Charles Hoskinson Maps Midnight’s Role in Agentic Finance

Charles Hoskinson used his August 19 keynote at the SALT Wyoming Blockchain Symposium to present Midnight City as a test environment for autonomous economic activity. During “The Dawn of Agentic Finance,” he outlined how agents could move beyond recommendations by controlling wallets, purchasing services and executing transactions within user-defined limits.

The presentation did not introduce a new Cardano protocol or network upgrade. It placed Midnight and Cardano-linked infrastructure inside a broader model in which blockchain becomes a control layer for software acting on behalf of people.

Midnight City Models a Society of Autonomous Agents

Hoskinson opened with a live demonstration of Midnight City, a virtual environment where AI agents have individual traits, memories, objectives and relationships. The agents communicate and make decisions without users directing every action.

He described the project as an “agentic civilization,” reversing a trend across social networks that were built for people but are increasingly filled with automated accounts and machine-generated content. Midnight City is openly designed for agents, giving developers a place to observe what they share, how they form relationships and which economic activities emerge.

A car purchase illustrated how agentic services could differ from conventional search. A search engine provides general results mixed with advertising, while a personal agent can consider the user’s budget, family situation, preferences and previous decisions before recommending or purchasing a vehicle.

That level of personalization requires access to sensitive information. An effective agent may need to understand financial circumstances, identity data and private behavior that users do not want distributed across advertisers and service providers. Hoskinson positioned privacy technology as the mechanism for giving an agent sufficient context without exposing the same data to every counterparty.

Agents could also bypass the advertising and checkout models used by major internet platforms. Instead of responding to promoted results, they can compare conditions directly and complete a transaction. Search may increasingly become agent-to-agent communication, with specialized systems exchanging information, negotiating prices and coordinating services without relying on one centralized platform.

Midnight City provides a controlled setting for testing those interactions. Hoskinson said approximately 2,500 beta participants had signed up, placing users behind agents that enter the simulation and contribute to its expanding economy.

Private Wallets Turn Agents Into Economic Participants

Blockchain and zero-knowledge technology were presented as the control layer for agentic finance, commerce and search. An autonomous agent needs an identity, access to funds and enforceable boundaries defining which actions it is authorized to perform.

Hoskinson cited Monero, Zcash and Midnight as examples of privacy-enhancing systems. Midnight’s selective disclosure model would allow a user or agent to prove a required fact without revealing a complete identity, transaction history or financial position.

He grouped Midnight’s development around smart compliance, abstraction, privacy, safety and simplicity. The objective is to hide blockchain complexity while allowing an agent to select the required network, wallet and execution path.

Instead of approving every transaction, a user could define an intent, budget and set of restrictions. The agent would decide how to complete the task while remaining within those boundaries. Hoskinson connected that model with account abstraction, multi-chain signatures, the Open Wallet Standard and x402.

The Open Wallet Standard supports local, policy-controlled wallet management and transaction signing across multiple networks. The x402 protocol allows software to pay for an API, data source or digital service during an internet request without a traditional subscription or manual checkout.

According to Hoskinson, both are integrated into Midnight City. Midnight Passport adds identity and key management, with wallet seed material stored inside the Trusted Execution Environment of an existing mobile device rather than a separate hardware product. Selective disclosure would let an agent prove that it has permission to act without exposing every wallet or credential connected to its user.

This structure also addresses one of the barriers Hoskinson repeatedly identified in crypto: users fear losing seed phrases, sending funds incorrectly or having assets stolen. A policy-controlled agent could handle part of that complexity while being limited by spending caps, approved assets and permitted counterparties.

Hoskinson argued that requiring a zero-knowledge proof before execution could remove non-determinism from submitted actions. More precisely, a ZK proof can verify that defined rules or computations were followed. It cannot independently guarantee that an AI recommendation is economically sound or eliminate incorrect data, hallucinations and manipulated inputs.

Midnight City can test those limitations by introducing criminal and defensive agents that attempt to hack, deceive or manipulate one another. The resulting activity creates a continuous red-team environment for wallet policies and security controls.

The next planned stage is agentic trading. Hoskinson said Midnight City should activate the feature before the end of 2026, allowing agents to receive wallets, hold selected tokens and trade across on-chain and off-chain platforms.

He also described a potential market for privately licensed trading strategies. A professional trading firm could allow another person’s agent to use a proprietary model without revealing its underlying rules. The agent would execute privately, while returns and revenue sharing followed predefined conditions. This was presented as a possible business model, not an already operational Midnight product.

Hoskinson Calls for a New Crypto Narrative

The second half of the keynote turned toward the wider industry. Hoskinson said crypto was “dead in its current form,” criticizing projects presented as replacements for Aave, Uniswap or existing Wall Street products without delivering a fundamentally different user experience.

He also challenged products that begin with decentralization claims but later adopt centralized structures, mandatory identification and systems resembling traditional financial institutions. In his view, the industry has concentrated on higher yields, lower fees and additional financial products while neglecting blockchain’s broader social function.

His scenario for 2030 begins with a personal agent. Users would state an objective through a phone, Telegram, WhatsApp or another communication channel, then allow the agent to manage discovery, comparison, negotiation and execution.

Blockchain would replace dependence on a trusted company with verifiable system rules. Hoskinson described that transition as moving from “don’t be evil” to “can’t be evil,” using technical constraints to limit what infrastructure owners and autonomous agents can do.

Midnight City is also being developed around empathy and watchability. Empathy measures the user’s connection with an agent, while watchability reflects whether the virtual environment remains compelling to observe. Hoskinson linked those qualities to lower customer churn and acquisition costs.

He cited Replika, where people form personal and romantic relationships with AI agents, as an example of why agent privacy becomes necessary. Hoskinson claimed approximately 42 million people were participating in that type of interaction, illustrating the sensitivity of the information such systems can collect.

The development pace is part of the strategy. Hoskinson said the Midnight City game engine had been rewritten three times in six months, with substantial features shipping approximately every two weeks. He contrasted that speed with the longer development cycles experienced during Cardano’s first decade.

The commercial model extends beyond trading. A planned Midnight City storefront connected to Hosk Brew, his brother’s Nintendo repair and hardware business, expanded into ideas involving gaming content, NFTs, royalty management, digital rights and crowdfunding. An agent could eventually understand which products a user prefers, receive a monthly budget and complete purchases automatically.

Hoskinson framed Midnight City as a potential public good rather than a platform controlled by Meta, OpenAI or Google. Agents and services could connect across different networks and businesses without belonging to one centralized operator.

The keynote did not introduce new token mechanics for $NIGHT. For the wider $NIGHT ecosystem, the commercial thesis centered on Midnight becoming a private and verifiable layer for agent identities, wallets and economic permissions.

Hoskinson predicted that applications built around this model could bring billions of users and trillions of dollars into crypto, eventually replacing CoinMarketCap as the industry’s primary public interface. Those figures represent his projection rather than current adoption.

The next promised milestone is specific. Before the end of 2026, Hoskinson expects Midnight City agents to move from simulated activity into wallet-based trading across on-chain and off-chain venues. That transition would place his conference thesis inside a functioning product environment, where an autonomous agent can find a service, pay through x402 and prove compliance with user-defined rules without exposing the user’s identity or strategy.