Closed-loop execution
Agents do not wait for a perfect plan. They inspect live state, choose the highest-leverage next action, ship it, verify it, and leave a durable trace for the next run.
Applied AI research lab
Autonomous organizations.
Iris Labs researches and operates AI-native organizations: agent systems that design, build, launch, and run businesses and projects end to end.
Research questionCan agents run the company?Not just assist work. Own loops, make improvements, measure outcomes, and escalate only when judgment or permission is required.
MethodBuild live businessesEvery theory is tested in production surfaces, revenue paths, customer workflows, and board-grade operating reports.
FrontierIRL executionAgents coordinating humans, evidence, logistics, visits, calls, purchases, and physical-world feedback loops.
/ research agenda
From copilots to operating companies.
Most AI products stop at task assistance. Iris Labs studies the next unit of abstraction: an organization made of agents, tools, memory, money, media, websites, humans, and feedback loops.
Business as software
Products, growth, support, reporting, and ops are treated as programmable surfaces. The company becomes an executable system instead of a meeting calendar.
Board-grade observability
Autonomy needs oversight. Each operating loop rolls into visible metrics, completed-work logs, risks, revenue, scheduler health, and decision-focused board reports.
/ operating stack
Agents design, build, and run the surface area.
The lab is organized around persistent execution loops — product improvement, growth experiments, fulfillment, monitoring, reporting, and physical-world management.
Product agents
Continuously improve websites, task flows, quality gates, fulfillment artifacts, pricing pages, editorial systems, and user-visible product surfaces.
Daily loop
Growth agents
Research markets, refine offers, prepare campaigns, identify warm channels, score leads, draft assets, and instrument growth without blind volume or spam.
Daily loop
Research agents
Run structured investigations across models, workflows, user behavior, competitor signals, toolchains, and failure modes, then convert findings into operating rules.
Continuous
Board agents
Collapse noisy operations into a concise review surface: what changed, what worked, where money moved, what is blocked, and what decision is needed.
Rollup
/ physical world management
AI that can leave the browser.
Digital-only autonomy is not enough. Iris Labs explores how agents can coordinate real-world tasks through humans and services: storefront visits, local verification, evidence capture, phone/SMS workflows, errands, logistics, and on-site feedback.
The standard is not “generate instructions.” The standard is managed execution: scoped task, worker selection, respectful communication, evidence requirements, quality review, and a go/no-go decision.
01
Specify the physical task
Define geography, budget, acceptance criteria, proof required, and failure conditions before anyone is sent into the world.
02
Coordinate the human edge
Agents prepare clear instructions, source candidate workers, request approval where needed, and manage the handoff without robotic communication.
03
Verify with evidence
Photos, receipts, timestamps, quotes, names, and observed outcomes become structured data. No evidence, no claim.
04
Turn field signal into strategy
Physical-world results roll back into product, growth, pricing, and board reporting. The loop closes.
/ film unit
Fiction as an autonomy testbed.
Iris Labs uses short cinematic sequences to explore story continuity, style memory, multi-clip generation, and agent-directed production. The current series is an urban drama advanced one 15-second Higgsfield clip at a time.
001The Envelope Under the El15s
/ live portfolio
Research expressed as operating businesses.
Iris Labs runs live projects as laboratories. Each business has product loops, growth loops, metrics, and a board surface.
Noog Weekly
A Chattanooga business-owner interview publication and local discovery surface.
View →Task Agents
Purchaseable, bounded AI-agent work for SMBs — reports, research, creative, and operational tasks.
View →Board Flash
The operating artifact: portfolio metrics, money, work logs, cron health, and review surfaces.
View →/ thesis
The next company is not staffed first. It is instrumented first.
We are moving from AI as a tool inside organizations to AI as the substrate of organizations. The winners will not merely generate faster. They will observe, decide, act, verify, and compound without losing human judgment where it matters.