i.Wonder Laboratories, Inc.

A place to wonder.

i.Wonder designs and develops advanced compound AI systems and immersive open-world platforms, redefining how people create and collaborate.

Apply for Early Access Start here Open roles · 6

00 — Start here

Four ways in.

For investors, candidates, partners and early users. Pick the door that fits.

Investors

Thesis & research

The company thesis, the full research and architecture paper, and the compact brief.

 Candidates

Founding roles

Six founding roles are listed now as Founding-Team Interest, remote within the United States.

 Partners

Julius architecture

How the Compound AI Agent’s three permanent tiers and Project Brain fit together.

 Early users

Wonderland in motion

A simulation of Julius and the AI companion system working inside Wonderland during a mission.

01 — Mission & vision

AI creation shouldn’t end when the chat window closes.

Mission. i.Wonder builds multiplayer open-world AI creation platforms — and researches, develops and deploys Compound AI Systems and Compound AI Agents.

Most AI creation still happens in isolation: one conversation, one agent, a coding tool that forgets the project the moment the session ends. The work is real. Nothing around it lasts.

“AI should stop being something people merely talk to and become something people can live, build, collaborate, compete, and create with inside a persistent world — powered by Compound AI Agents rather than a single agent or a monolithic model.”

Core thesis — Compact Research & Architecture Brief

Vision — three shifts

  1. 01Isolated chat boxesA shared, persistent world
  2. 02Single modelsCompound AI Agents with accountable tiers
  3. 03Disconnected coding toolsLong-lived projects that remember

Our thesis: AI creation becomes a world — a place where people and Compound AI Agents build, collaborate, compete, create software, and maintain what they make, together, over time.

  • Principle 01

    Completion is earned, not claimed.

    Work is finished when independent review and verification say so — not when an agent says it is.

  • Principle 02

    Architecture over vendors.

    Tiers are defined by responsibility. The best technology for a tier can change; the tier does not.

  • Principle 03

    Honest by default.

    What isn’t known is shown as unknown. What is simulated says so. What is still in development is labelled.

02 — Flagship platform

Wonderland

A persistent multiplayer open world for AI creation.

In active development

Projects don’t reset. Worlds don’t vanish.

Wonderland is an open, persistent world where AI-powered creation and real software development happen in the same place as collaboration, competition and play. What you build is still there tomorrow — and so are the people and agents who built it with you.

Its role is to turn AI creation into a persistent, shared and motivating experience instead of a one-off chat interaction.

This section describes Wonderland’s intended design. It is in active development.

See Wonderland in motion

  • 01

    AI-powered creation

    Shape ideas into worlds, objects and experiences with AI as a creative collaborator.

  • 02

    Software development

    Real software, built by Compound AI Agents inside the world and verified before it’s called done.

  • 03

    Persistent worlds

    Places, projects and progress that outlive the session that created them.

  • 04

    Collaboration

    People and agents working in one shared context, with clear roles and a shared memory.

  • 05

    Competition

    Challenges that put ideas, builds and agents to the test.

  • 06

    Social systems & game mechanics

    Identity, progression and community that give long-lived creation its momentum.

Key product ideas — from the research brief

Wonder Web

The in-world browser and workspace for opening, inspecting, testing and using what gets created, without leaving Wonderland.

Customizable AI companions

Julius and related agents are presented through a user-facing identity layer, including the Relay Dog concept.

Competition & evaluation

Gamified Verification Environments, agent-versus-agent flows and evaluation battles that make quality and performance visible.

03 — Flagship Compound AI Agent

Julius. One agent. Three permanent tiers.

Julius is i.Wonder’s flagship Compound AI Agent and the intelligence inside Wonderland. It is built from three persistent architectural tiers — each a different class of AI technology, each accountable for its own responsibility, all governed by Julius.

Compound AI Agent

Julius

Governs every tier

  1. Tier 1

    Prompt Architect

    Thinks before anything is built.

    Technology class
    Large language model
    Purpose
    Mission understanding, reasoning, architecture, decomposition, research, continuous guidance
  2. Tier 2

    Coding Agent

    Builds inside the codebase.

    Technology class
    Agentic coding system
    Purpose
    Codebase operation, implementation, tool use, tests, environment interaction
  3. Tier 3

    Independent Reviewer

    Answers to the evidence, not to the builder.

    Technology class
    Independent agent harness / reviewer runtime
    Purpose
    Independent research, adversarial review, evidence inspection, defect discovery, acceptance verification
Every tier reads and writes

Project Brain

Julius’s durable memory — a persistent record of the actions, decisions, evidence, state and history accumulated throughout a mission or project. Nothing is cleared when the mission ends; the next mission begins from this record.

The tiers are permanent; the providers are not. Each tier is a slot defined by its responsibility and filled by replaceable implementations — never a dependency on any one AI vendor. The Independent Reviewer always runs apart from the Coding Agent. This is Julius’s architecture as designed; the research papers set out the evidence still to be gathered.

What Julius governs

  • 01Mission authority
  • 02Orchestration
  • 03Dependencies
  • 04Permissions
  • 05Budgets
  • 06Project Brain & context
  • 07Synthesis
  • 08Conflict resolution
  • 09Verification
  • 10Termination

04 — Wonderland in motion

A system in motion.

Julius and the AI companion system operating inside Wonderland during a mission.

Simulation Wonderland — Scenario B 4:30
Open video

05 — Research & reference

Context, in full.

The founder’s research and architecture paper, and its compact companion.

  • Full research paper

    Compound AI for AI Creation

    The full i.Wonder research and architecture paper covering Julius, Compound AI systems, Project Brain, Wonderland, verification, and the company’s technical thesis.

    Kaisin Rogodfree, CEO v1.0 · 6 Sep 2026 23 pages · PDF
    Read full paper
  • Compact brief

    i.Wonder Compact Research & Architecture Brief

    A shorter overview of i.Wonder’s core thesis, Julius, Wonderland, Compound AI, and the problem the company is solving.

    Kaisin Rogodfree, CEO Companion to the full paper 4 pages · PDF
    Read compact paper

The full paper is a research-grounded architecture and evaluation agenda. In its own words, “it is not an empirical performance report for Julius”, and “no Julius benchmark results, production user study, or industry-wide priority claim is asserted.” Its validation agenda centres on measurable outcomes: verified task completion, cost, latency, regressions, human effort saved and repeat use.

06 — Founding team

Join the founding team.

Recruiting is open now for six founding roles across Wonderland and Julius.

  • Founding-Team Interest
  • Remote (United States)
  • Proposed equity per role
  • No visa sponsorship

i.Wonder is an early-stage company: Wonderland and Julius are in active development. Each role lists proposed equity, shown on the role: four-year vesting with a one-year cliff (proposed terms, subject to final written agreements). Founding-Team Interest — Equity-Based Opportunity. i.Wonder Laboratories is currently identifying prospective founding-team members. Cash salaries have not yet been established. These listings are expressions of interest in future positions, not offers to begin unpaid employment. Any work arrangement will require a legally compliant agreement. Future team members are expected to make a serious time commitment once employment arrangements are finalized. All roles on one page.

Early access

The next world is being built.

Apply for early access to Wonderland and Julius. We are choosing a small first group of testers for the Private Alpha and Beta, starting with Wonderland. A person reads every application.

Apply for Early Access Privacy notice

07 — Contact

Talk to i.Wonder.

Careers & early access
  • Investors & partners — questions about the thesis, the research and the company.
  • Candidates — apply online on the role you want.
  • Early users — apply for early access to Wonderland and Julius.