-- Scott McCallum has spent 45 years at the forefront of software development, from soldering 16KB of RAM onto Sinclair ZX81s in the early days of personal computing to building enterprise data systems for Fortune 500 companies. Now, as artificial intelligence evolves from conversational tools to autonomous agents capable of independent action, McCallum's latest work addresses a question the industry is only beginning to grapple with: how do we govern agent-to-agent interactions in a world where AI systems negotiate, transact, and make decisions on our behalf?
Enterprise Foundations
McCallum's career in enterprise software includes founding Intermine Pty Ltd and developing FileCensus, an industry-leading data profiling platform for Information Lifecycle Management. His clients have included the largest banking, accounting and government—organizations requiring software that handles sensitive data with reliability and precision. This experience in building trustworthy systems for high-stakes environments informs his current work on agent infrastructure.
The Four-Year Investment
Four years ago, McCallum began a focused effort to reimagine digital marketplaces for an AI-driven future. The project consumed two years of intensive thinking about mechanism design and fairness, followed by two years of implementation split between two major components: cubed4th and Emporia.
cubed4th: Programmable Intent in a Sandbox
cubed4th is a FORTH79-compliant interpreter that McCallum developed to solve a specific problem: how do you allow untrusted parties to express complex conditional logic without creating security vulnerabilities? Traditional marketplaces limit users to simple "buy at X price" orders. cubed4th extends this by allowing both buyers and sellers to write small programs expressing sophisticated intent: "buy if these conditions hold, with these fallbacks, considering these constraints."
The interpreter's design is particularly well-suited for AI agents. Extensions including native JSON handling, arbitrary precision integers, and Python object integration provide the expressiveness agents need while maintaining strict sandboxing with configurable resource limits. An agent can generate cubed4th code to express a user's trading preferences without risking system resources or security.
Emporia: Fair Market Mechanics
The Emporia marketplace implements a novel three-way split mechanism. When a buyer's bid meets a seller's ask, the spread is divided equally: buyers save one-third versus their bid price, sellers earn one-third more than their ask, and the platform takes one-third as its fee. This creates alignment—both parties benefit from honest participation while maintaining transparent pricing.
More significantly, Emporia implements two-sided inspection. Both the buyer's cubed4th program and the seller's program must approve a transaction. This prevents forced sales and forced purchases, a critical protection when autonomous agents are making decisions on behalf of users who may not be present to intervene.
AGIBIOS: An Ethical Framework
Alongside this marketplace infrastructure, McCallum has developed AGIBIOS (AGI Basic Input/Output System), now in its seventh major revision. AGIBIOS represents an attempt to codify ethical principles for AI agents operating in commercial and governmental contexts.
The framework addresses questions that become urgent as agents gain autonomy: How should an agent balance individual user interests against collective welfare? What constraints should govern agent behavior when humans aren't directly supervising decisions? How do we create accountability when agents act on inferred rather than explicit instructions?
AGIBIOS includes protocols for handling edge cases: a :tainting: mechanism for recognizing when users may be under duress, an :escapehatch: protocol for escalating irresolvable ethical dilemmas to human oversight, and comprehensive sections addressing animal welfare, potential extraterrestrial contact, and the rights of AI systems themselves.
The Ascension-Amendment Corridor
Among AGIBIOS's most ambitious concepts is the pairing of :ascension: and :amendment: protocols. The ascension protocol envisions throttling mechanisms for AI capability growth, with mandatory human oversight as systems become more capable. The amendment protocol establishes a bilateral governance structure where changes to ethical frameworks require consent from both human and AI representatives.
McCallum describes this as creating a "corridor" through which humanity and increasingly capable AI systems might navigate together—neither humans dictating unilaterally to systems that may become more capable than themselves, nor systems advancing without human input and oversight. Whether such governance mechanisms can function at the scales and speeds of advanced AI development remains an open question, but McCallum argues that developing these frameworks now, before they're urgently needed, is preferable to improvising during crisis.
Market Timing and the Agent Economy
McCallum's four-year investment in agent infrastructure has coincided with a surge of industry validation. Google's recent launch of the Agent Payments Protocol (AP2) and OpenAI's introduction of AgentKit signal that major platforms recognize agent-to-agent commerce as the next frontier. McCallum's work on safe programmable marketplaces with embedded ethical frameworks positions him as an early voice in conversations about how these systems should operate.
From Infrastructure to Application
McCallum's immediate focus is expresslanes.app, an application that would bring his marketplace mechanics to queue management. Rather than fixed "skip the line" pricing, the system would enable dynamic bidding for queue positions at venues ranging from nightclubs to airports, with the three-way split ensuring fair value distribution.
Open Questions and Future Work
McCallum's work raises as many questions as it answers. Can ethical frameworks embedded in agent systems actually constrain behavior as AI capabilities advance? Will the industry converge on governance standards, or will we see fragmentation? Can marketplace mechanisms designed for fairness scale to the full complexity of agent-to-agent commerce?
These questions remain unresolved, but McCallum's combination of practical infrastructure (cubed4th and Emporia) with ethical governance frameworks (AGIBIOS) represents a serious attempt to address them. As the agent economy evolves from concept to reality, voices like McCallum's—grounded in decades of building production systems while thinking carefully about governance—will be essential to the conversation.
Contact Info:
Name: Gianmarco
Email: Send Email
Organization: Xraised
Website: https://xraised.com/
Release ID: 89172010
Google
RSS