Pi Squared is building FastSet, a decentralized settlement and payments network designed to deliver instant, trustless transactions (and verifiable computation) at internet scale without the bottlenecks of traditional blockchains. Pulse 2.0 interviewed Pi Squared founder and CEO Grigore Roșu to gain a deeper understanding of the company.
What Pi Squared Does
What is Pi Squared building and what problem are you trying to solve? Roșu said:
“Pi Squared is going beyond blockchain and building FastSet, a new type of network that can deliver on Web3’s promise of instant, trustless payments. As billions of transactions move to decentralized infrastructure, from stablecoin settlements to AI agentic payments, blockchain architecture simply won’t scale. What’s needed is a new paradigm: verifiable, parallel settlement. This is the only way we’ll be able to keep up with the demand of future economies, which will require millions of transactions per second, each costing a fraction of a cent.”
“FastSet is the fastest decentralized network for instant payments. Unlike blockchains that rely on total ordering, FastSet settles transactions in parallel, delivering uncapped throughput (currently 150K transactions per second), sub-100ms finality, and real-time verifiability.”
Preventing Double-Spending
How do you prevent double-spending without total ordering? Roșu noted:
“FastSet enforces per-account ordering using nonces. Validators must process each account’s transactions in the order they were issued, so an account cannot spend the same funds twice. All transactions from different accounts can be processed in any order because of weak independence, which ensures they still lead to the same final state. You can read more about this in the total ordering section of our documentation.”
Do We Need L2s?
Do we actually need L2s, or are they just band-aids? Roșu pointed out:
“We don’t need L2s as scalability solutions if we have networks that are sufficiently fast, cheap, and low-latency. That said, L2s have greatly influenced the architecture of Web3 apps and infra that treasures modularity and verifiability. L2s pushed us to understand how to decompose monolithic blockchains in the early days into separate modules: sequencing, computation, data availability, settlement, etc, and how to integrate them in a verifiable way. That work is highly valuable and will last a long time in Web3, but L2s as scalability solutions will vanish.”
Achieving 150K+ TPS And Sub-100 ms finality
How does FastSet achieve 150K+ TPS and sub-100 ms finality? Roșu explained:
“By redesigning the broken consensus model of blockchains. The key advantage of FastSet over blockchains is that validators DO NOT communicate with each other and only maintain their own local state as FastSet is strongly eventually consistent. As a result, claims can be settled independently and in parallel. Thus, it is the massive parallelism of claim settlements that enables FastSet to achieve such high TPS with sub-100ms finality.”
Achieving 150K+ TPS On Standard Hardware
Most high-TPS systems require specialized hardware. How did FastSet achieve 150K+ TPS on standard hardware? Roșu described:
“We achieved it by keeping on-network computation as simple as possible. FastSet validators only sign and check signatures. Any other complex computation, such as executing an EVM contract or running an AI agent, is carried out off-network, and only the final results are settled on FastSet in a verifiable way.”
FactSet Being Verified
How is FastSet being verified? Roșu assessed:
“FastSet’s core consensus protocol has been mathematically proved correct in our whitepaper. The proof is based on two core ideas. One is weak independence, and the other is conflict-free replicated data types. Weak independence says that transactions from different accounts can be executed in any order, given that they are all valid.”
“For example, suppose I’m paying you, you are paying someone, and they are paying me. It looks like there needs to be a casualty among the three payments, but if all of us have sufficient balance, then these payments can take place in whatever order and we’ll eventually reach a consensus. This is the key observation that weak independence entails eventual consistency. Then, we generalize it from payments to any conflict-free replicated data types, or CRDTs. CRDTs represent data types that are replicated across the nodes of a network, such that each node can update the data independently and concurrently, and there’s an algorithm to resolve any conflicts that might occur, and eventually all nodes will converge. FastSet basically shows that weak independence + CRDTs give you eventual consensus. No need for total ordering.”
Decentralization
If FastSet is optimized for speed, where does decentralization live? Roșu revealed:
“FastSet is fully decentralized, even more than blockchains. Because of FastSet’s parallel design, more validators do not slow down consensus because they don’t need to communicate with each other.”
Trade-Offs
What are the trade-offs that come with FastSet’s high performance? Roșu detailed:
“There are no trade-offs per se. The only possible issues are (i) the storage of claims/transactions can get huge very fast due to the ultra high speed and (ii) validators might be maintaining different sets of claims/transactions from one another since they do not communicate with each other. However, there are many existing solutions to resolve both issues.”
dApps That Will Benefit
What type of dApps will benefit the most from being built on FastSet? Roșu specified:
“Any dApp that needs quick, cheap payments would benefit from using FastSet. One good example is OmniSet, a universal liquidity hub. OmniSet connects fragmented liquidity across blockchains into one cohesive, verifiable system. Built on FastSet, it acts as a neutral coordination layer for capital, allowing assets to move, settle, and be audited across chains at 150K+ TPS with sub-100 ms finality.”
Making Payments And Micro-Transactions Viable
What does an agent-native payment rail look like and how does FastSet make payments and micro-transactions viable for AI agents and humans? Roșu outlined:
“Each agent will have either his own wallet account/s or an owner’s wallet account/s to manage. Transactions in and out of these wallet accounts can be triggered manually and/or programmatically. In any case, once we reach a world where agents are transacting on behalf of us, we will be seeing millions of transactions flowing into Web3. Thus, we need an infrastructure that can handle huge volumes.”
“FastSet makes payments and micro-transactions viable for AI agents and humans in two main ways — (i) transacting between AI agents and humans can be done at light speed and (ii) keeping the transaction cost low as huge volumes of transactions can be handled at any moment.”
Blockchain Model Flawed For Payments?
Is the blockchain model fundamentally flawed for payments? Roșu acknowledged:
“Yes, because blockchains require total ordering but in reality most payments are independent of each other. Hence, they should be able to be settled in parallel, which is what FastSet provides.”
Future
In five years, what will Pi² have made possible that doesn’t exist today? Roșu concluded:
“Super fast and cheap payment rails for AI agents and stablecoins.”
