From open-source agents to personal AI and community-centered infrastructure, Emad Mostaque’s Intelligent Internet is exploring who will own—and benefit from—the next wave of artificial intelligence.
Artificial intelligence is becoming an essential tool for business, research and creativity. Yet a deeper question sits behind the race to build increasingly capable models: who will control the systems, and how widely will their benefits spread?
From Stable Diffusion to a larger ambition
Emad Mostaque became widely known as the founder and former CEO of Stability AI, whose work helped popularize openly available generative-image technology. His next venture, Intelligent Internet, asks a broader question: as AI becomes part of nearly every profession and institution, who will control it—and who will benefit?
The company’s answer is not a single chatbot. It is a collection of agents, creative tools, retrieval infrastructure and ideas about locally accountable AI. The vision is expansive; its commercial success is not yet established. But the work offers a useful map of where AI may be heading.
II-Agent: moving from answers to action
Intelligent Internet’s open-source II-Agent is designed to handle practical, multistep work, including research, software development and content creation. The company reported scores of 61.8% on Terminal Bench and 45.1% on SWE Bench Pro with its February 2026 V1 release. These are company-reported benchmarks, not independently reproduced results.
For entrepreneurs, the important distinction is between an AI that gives advice and an AI that can use tools, carry out tasks and produce artifacts. The latter could help small teams test products, analyze markets and automate repetitive work. Reliability, oversight and security remain essential when agents act on real systems.
Zenith: better systems around the model
Zenith explores the software “harness” that plans, coordinates, tests and verifies a model’s work. In September 2026, Intelligent Internet reported that Zenith using DeepSeek V4.1 Flash achieved strong results on its AutoResearchExam evaluation at a cost of roughly $2.82 per task. The company’s own evaluation should not be treated as a universal performance ranking.
The strategic idea is compelling: a smaller, less expensive model guided by a strong workflow may outperform a more expensive model on certain tasks. For businesses deploying agents at scale, cost per successfully completed task may matter more than the model name on the invoice.
Genii: a personal assistant in everyday messages
Introduced in July, Genii is a personal AI assistant designed to live inside iMessage rather than requiring a separate app. Intelligent Internet describes an assistant that can remember context, communicate naturally and help coordinate tasks across everyday tools.
The product addresses a real adoption problem: people do not necessarily want another dashboard. They want help in the places where they already communicate. Whether such assistants become indispensable will depend on trust, privacy, permissions and dependable execution.
Factory: creative production within reach
Factory is Intelligent Internet’s generative-media workspace for developing images, video, voice and other creative assets. Its 2026 updates included website-to-video and whiteboard-explainer workflows, reusable assets and custom skills.
For independent creators and small businesses, reducing the time between an idea and a usable creative asset can have real economic value. The opportunity is not simply more content, but faster experimentation and more accessible production.
Evoke: the less visible infrastructure that matters
The company’s October 2026 Evoke for Postgres announcement focuses on combining semantic and keyword retrieval inside PostgreSQL. Intelligent Internet says the system uses a 30-million-parameter CPU-capable model and avoids a separate embedding pipeline.
This is a less flashy but potentially important problem. Agents working across documents, conversations and business records need to retrieve the right information consistently. Simpler retrieval architecture could make long-running AI applications easier and less costly to maintain, though independent performance testing is needed.
Beyond products: who owns the intelligence age?
Mostaque has also written about “Champions”—locally rooted public-benefit institutions intended to keep relationships, responsibility and economic value close to the communities served by AI. This is a proposal for institutional design, not an established system.
It raises worthwhile questions: Will the greatest benefits of AI accrue only to the owners of compute and distribution? Can open infrastructure make small businesses and communities more capable? How should governance work when intelligent systems influence daily life? Answers will require more than technological optimism; they will require sound incentives, accountable institutions and measurable outcomes.
Hear Emad Mostaque on Moonshots
Readers who want to hear Mostaque explain his views directly can find him on Peter Diamandis’s Moonshots podcast. In the October 7, 2026 episode “Why Altman Says ‘Accept Some Bad Things Happening’ and What It Means for AI Safety,” Mostaque joins a discussion about AI consciousness, safety and the possibility of a major push toward superintelligence. He also appeared in an August 2026 conversation on open models, compute economics and the AI race.
These long-form discussions are valuable complements to company announcements: they let listeners hear his arguments and decide for themselves which forecasts are persuasive.
Why Charlotte should pay attention
Charlotte’s banks, universities, startups, agencies and local businesses all have a stake in how AI develops. For a small company, affordable agents could expand what a lean team can accomplish. For developers, open tools could reduce dependence on a single platform. For investors, the opportunity is to distinguish impressive demos from products with adoption, retention and sustainable economics.
Intelligent Internet is worth following because it is pursuing both practical AI tools and a larger question of human agency. Its ideas are ambitious and their outcomes uncertain. The most useful response is neither uncritical enthusiasm nor dismissal: explore the tools, evaluate the evidence and ask how intelligence can help more people create genuine value.
Explore the work
Intelligent Internet · Company blog · Research and releases · Genii announcement · Moonshots episode with Emad Mostaque
Editor’s note: This independent overview draws on Intelligent Internet’s public announcements and the Moonshots podcast. Company-reported benchmarks and product capabilities have not been independently tested by Creative Loafing Charlotte. No endorsement or commercial relationship is implied.