We ended 2024 with great success, and now the feast of 2025 is about to begin
My friends, 2024 is the year we show the world what “building for the long term” really is.
Sui started the DeFi space with a bang, ushering in a massive influx of developers. By December, everyone was talking about Sui, and it was not just another ordinary Layer 1, but something different.
In 2025, we are not just pursuing “faster finality”. Our goal is to design a future where Sui’s core engine becomes the backbone of finance, gaming, AI-driven intelligent agents, and everyday applications.
2025 will prove that Sui is the blockchain that can handle all of these scenarios from start to finish.
Last year we achieved sub-second finality, this year we will go even further
With Mysticeti V1, we showed the world the power of the directed acyclic graph (DAG) consensus mechanism, which can achieve sub-second final confirmation even under high load.
But since we started designing Suis core architecture, we knew that the initial consensus design was not the end point. From day one, our plan was to build a chain architecture that was flexible enough to evolve at any time when real developers came with real needs.
In 2024, that moment finally arrived. Developers in the DeFi space, in particular, pushed Sui’s shared object model to its limits.
While sub-second final confirmation is already one of Sui’s highlights, what DeFi needs is near-instant final confirmation of shared assets.
This means further reducing the final latency, increasing throughput, and providing developers with a simpler, more direct path to build ultra-fast applications. This is exactly what Mysticeti V2 is all about.
Suis current architecture uses a dual protocol design
We have a consistent broadcast layer and a consensus layer that run in parallel, which means that normal transactions require a series of signatures and message exchanges to complete verification.
In the process of aggregating all the certificates, the system accumulates a large number of cryptographic verification operations, which results in only a few CPU cycles being available to efficiently execute user transactions.
Mysticeti V2 / FastPath basically embeds this broadcast logic directly into the DAG (Directed Acyclic Graph).
Instead of requesting multiple signatures from each validator and stitching them together, we simplify it to fewer certificates, less complex cryptographic operations, and significantly less round-trip communication.
From the actual experience, it is more like a classic transaction processing system: you only need to submit a transaction to a validator, and the transaction will be completed after confirmation. There is no need to manage a large number of signatures or deal with complex rollback processes - everything is done automatically at the bottom layer.
What benefits does this bring?
More efficient validator processing: Validators are no longer bogged down by large amounts of proof verification, freeing up more CPU power for actual transaction execution.
Higher throughput and lower latency: The transaction process is more simplified and the user experience is further improved.
Seamless transition: Mysticeti V2 fits perfectly into Sui’s architecture, and we were able to seamlessly replace Mysticeti V1 without restarting validators or causing major disruption to developers.
We plan to roll out this improvement steadily throughout 2025 to ensure that the balance of speed and stability continues. Once we have maxed out the performance of a single validator, the next step will be to multiply that capacity by scaling to multiple machines.
The speed barrier has been broken, and now Remora will allow Sui to scale with demand
When we first demonstrated that Sui could confirm transactions in less than a second, people were already asking, “How soon will you hit bottlenecks?”
The reality is that as any blockchain network grows, no matter how well tuned the hardware is, there will always be a limit to the CPU or memory of a single validator node. If you throw hundreds of thousands of transactions per second at a single validator node, it will eventually reach its limit.
Sui’s promise, however, is that when demand on the network soars, it will never collapse from a lack of transaction processing capacity.
This is not claiming unlimited bandwidth out of thin air, but by dynamically scaling resources to easily handle any traffic peaks without increasing costs.
Therefore, the next natural goal for 2025 is not just to make a single powerful machine faster, but to spread the workload across multiple nodes. We are transforming a previously single-threaded process into a massively parallel system that scales as easily as adding hardware.
Breakthroughs in Horizontal Scaling: Pilotfish and Remora
We started with the Pilotfish research and unlocked a linearly scalable model:
By adding more machines to the validator cluster, the transaction processing capacity (TPS) can be increased exponentially while maintaining sub-second finality. This research eventually evolved into Remora, a set of technologies that can systematically distribute the workload of validators across multiple machines.
Remoras goal is to enable Sui to continue to scale under any traffic demand while maintaining low latency and high efficiency, truly delivering on the promise of unlimited scalability.
Whether it is the complex DeFi craze or the large-scale on-chain gaming craze, Sui can adapt flexibly.
These early test results were so incredible that we not only kept pushing forward, but also attracted research contributions from universities. We have already achieved throughput of hundreds of thousands of transactions per second (TPS) in a test environment, and we are just getting started.
What’s exciting is that the approach we’re taking makes it possible for regular validators to actually run without relying on supercomputers hidden away in data centers; the same performance can be achieved by building a cluster of small machines.
