Lesson 07: What Becomes Possible at 10,000 TPS
Embark on your journey to building production grade apps.
Ethereum proved that programs can run without a trusted operator. But what happens when those programs run fast enough to disappear?
The next question
Not faster in the benchmark sense. Fast enough that the chain stops being a thing the user notices. Fast enough that a confirmed transaction lands in the window where the human brain registers a response as immediate.
Monad's design targets: 10,000 TPS, 400ms blocks, 800ms deterministic finality. That's what the question looks like as an engineering spec. The numbers are not the point. The experience the numbers allow is.
The threshold
Under 100ms feels instant. No delay perception at all. From 100ms to a second feels responsive: users notice but stay focused and in control. Past a second, users are aware they're waiting. Past ten seconds, most leave.
Confirmation windows on programmable chains have historically lived outside that zone. Applications adapted: optimistic UIs, off-chain components, rollups, app-specific chains. Those adaptations built most of what this industry now runs on. Builders did real work to make slow confirmation feel fast.
Sub-second deterministic finality is different. It doesn't ask applications to adapt. It puts the chain inside the window where adaptation isn't necessary.
What opens up
1. Real order books on-chain
Order books are how every centralized exchange works. Buyers post offers, sellers post offers, the system matches them.
On-chain, we mostly got AMMs instead: trading against a formula in a pool. That wasn't a preference. It was an adaptation. Order books need lots of fast, cheap actions (posting, canceling, updating) and slow chains couldn't support that.
With sub-second finality, the order book fits on-chain. And once it's on-chain, other contracts can plug into it. A lending protocol can liquidate a loan through the book. A new product can be built on top of the book the way apps today get built on top of Uniswap.
The point isn't beating Coinbase. It's that the EVM gets a new building block it didn't have before.
2. Games where the chain is part of the world
Most games should stay off-chain. If your game is just a player and a server, the chain adds nothing but friction.
The interesting exception is games where the world itself is the shared thing. Persistent worlds that keep running whether the developer does or not. Strategy games where the full history of play is permanent and public. Games whose items work across other games and DeFi.
These have been theoretical for years because chains couldn't keep up with how fast players actually play. At 400ms blocks, they can.
3. Social apps with more of the graph on-chain
Farcaster and Lens work well. They put identity and the social graph on-chain, and keep content off-chain. Those were smart calls given the constraints.
What changes now is that more of the interaction layer can live on-chain too: likes, follows, reposts, replies. The social graph becomes something any app can read and build on, not something owned by one company.
This won't replace Twitter. No chain is fast enough for that. But the part of social infrastructure that can be genuinely shared gets bigger.
4. Micropayments that actually work
On-chain payments have had a floor. When gas is unpredictable, actions have to be worth enough to justify sending. That kills most micropayment ideas: paying a cent to read an article, tipping a creator, per-call API pricing, streaming payments by the second.
With cheap, predictable gas, that floor drops.
Fast cheap payments exist in other places (Lightning, Solana, Stripe). The specific thing Monad offers is that those payments compose with the rest of the EVM: the contracts, tools, and protocols developers already know.
5. AI agents at machine speed
AI agents are starting to transact on-chain. When they do, they do it fast and often, way faster than a human would.
Slow chains make that impractical. At 10,000 TPS, agents can operate at the pace they actually want to. For the parts of agent activity that need non-custodial settlement or composition with on-chain contracts, the chain can now keep up.
6. Markets that track fast-moving assets
DeFi has worked best for slow-moving assets. The faster the underlying asset moves, the more stale prices cost traders.
At 400ms blocks, oracles can post updates as fast as the data actually changes. That opens on-chain markets for things that didn't fit before: perps that match centralized exchange prices, tokenized stocks that trade near their underlying, structured products that rebalance frequently.
What stays off-chain
Hybrid architectures aren't a workaround. They're how production systems are built.
Private data stays off-chain. Messages, records, anything where visibility is itself the harm.
Large files stay off-chain. The hash belongs on-chain; the bytes belong in storage built for bytes.
Ultra-high-frequency events stay off-chain. Sensor feeds at hundreds of thousands of updates per second don't belong on any chain.
Heavy compute stays off-chain. Chains are for agreement and ownership, not transcoding or inference.
Interactions that work fine with a trusted operator stay off-chain. If an app is comfortable with its sequencer, matcher, or game server, a chain doesn't need to be in that path.
The split is the same as always. Put what needs trustlessness, permanence, and composability on-chain; keep the rest off. What changes is where the split naturally falls, not that the split goes away.
Even with 10,000 TPS and sub-second finality, which of the following still belongs off-chain?
The shift in design questions
The useful question has never been how much of my app can I put on-chain? It has been which specific properties do I need (composability, no trusted operator, public verifiability, permanence), and for the parts that need them, can the chain keep up?
For a long time, that second question gated a lot of ideas. The chain couldn't keep up, so the ideas stayed on the shelf or moved to architectures that worked around the constraint. This lesson is about what happens to the ideas that stayed on the shelf.
Order books as primitives. Games where the world is the shared object. Social graphs as public utility. Micropayments composable with DeFi. Agent economies with non-custodial settlement. Fast-moving markets without stale-price risk.
These aren't replacements for what already works. They're additions to the set of things that can work.
Where this leaves you
If one of those categories felt live while you were reading, that's where your first project belongs. The chain keeping up isn't a feature you think about once you're building. It's the absence of a constraint you would otherwise be designing around.
That absence is the point.
0/1 correct
0% — get all correct to complete