Artificial intelligence is entering a phase where the main constraint is no longer model capability but the lack of economic and governance infrastructure, according to a16z crypto.
In a series of forward-looking assessments, members of the firm’s research and investment teams argued that AI systems are now capable of performing substantive work, from academic research to financial tasks, but lack the identity, attribution, and compensation mechanisms required to operate at scale.
As a result, a16z positions blockchain-based systems as infrastructure rather than speculative technology.
AI Moves From Assistance To Discovery
Scott Kominers, a member of the a16z crypto research team and a professor at Harvard Business School, said AI models have progressed rapidly over the past year.
He described a shift from systems that struggled with abstract instructions to models capable of reasoning through complex problems in ways comparable to advanced graduate students.
Kominers said AI is increasingly being used for discovery in reasoning-heavy fields, including mathematics, where models have demonstrated the ability to solve elite university-level problems.
This has enabled a new research style focused on conjecture, synthesis, and exploration, even when intermediate outputs are imperfect.
However, he noted that these workflows often rely on layered systems in which multiple models evaluate and refine each other’s outputs, raising unresolved questions about interoperability and how contributions should be recognized and rewarded.
From KYC To 'Know Your Agent'
Another challenge highlighted by a16z is identity.
Sean Neville, cofounder of Circle and chief executive of Catena Labs, said the agent economy is now constrained less by intelligence and more by trust.
Neville said non-human identities already outnumber human employees across financial systems, yet most remain effectively “unbanked.”
He argued that just as financial institutions rely on know-your-customer rules for humans, AI agents will require cryptographically verifiable credentials, a framework he called “Know Your Agent.”
Without such systems, Neville said, merchants and platforms will continue blocking autonomous agents from transacting at scale.
AI’s Strain On The Open Web
Liz Harkavy, an investor on the a16z crypto team, warned that AI agents are imposing an “invisible tax” on the open web by extracting value from content while bypassing ad- and subscription-based revenue models.
She said existing licensing deals have proven insufficient and argued that the web needs real-time, usage-based compensation systems.
Blockchain-enabled micropayments and attribution standards, she said, could allow value to flow automatically to contributors.