## Intel — 2026-01-04

### 1) Data Pattern Recognition

**Development velocity (GitHub)**
- **Last 24–48h (Jan 3–4):** **1 PR opened / 1 merged**, **2 new issues**, **4 active contributors**.
- **Month-to-date (Jan 1–Feb 1 window snapshot):** **2 PRs merged**, **16 new issues**, **5 active contributors**, **~160 additions / 36 deletions / 17 commits**.
- **Change profile:** fixes + refactors outweigh new features; notable focus on **reliability/perf** (MultiStep provider execution) and **correctness** (SQL data isolation).

**Notable code trends**
- **Performance/throughput:** PR **#6263** parallelizes provider execution in MultiStep (`Promise.allSettled`) → likely lower latency for multi-provider agents; aligns with community interest in multi-model agents.
- **Operational quality:** new **logging linter** in `eliza/config` + enhanced logging → increasing emphasis on observability and standardization.
- **Data isolation stability:** PR **#6316** addresses a PostgreSQL `SET LOCAL` parameterization failure that “broke all DB operations” when `ENABLE_DATA_ISOLATION=true` → indicates maturity work on multi-tenant / isolation features.

**Community engagement patterns (Discord)**
- Support load concentrates around **core developer(s)** (Stan) answering deployment + model routing questions; **Borko** covers docs; **peer support** by Omid Sa appears in troubleshooting.
- Repeating themes:
  - **“How do I configure multiple models/providers?”** (OpenAI + Anthropic via OpenRouter)
  - **“How do I deploy with custom plugins?”** (containers/cloud)
  - **“Why doesn’t X plugin/version work?”** (Discord plugin version drift)
  - **“Why can’t my agent recall config/bio?”** (knowledge/lore regression or config mismatch)
  - **“Is ElizaOS more than a wrapper?”** (positioning skepticism)

**Feature adoption signals**
- **OpenRouter plugin** is acting as the “recommended path” for multi-provider configuration (high leverage; currently under-documented).
- Early pull for “agent-as-product” funnel ideas (agentic AI crash course) suggests marketing-led adoption initiatives are emerging in parallel to core engineering.

---

### 2) User Experience Intelligence

**Feedback themes (categorized by impact)**

**High impact (blocks usage / breaks trust)**
1. **Website integration error: “Model not found”** (API key + agent ID flow unclear or brittle)
   - Impact: blocks embedding agents into websites; direct revenue path risk for Cloud.
2. **Agent memory/config regression: missing “knowledge & lore” or inability to recall bio**
   - Impact: core agent UX feels unreliable; increases “it’s just a wrapper” skepticism.
3. **Plugin versioning inconsistency (Discord plugin v1.3.3 → v1.3.5, failed publish v1.3.4)**
   - Impact: onboarding friction; increases support burden; creates perception of instability.

**Medium impact (conversion / usability)**
4. **Public agent UX and wallet flow issues (GitHub issues)**
   - #6317 “Connect wallet should go straight to wallet options”
   - #6318 “Scroll should work on whole page”
   - Broader public-agent funnel issues also present in month-to-date list (separating public agent states, message limits, chat summaries quality, agent cards w/ chat count, free credits).
   - Impact: conversion + retention; especially for unauthenticated visitors coming from social.

**Lower impact (feature expansion / nice-to-have)**
5. **Twitter poll creation capability**
6. **Calculation/forecasting functionality guidance** (unclear whether via tool/function, model choice, or plugin)
7. **Crypto earning in Hyperscape / DegenAI update chatter** (community interest; unclear product fit)

**Sentiment and narrative**
- Positive: “Linux of autonomous AI agents” framing resonates; “190 plugins” cited as differentiation.
- Negative/at-risk: skepticism that ElizaOS “wraps models” unless reliability + composability are felt in first-run experience.
- Sentiment driver: UX friction (public agent flows, wallet connect, broken integrations) amplifies doubts more than missing features.

**Usage vs intended design (observed mismatch)**
- Users want **task-based model routing** (Anthropic for calc, OpenAI for reasoning) inside one agent; current guidance is “configure via env / OpenRouter” but lacks a clear “policy” abstraction users can understand (rules, fallback, per-tool model selection).
- Users attempt **website embedding** and expect a simple API-key-driven integration; errors like “Model not found” indicate hidden coupling between “agent ID”, provider configuration, and environment.

---

### 3) Strategic Prioritization (Impact × Risk × Dependency)

#### P0 — Fix conversion blockers + reliability regressions (1–3 days)
1. **Triage and resolve “Model not found” website integration**
   - Deliverable: reproduce + fix, or publish a canonical integration recipe + clearer error messages.
   - Dependency: API endpoint validation, agent/provider resolution logic, docs.
   - Success metric: reduction in Discord “model not found” reports; successful embed sample app.

2. **Stabilize agent config/memory expectations (“knowledge & lore” / bio recall)**
   - Deliverable: compatibility note + migration guide if schema changed; add validation warnings in CLI/UI when fields are ignored.
   - Risk: medium (touches config parsing + memory initialization); high trust payoff.
   - Success metric: fewer “can’t recall bio” reports; improved first-chat coherence.

3. **Plugin versioning integrity for Discord plugin**
   - Deliverable: audit release pipeline, enforce semantic version + published artifact verification, add “latest compatible versions” table.
   - Success metric: decreased install/version mismatch questions; fewer support pings.

#### P1 — Make observability “real” across ecosystem (this week)
4. **Roll out logging linter adoption plan**
   - Deliverable: checklist + CI recommendation for plugins/projects; add a “logging standard” page.
   - Dependency: eliza/config linter integration examples.
   - Success metric: reduced log format variance; faster debugging turnaround; fewer “black box” complaints.

5. **Document multi-model / multi-provider patterns (OpenRouter-first)**
   - Deliverable: 2–3 copy-pastable examples:
     - “Reasoning vs tools/calc” routing pattern
     - fallback + retries
     - per-environment model selection (dev vs prod)
   - Success metric: fewer repeated Discord questions; increased successful deployments.

#### P2 — Optimize top-of-funnel UX (next 1–2 sprints; product + frontend)
6. **Public agent experience rework (unauth vs authed vs owner)**
   - Align with the open issues cluster (separate states, message limits, chat counts, summaries quality, credits changes).
   - Rationale: this is the **highest-leverage** area for growth; fixes skepticism by making value immediately visible.
   - Success metric: visitor→signup conversion, chat continuation rate after gating, reduced bounce.

7. **Wallet connect flow improvement**
   - Implement #6317 and validate with first-time user testing.
   - Success metric: connect completion rate; time-to-first-chat.

---

### Recommended Resource Allocation (next 7 days)

- **1 backend engineer (primary):** website integration “Model not found” + agent/provider resolution + error messaging.
- **1 platform engineer:** plugin versioning + release validation; logging linter CI guidance.
- **1 developer experience writer/engineer:** multi-model docs + cloud container deployment docs; publish “known-good” templates.
- **1 frontend engineer (part-time):** quick wins (#6318 scroll, #6317 wallet options) while scoping the broader public-agent state redesign.

---

### Watchlist / Leading Indicators

- **Support load:** number of daily Discord questions repeating “multi-model”, “deploy plugins”, “model not found”, “discord plugin version”.
- **Adoption:** OpenRouter plugin usage as default multi-provider path (proxy via docs clicks, template forks).
- **Reliability perception:** frequency of “wrapper” skepticism messages after UX fixes ship.
- **Funnel health:** public agent page engagement once state separation + message gating is implemented (conversion + retention).