Skyfall AI Introduces the World’s First AI CEO, Steve, and Announces $2M Startup Program
Skyfall AI, enabling the future where an entire business can operate autonomously, led by Sam Pasupalak, Kaheer Suleman, and Sumit Pasupalak, has introduced Steve, the world’s first autonomous business harness that can run digital businesses such as SaaS products and Shopify apps. Alongside the launch of Steve, Skyfall AI is announcing a $2 million Startup Program, through which Skyfall will acquire 5 additional startups to run completely autonomously on Steve.
Steve, the world’s first AI CEO, combines World Models, RSI, simulation, and specialist AI agents to manage business operations, continuously learning from experience to improve decisions and performance. Human owners retain control over strategic direction and critical decisions, while Skyfall plans to expand Steve across more business types.
Testing Across Different Business Needs.
Skyfall has deployed Steve across three businesses, with early results showing encouraging progress. The first business recorded 2.3x revenue growth over the last quarter. The second, after five years of declining revenue and high customer churn, has stabilized revenue and started acquiring new customers. The third joined the platform just a week ago, making it too early to assess results.
These businesses have different revenue patterns, customer needs, and operating pressures. They give Skyfall different conditions in which to test Steve's world models, agent coordination, and self-improvement loop.
Public Dashboards for Anonymized Businesses.
Skyfall plans to launch public dashboards for each business alongside the announcement, with identifying details withheld. Each dashboard will provide visibility into business goals, metrics, agent activity, decisions, scheduled jobs, and harness updates, offering readers a view of how Steve manages operations, adapts plans, acts within permissions, and continuously improves its own systems.
World Models Enable Continuous Planning and Replanning.
Steve uses executable code-based world models to map relationships between customers, product usage, revenue, and costs, drawing on business records and inputs. These models allow Steve to simulate potential actions, estimate outcomes, and continuously plan and replan operations around business goals.
For example, Steve can compare advertising strategies by simulating potential sales and costs before committing resources. Each proposal includes a forecast, time frame and confidence level, while the policy engine validates actions before execution. Actual results are then compared against forecasts, with any differences informing model updates and subsequent plans.
Steve uses recursive self-improvement (RSI) to update its code, skills, agent configurations, tests and world models. By linking problems and missed forecasts to specific system components, Steve records changes and evaluates their impact on subsequent runs.
These improvements can reshape task allocation, agent coordination and testing processes, with the policy engine ensuring all changes remain within the owner's rules.
Two examples demonstrate this process:
Improving customer emails: Steve tested different lifecycle email approaches, identified the best-performing versions, and updated its instructions to improve future campaigns.
Resolving agent deadlocks: When growth and engineering agents repeatedly passed a website deployment task between themselves, Steve updated its orchestration system to identify and resolve coordination conflicts, preventing repeated delegation loops.
A Policy Engine Governs Execution.
Before execution, Steve's policy engine evaluates each decision against owner-defined rules and learned policies, considering action type, spending, and customer impact. Actions are assigned one of four outcomes: automatic execution, execution with notification, approval required, or blocked. Each decision retains a record of the applicable rules and checks.
For example, refunds of up to $30 could be processed automatically, while larger amounts require owner approval. This framework allows routine operations to continue within predefined permissions while keeping business-critical decisions under human control. All system improvements remain subject to the owner's goals and permissions.
Expanding the Portfolio Through a $2M Startup Program.
To accelerate testing across diverse software and digital business models, Skyfall has allocated $2 million to acquire five early-stage startups. Founders interested in transitioning their products to fully autonomous operations powered by Steve can apply to join the program.
