Nightframes: Charles Hoskinson's Pitch for What Comes After the Smart Contract
At TOKEN2049 and Midnight City Tokyo, Charles Hoskinson called today's smart contracts "pretty dumb" and introduced Nightframes: private computation on Midnight, built for a world of AI agents. Here is what we know so far.
Charles Hoskinson spent the past week telling two audiences the same thing: the smart contract has run its course, and he thinks he knows what replaces it. He calls the replacement Nightframes.
He introduced the idea on stage at TOKEN2049 in Singapore on October 7, then gave it a full hour at Midnight City Tokyo in a talk titled What comes after the smart contract? Welcome to Nightframes, posted on October 10. Here is what has been said so far, what it means and what is still missing.
The argument: smart contracts are too limited
In Singapore, Hoskinson called today's smart contracts "pretty dumb". His complaint comes down to two things.
They cannot compute much. Every node on a public chain re-runs every contract, so anything heavy is too slow or too expensive to put on-chain.
They cannot keep a secret. Everything a contract touches is visible to everyone. That works for a public token swap. It does not work for payroll, medical records, a credit decision or a business deal.
Most of what companies and people actually do needs both real computation and privacy, which is why so little of it has moved on-chain.
Where Nightframes sit in the story
The Tokyo talk frames this as a generational step.
Bitcoin gave us digital money. Ethereum made transactions programmable. Cardano was his third-generation design. Nightframes are his answer to what comes next, and he pitches them as a new kind of computing as much as a new kind of crypto.
What a Nightframe is
Based on coverage of the Singapore talk, a Nightframe is a unit of private computation built on Midnight that combines three technologies:
- Zero-knowledge proofs prove a statement is true without revealing the data behind it. You can prove you are over 21 without showing your birthday.
- Multi-party computation lets several parties compute a result together while each keeps its own inputs private. Coworkers can work out their average salary without anyone revealing their own.
- Trusted execution environments are sealed areas of a processor where code runs out of view of the rest of the machine, including its operator.
The heavy or sensitive work happens off-chain or in a private layer. The chain settles the result and holds the proof that the work was done correctly. You get the guarantees of on-chain settlement without publishing every intermediate step.
Each of the three has a weakness on its own. Zero-knowledge proofs are slow for big computations, multi-party computation needs coordination, and secure hardware asks you to trust a chip maker. The bet is that together they cover for each other.
The part that makes it different: agents
This is where Nightframes go beyond "private smart contracts." In Tokyo, Hoskinson described them as the agents of Midnight City, the simulated city where autonomous AI agents already work, trade and generate real Midnight transactions and zero-knowledge proofs. He called Nightframes the most social things ever built, because they are designed for thousands of agents working together, and described the project as built for AI agents first.
Read that way, a smart contract is a fixed program that waits to be called. A Nightframe is closer to a private workspace that an agent can act inside: holding confidential data, doing real computation, working with other agents and proving to the outside world that it followed the rules.
He gave two examples of where this leads. One is microfinance, where a loan repayment could feel like paying with cash, stay fully private and still prove the facts a lender needs. The other is agentic trading inside Midnight City, where he expects users to be able to fund agents that run trading strategies across other chains within the next few months. Both are forecasts, not shipped features.
What we do not know yet
It is worth being clear about the gaps.
- There is no public specification. Everything above comes from talks and press coverage. Some reports call the same stack "Night Mode", and it is not yet clear how the names relate.
- No dates. Permissionless contract deployment only reached Midnight mainnet on October 2. Nightframes are a direction, and timelines are Hoskinson's estimates.
- The engineering is hard. Combining three privacy technologies is complex, secure hardware has a history of vulnerabilities, and developers will need to learn Midnight's Compact language.
Why we are paying attention
We think the diagnosis is right. In our own work, the two reasons a business decides against putting something on-chain are almost always the same two Hoskinson named: the data cannot be public, and the logic is too heavy for a contract. If Nightframes deliver private computation that agents can use and outsiders can verify, a lot of ideas that never made sense on a blockchain start to.
We run infrastructure on Cardano and build with AI agents every day, so this sits right where our work meets. If you are weighing what private smart contracts or AI agents could do for your business, that is the kind of problem our consulting team takes on.
We will follow this one closely and report back when there is a specification to read.
This post is for information only and is not financial advice.