Space and Time: where AI agents meet verifiable onchain data

Space and Time was the first Web3 company to receive funding from Microsoft’s M12 venture arm, backing that helped put the project on the map early. Since then, the company has launched its purpose-built SXT Chain and expanded its Proof of SQL technology across DeFi, institutional finance and AI. In this interview, co-founder and CTO Scott Dykstra discusses what the team has learned since mainnet, where AI agents are already using verifiable onchain data, and how Space and Time’s strategy is evolving as blockchain infrastructure matures.

QS to Scott Dykstra, Space and Time Co-founder and CTO

You built a purpose-specific chain rather than operating as a general L1 competitor. Now that SXT Chain has been live for over a year, has that architectural bet paid off the way you expected, or has the rise of things like rollup ecosystems pulled the roadmap in directions you didn’t originally anticipate?

Choosing to build a purpose-specific chain was a necessity in order to solve a technical problem for Ethereum (allowing smart contracts to query verifiable data) rather than a bet on what blockchain approach would be most popular later. Space and Time’s Proof of SQL technology needed a purpose-specific chain of nodes sitting next to Ethereum, working together to verify Ethereum data and return query results back to customers’ smart contracts. It’s almost like an oracle for querying data.

That said, the corporate-deployed rollup ecosystem blossoming over the last year (Base, Stripe’s Tempo, Circle’s Arc, Robinhood chain, etc) has been fantastic for Space and Time as we are EVM-focused. As more developers onboard to these EVM chains, we can support as the onchain database for each.

Proof of SQL lets Prover nodes handle queries and generate proofs in under a second. As the network takes on more diverse workloads, from institutional finance to student credential records in Indonesia, has that performance held steady, or have you had to make trade-offs between proof speed and the complexity of queries you can support?

Most queries that our clients run are still sub-second (which is very cheap to execute), particularly for workloads that are run constantly all day long. Examples are queries that check collateral before a trade executes.

Of course there are times when larger, complex queries over many months of Ethereum transactions can slow down processing times, but these are generally executed by a client as needed (rather than constantly run all day). An example of these more complex queries that might need to aggregate a lot more Ethereum history would be a lending protocol querying total deposits over time from a specific wallet for rewards calculations.

You’ve talked about smart contracts eventually acting as controls or safeguards for autonomous AI agents. Concretely, what does that look like in practice today: are there live examples of an AI agent’s onchain actions being gated or verified through Space and Time data before execution?

Trading onchain with AI agents is still in its infancy. This space is evolving, and we’re starting to see users who are using smart contracts to control capital allocations to sub-accounts (which was a concept pioneered by Hyperliquid). In this way, the contract can query Space and Time to check an agent’s trading performance over a time period and rescind or add more capital to that agent’s sub-account.

In the lending space, we’re seeing more vault management with AI agents and onchain queries as well. If an agent wants to move assets into a yield vault, execution is allowed only if a proof shows that the vault has maintained sufficient liquidity, stayed below a maximum historical drawdown, and has not exceeded exposure limits (via queries from a smart contract to Space and Time).

These are the two use cases we’re seeing live, but many more will likely emerge as AI agent trading becomes more prevalent.

The Genesis Validator Rewards Program onboarded 34 validators quickly. What have you learned since mainnet about validator economics and decentralization at that scale, and is 34 close to where you want the set to stabilize, or are you actively working to grow it?

At this moment there are 21 active validators, which seems to be the sweet spot for balancing network economics and security. In theory, the more validators that join the network, the more secure our data. But of course each validator shares in the rewards received by clients for inserting data and using Space and Time. By decreasing the size a bit (it’s still quite a large network, for example, a typical Chainlink oracle has a 21 node threshold for participation), the SXT network increased the staking reward rate by over 20% to a ~9.7% target annualized rate, which is hefty!

You’ve described a future “DataFi” model where contributors earn fees when their datasets get queried. How close is that to actually functioning today, is there real usage and real payouts happening, or is it still mostly a roadmap item?

We’ve put this specific roadmap item on the backburner for now, as most customers liked the concept but frankly don’t have datasets valuable enough for other customers to be interested in buying. We discovered that most of our customers’ datasets are useful to themselves, and very specific to their own use case (rather than something they want to sell via a data marketplace). This also explains why onchain data marketplaces have shut down or struggled to get adoption in the past. If this changes in the future, we’ll be ready to make “DataFi” a reality.

Dreamspace lets people generate and deploy fully functional onchain apps without writing code, with contract logic built automatically. What’s your actual review or audit process for that AI-generated code before it touches real users’ funds?

We don’t recommend that users deploy sensitive smart contracts without first getting an external audit from a real auditing firm. However, Dreamspace does a great “first pass” audit to get you testing the contract functionally with a user interface quickly. Dreamspace is an “audit” button that kicks off an audit of the contract by leveraging pages and pages of prompts we collected from top auditing teams to assess common vulnerabilities. By no means is this top-hacker-secure, but it’s a great start.

Space and Time started as infrastructure for tokenized assets, stablecoins, and institutional DeFi, but Dreamspace opens the same stack to non-technical creators building simple consumer apps. How do you think about serving both ends of that spectrum without one pulling focus from the other?

It’s important to note that Space and Time will always be the top priority for our engineering efforts. Makeinfinite Labs built Dreamspace at the infancy of the vibe coding explosion as a fun way to give nontechnical users a better UI for building onchain apps… but it’s not particularly a research moat like Proof of SQL is. Building Space and Time required five years of new cryptographic research (using novel ZK-proofs and hardware acceleration of mathematical circuits), inventing entirely new technology to solve a specific problem for Ethereum: a way for smart contracts to query proven data. The majority of the value will always be in the later, and thus our focus will be as well.