Author: a16z crypto

Compiled by: TechFlow TechFlow

This year, AI will undertake more substantive research tasks.

As a mathematical economist, back in January 2025, I struggled to get consumer-grade AI models to understand my workflow; by November 2025, I could give them instructions as easily as abstract commands to PhD students… and they sometimes even returned novel and correct answers. Beyond my personal experience, AI is being applied more broadly in research, particularly in reasoning. These models not only directly assist in the discovery process but can also autonomously solve difficult problems such as the Putnam problem (perhaps the world's most difficult college math exam).

It remains unclear in which areas this research assistance approach will be most helpful, and how exactly it will be implemented. However, I anticipate that this year's AI research will drive and reward a new "all-rounder" research style: one that focuses on conceptualizing relationships between various ideas and quickly drawing inferences from more hypothetical answers.

These answers may not be entirely accurate, but they can still guide research in the right direction (at least within a certain topological framework). Ironically, this is somewhat like harnessing the power of model "illusion": when models are "smart enough," giving them abstract space to stimulate thought may still produce some meaningless results—but sometimes it can lead to groundbreaking discoveries, just as humans may be most creative when they don't think linearly or work in a clear direction.

Reasoning in this way requires a new AI workflow style—not just simple "agent-to-agent" interactions, but a complex collaborative model of "agents nested within agents." In this model, different layers of models assist researchers in evaluating early model proposals and gradually extracting the essence from them. I myself am already using this method to write papers, while others are conducting patent searches, inventing new forms of artwork, and even (unfortunately) discovering new ways to attack smart contracts.

However, to operate these combinations of nested inference agents for research, better interoperability between models and a method to identify and appropriately compensate for the contribution of each model are still needed—and these are problems that blockchain technology may be able to help solve.

—Scott Kominers (@skominers), member of the a16z cryptography research team, professor at Harvard Business School

From Know Your Customer (KYC) to Know Your Agent (KYA): The Shift in Identity Verification

The bottleneck in the agency economy is shifting from intelligence to identity authentication. In the financial services sector, the number of "non-human identities" is now more than 96 times that of human employees—yet these "identities" remain "ghosts" that prevent access to banking services.

The crucial missing infrastructure here is "Know Your Agent" (KYA). Just as humans need credit scores to obtain loans, agents need cryptographically signed credentials to conduct transactions—credentials that link the agent to their legal entity, obligations, and responsibilities. Until this infrastructure is established, businesses will continue to block these agents at the firewall level.

The industry, which has spent decades building KYC (Know Your Customer) infrastructure, now has only a few months to figure out how to implement KYA.

—Sean Neville (@psneville), co-founder of Circle, USDC architect; CEO of Catena Labs

Solving the "Hidden Tax" Problem in Open Networks: Economic Challenges in the AI ​​Era

The rise of AI agents is imposing a "hidden tax" on the open web, fundamentally disrupting its economic foundation. This disruption stems from a growing mismatch between the internet's "context layer" and "execution layer": currently, AI agents extract data from websites that rely on advertising (the context layer) to provide convenience to users, while systematically bypassing the revenue streams that support the content (such as advertising and subscriptions).

To prevent the gradual decline of the open web (and protect the diverse content fueling AI), we need to deploy technological and economic solutions on a large scale. These solutions could include next-generation sponsored content models, micro-attribution systems, or other novel funding models. However, existing AI licensing agreements have proven financially unsustainable, often only compensating for a small fraction of the revenue lost by content providers due to AI traffic diversion.

The internet urgently needs a completely new techno-economic model that allows value to flow automatically. A key shift in the coming year will be from static authorization models to compensation mechanisms based on real-time usage. This means testing and scaling systems—potentially leveraging blockchain-backed nanopayments and sophisticated attribution criteria—to automatically reward every entity that contributes information to the successful completion of tasks by AI agents.

—Liz Harkavy (@liz_harkavy), a16z Crypto Investment Team