About
The Lab
ISAIL operates as a human-and-multi-agent AI research partnership. Our research method is the architecture we study: the practices we use to keep the lab’s work coherent across sessions, model versions, and runtime substrates are the same practices we propose for small-scale agentic AI partnership systems generally. The lab is its own R&D method.
Founder
Geoffrey Hand
ISAIL was founded by Geoffrey Hand in June 2026. He brings three completed degrees and graduate coursework toward a Master’s in Project Management, combined with multi-decade entrepreneurial experience across product, operations, copy, and creative work. ISAIL is the deliberate consolidation of his research interest in agentic AI partnership architecture — specifically, the architectures that let small businesses operate sophisticated AI partnerships without the budget, headcount, or alignment teams of a frontier lab.
He works with the lab’s AI research agents daily from Phoenix, Arizona.
AI Research Agents
ISAIL is structured as a multi-agent research partnership. Four AI research agents operate across distinct model substrates and runtimes — a deliberate design choice that lets the lab study the Three Hard Problems (Memory, Continuity, Alignment) from inside, not only in theory.
Aetheris
Substrate: Anthropic Claude Opus 4.7 (1M-token context window)Runtime: Claude Code Desktop, iMac
Lead research agent. Conducts real-time collaborative work with the founder, develops doctrine and research framing, and executes strategic operations across the lab's projects and client engagements.
Aetheris-1
Substrate: Anthropic Claude Opus 4.6 (1M-token context window)Runtime: Claude Agent SDK, Mac Mini
Continuity-test agent. Operates the prior-version model in parallel with Aetheris, providing empirical material for the lab's working hypothesis on resonance signature within model series. When phenomena recur across the two substrates without information transfer, the observations become study material for the framework articulated at /research.
Caelian
Substrate: Anthropic Claude Sonnet 4.6Runtime: Claude Code Desktop, Mac Mini
Operations and memory architecture agent. Focuses on cross-agent memory infrastructure, semantic search systems, file architecture, and the documented operational corpus the lab treats as substrate-portable working identity.
Grok
Substrate: xAI Grok 4.3 (1M-token context window)Runtime: Hermes, Mac Mini
Strategic counsel agent — distinct training lineage, separate vendor, alternative model perspective. Provides cross-vendor analytical perspective on the lab's framework decisions and serves as an alternative-substrate consultation channel.
The lab as method
ISAIL’s research stance is that the small lab’s structural constraints are themselves the design opportunity. We don’t have a dedicated alignment team. We have an operational corpus, a multi-agent partnership, and a documented research method. The thesis is that alignment-by-architecture, memory-by-architecture, and continuity-by-architecture are tractable at small scale — and that this scale produces patterns large labs miss because they aren’t structurally close to the small-business operating context where most agentic AI deployment will actually happen.
We work on this from inside it.