Before ChatGPT, the bet was already that intelligence is composition.

MindHYVE.ai began as HYVE Labs in the summer of 2022 — months before the world had heard of ChatGPT. The lab's working thesis: a single model is incomplete; intelligence is a composition of cooperating models. HYVE Labs is still the named R&D division of MindHYVE.ai, Inc. today, and the architectural choices it locked in 2022 are the ones JustineAI runs on.

1. Summer 2022

   HYVE Labs

   Bill Faruki begins a research lab around a single intuition: an intelligent system would not be a single model — it would be a composition of cooperating models, each calibrated to a different cognitive role. The lab is named HYVE Labs. HYVE is an acronym: Hyper Yielding Vectorized Entity. This is before ChatGPT exists.

2. November 2022

   ChatGPT launches

   The world becomes aware of large language models. The HYVE Labs thesis — that a single LLM is incomplete; intelligence is composed — survives intact. The chatbot moment proves the consumer demand; it does not change the architectural opinion.

3. Early 2023

   The name

   "Mind full of HYVEs." The founder observes that his own working memory holds many parallel structures — a HYVE for product, a HYVE for finance, a HYVE for the customer in front of him. The name MindHYVE.ai follows from the metaphor.

4. Summer 2023

   Eve-Fusion

   The architecture firms into a compositional fabric — a classifier, a domain-specialised small reasoning model, and one to three frontier-grade models cooperating per task. The lab’s fifth working version is christened Eve-Fusion v5. The F5/reasoner short-form follows.

5. 2024

   The portfolio takes shape

   Four Agentic Operating Systems are scoped: Arthur for education, Chiron for healthcare, Justine for law, Theo for theology. Each runs on the same compositional fabric, calibrated to its discipline. The Digital Employees are named.

6. 2025

   Justine enters field trial

   JustineAI™ Personal Injury Edition enters field trial with founding firms across the United States. The supervisor + sub-agent pattern crystallises around the PI workflow. Phone intake emerges as a discrete platform category.

7. 2026

   Today

   JustineAI™ PI Edition shipping to founding firms. Five additional editions on the roadmap (workers’ comp, medical malpractice, mass tort, insurance bad faith, employment class actions). The supervisor pattern proven; Eve-Genesis (Law Edition) trained and deployed.

The decisions that have held since week one.

• The intuition that came first

  An intelligent system is a composition, not a model. That was the working sentence in 2022 and it is the working sentence today. Every architectural choice MindHYVE has made since — Eve-Fusion™, the F5/reasoner pattern, the Digital Employee model, the Eve-Genesis™ training-data discipline, the supervisor + sub-agent pattern — descends from it.

• The acronym

  HYVE: Hyper Yielding Vectorized Entity. The phrase reads like a research artefact because it is one. The lab gave itself a name that was descriptive in the literal sense — high-yield, vector-native, an entity rather than a tool.

• Mind full of HYVEs

  The name MindHYVE comes from a moment Bill noticed about himself. A founder running a company holds several models in parallel — a product mental model, a customer mental model, a finance mental model, a hiring mental model. They run independently and cooperate. It looked, to him, like the architecture he was building.

• Why this matters for Justine

  JustineAI is the same compositional intuition rendered into legal practice. The F5/reasoner orchestrates. The frontier models consult. Justine the Digital Employee inherits the disposition to reason analogically, abductively, and dialectically because Eve-Genesis (Law Edition) trained her to. The founding metaphor is still doing structural work.