider workflows in `MultiStep` got a meaningful performance upgrade: providers that used to run sequentially now execute in parallel (with fault tolerance), improving responsiveness when an agent consults multiple services/models in one turn (merged **PR #6263**).

**2) A critical database isolation bug fix is in flight.**  
A high-impact issue affecting PostgreSQL when `ENABLE_DATA_ISOLATION=true` was identified and addressed: PostgreSQL `SET LOCAL` commands can’t be parameterized, which broke database operations under isolation. A fix (plus unit + integration tests) is proposed in **PR #6316**.

**3) The community aligned on “ElizaOS as the Linux of agents” and how to build practical multi-model agents.**  
Discord discussions sharpened the mental model for ElizaOS (agent infrastructure + plugin ecosystem), and produced actionable guidance for multi-model setups (e.g., Anthropic + OpenAI together via OpenRouter) and deployment patterns (cloud containers for custom plugins).

---

### 2) Development Updates

#### Core framework (elizaos/eliza)

- **Performance: parallel provider execution in MultiStep (merged).**  
  **PR #6263** refactors `runMultiStepCore` to execute providers concurrently via `Promise.allSettled`, maintaining fault tolerance while reducing end-to-end latency for multi-provider runs. If your agent uses multiple providers/tools in a single step, you should see faster completion—especially when one provider is slow.

- **Database reliability: Postgres isolation fix (opened).**  
  **PR #6316** updates `plugin-sql` entity context behavior to use `sql.raw()` for `SET LOCAL`, avoiding Drizzle’s auto-parameterization (which caused `syntax error at or near $1`). The PR includes unit tests (`withEntityContext`) and an integration test against a real Postgres instance—exactly the kind of coverage you want for a “breaks everything when enabled” class of bug.

- **Product/UX direction emerging from new issues.**  
  A cluster of newly opened issues points to near-term work on public-agent UX and cost controls:
  - **#6313 — Separate public agent states** (unauthenticated / authenticated non-owner / owner) with tailored UI and gating.
  - **#6312 — Limit messages for non-signed up users (~2–3)** to reduce abuse/cost while keeping “try before signup.”
  - **#6314 — Public agent cards should show chat number** (total messages across users) as a lightweight social-proof metric.
  - **#6311 — Improve chat summaries** (current summaries aren’t helpful enough; request for clearer, more meaningful output).
  - **#6315 — Change free credits from $5 to $1** reflecting an active discussion on sustainable onboarding incentives.

- **Maintenance:**  
  **PR #6301** updates the LICENSE year to 2026 (routine housekeeping).

#### Plugins & ecosystem (from last week’s repo activity)

While this newsletter focuses on the week ending Jan 4, the most recent cross-repo context (Dec 28–Jan 3) suggests continued momentum:
- **plugin-twitter:** OAuth2 PKCE work underway (security + better auth ergonomics).  
- **plugin-openai:** improvements for image description handling and caching for media paths (performance + quality).  
- **registry:** new community plugin submissions continue (expanding the “batteries included” feel).

---

### 3) Community Spotlight (Discord)

#### “ElizaOS is the Linux of autonomous AI agents”
A recurring framing this week: ElizaOS as foundational infrastructure—**not just an app**, but a composable framework for building and deploying autonomous agents with a large plugin surface area. That analogy resonated with newcomers trying to understand “what ElizaOS is” and how it stays flexible.

#### Practical guidance: multi-model agents (Anthropic + OpenAI)
A standout Q&A: *“Can I implement two models in one agent at the same time?”*  
Community guidance emphasized using the **OpenRouter plugin** and configuring provider/model choices via environment settings. The takeaway: multi-model architectures are viable today—use one model for “calculation/forecasting” and another for general reasoning, and let routing/config drive the split.

#### Deployment advice: cloud containers for custom plugins
When asked about deploying agents that rely on plugins not yet supported in a managed environment, the community recommendation was clear: **use cloud containers** (bring your own runtime, keep your plugin set intact). This is a pragmatic middle ground between local dev and fully managed deployments.

#### Notable community project thread: “roseOS”
A community member shared an experimental agent framework built on ElizaOS with an emphasis on:
- explicit agency boundaries  
- constraint-aware reasoning  
- accountability layers  

Even if you’re not adopting that design directly, the thread is pushing useful conversation: how to make agents more auditable, more predictable, and safer to operate over long time horizons.

#### Support and troubleshooting highlights
- A plugin versioning issue was flagged (browser plugin versions differing between stores).  
- “Model not found” errors came up in website integrations (API key + provider config likely culprits).  
- An agent memory expectation gap surfaced (“can’t recall info from bio”), a reminder that memory plumbing + retrieval configuration matters as much as prompt text.

---

### 4) Token Economics (AI16z token & auto.fun)

This week’s visible signals were less about on-chain mechanics and more about **product knobs that influence cost, growth, and sustainability**—which ultimately shape token economics and platform health.

- **Onboarding incentives under review:**  
  The proposal in **issue #6315** to reduce free credits from **$5 → $1** suggests a push toward more sustainable acquisition and lower subsidy burn.

- **Public agent funnel tightening:**  
  The combo of **message limits for unauthenticated users (#6312)** and **separating public agent UI states (#6313)** reads like funnel optimization: preserve “instant try” while nudging conversion to accounts (and, downstream, paid usage).

- **auto.fun status:**  
  No concrete development notes or releases for **auto.fun** were included in this week’s dataset. If you’re tracking auto.fun closely, keep an eye on the next round of product changelogs/announcements—this week’s UX issues strongly imply foundational work is happening to support better discovery, sharing, and gated engagement.

*(As always: avoid relying on rumors for token-specific details—stick to official announcements for supply, emissions, listings, or contract changes.)*

---

### 5) Coming Soon

Based on what was merged, opened, and actively discussed, here’s what looks “next up”:

- **Public agent experience revamp:** clearer UI states, better sharing, and more intentional conversion flow (**#6313**, **#6312**).
- **Better agent discovery metrics:** “chat number” on public cards (**#6314**) and (eventually) richer analytics.
- **Higher-quality chat summaries:** improvements requested in **#6311**—expect iteration here as summarization becomes more central to managing long-running agent threads.
- **Data isolation robustness:** once **PR #6316** merges, teams using Postgres isolation should regain stability (and confidence) with better test coverage.
- **Docs/examples request:** community asked for practical, copy-pastable examples of multi-model configuration and cloud-container deployment patterns.

---

### 6) Resources

**Key PRs**
- PR **#6263** — Parallel provider execution in MultiStep: https://github.com/elizaos/eliza/pull/6263  
- PR **#6316** — Fix Postgres `SET LOCAL` under data isolation + tests: https://github.com/elizaos/eliza/pull/6316  
- PR **#6301** — LICENSE year update: https://github.com/elizaos/eliza/pull/6301  

**Key Issues (Product / UX / Growth)**
- **#6313** Separate public agent states: https://github.com/elizaos/eliza/issues/6313  
- **#6312** Limit unauthenticated messages to ~2–3: https://github.com/elizaos/eliza/issues/6312  
- **#6314** Public agent cards: show chat number: https://github.com/elizaos/eliza/issues/6314  
- **#6311** Improve chat summaries: https://github.com/elizaos/eliza/issues/6311  
- **#6315** Free credits $5 → $1: https://github.com/elizaos/eliza/issues/6315  

**Community guidance to revisit**
- Multi-model agent setup via **OpenRouter plugin** (configure provider/model via env)  
- Deployment via **cloud containers** for agents with custom/early plugins  
- Logging consistency: discussion referenced improved logging + a logging linter direction (watch for follow-up PRs/docs)

---