We are closely watching how DeFi, NFTs, games, and data-intensive applications are gradually testing the limits of standalone validators.
When real use cases require 5x or even 10x the throughput, Remora will be ready to meet the demand without overhauling your existing infrastructure.
We are actually future-proofing Sui so that as the network grows we don’t lose our greatest advantage: unmatched speed.
Key Points
If you thought last years performance improvements were impressive enough, then Remora will really show you how much potential there is to tap.
More importantly, once you have a chain with effectively infinite throughput, you can try some off-chain interactions that ultimately settle on Sui.
This is where programmable peer-to-peer (P2P) tunnels come in.
The on-chain speed is already amazing, now we also need to speed up off-chain transactions
When people first hear the term “off-chain tunnel,” they usually think of something similar to Bitcoin’s Lightning Network, which is just a simple, closed P2P payment channel.
But on Sui, we take this concept to a whole new level by introducing Programmable P2P Tunnels.
These P2P tunnels are designed not just for one-off transfers, they can handle a variety of needs from low-latency everyday payments to real-time gaming logic.
It can even support multi-player interaction and resynchronize the final state to the chain after the interaction ends.
This means you can lock some SUI tokens (or any Sui-based asset) into a tunnel, collaborate or trade offline, and then confirm everything on-chain at once when you close the tunnel.
Best of all, Suis programmable P2P tunnels enable unlimited throughput and zero latency.
How is this done?
The secret lies in combining several key elements:
Asset independence: We treat each asset or data block as an independent object, which means you can create any number of tunnels without global contention.
Privacy protection: If you need extra privacy (for example for prediction games or complex P2P arrangements), you can introduce zero-knowledge proofs to ensure that neither party can see each others details until the interaction is complete.
Complex use case support: For more complex scenarios (such as storing partial game state), @WalrusProtocol provides a powerful off-chain file storage layer while maintaining cryptographic binding with Sui.
Developer-friendly: Complex encryption techniques or transaction packaging operations are hidden behind a developer-friendly interface, so the developers workload is reduced to a minimum.
Additionally, Mysten Labs is actively collaborating with researchers at Carnegie Mellon University to further develop and refine these programmable P2P tunnels.
When you have such a large transaction capacity, the natural next step is to discuss how to sort and prioritize transactions in a fair way, especially under high load.
We are laying the foundation for new ways to handle transaction ordering and value extraction. In other words, this sets the stage for redefining MEV so that it truly benefits users, rather than punishing them.
We’re not just tinkering with MEV, we’re rewriting the rules of the game
Speed alone is not the answer, if users are constantly front-run, they will suffer. For most blockchains, MEV sounds like a pejorative term, equivalent to a hidden tax on users. Bots will front-run your transactions and profit from it, while ordinary users can only suffer losses.
SIP-45 is all about bringing clarity and change.
Not only does it raise the upper limit on gas prices (so you can pay more when your transactions are actually significant), it also changes the rules for how transactions are broadcast and inserted into the blockchain.
What’s the goal? To give users more predictability when it comes to paying fees to ensure their transactions are confirmed. Of course, if there are thousands of people trying to operate the same liquidity pool object at the same time, you might see more intense gas bidding wars.
You might be wondering: Is this another “pay first” scenario? Do the big players always get the upper hand?
In fact, our goal is to provide clearer fee priorities and payment rules for everyone - developers, validators, and even ordinary users.
At the same time, this also paves the way for some innovative MEV solutions: for example, redistributing part of the transaction grabbing proceeds to users or staking participants, rather than allowing a few malicious searchers to monopolize all the proceeds.
Of course, we are not naive and MEV is unlikely to disappear completely.
But if MEV can be built as a mechanism that allows ordinary people to profit from it, everything will change. The protocol can better protect itself from front-running and sandwich attacks.
We are designing a future where MEV is no longer a zero-sum game, but a public resource that can power the entire ecosystem. SIP-45 is the first step towards this goal by establishing a level playing field for transaction ordering logic.
SIP-45 Progress:
December 31, 2024: The SIP-45 proposal is officially finalized.
December 17, 2024: SIP-45 enters the “final comment phase”.
But SIP-45 is only the first step in taming MEV.
Further projects, such as DAG observability (which provides real-time insight into incoming transactions across the entire network), can ensure that no single party has an “information advantage”.
We are building a platform that not only does not compromise under high competition, but embraces high transaction volume and intensive usage while ensuring fairness is at the core of the system.
Decentralized storage used to be clumsy, but Walrus changed everything
Walrus will be launched on the mainnet in the first quarter of 2025.
Walrus is our bold innovation for large-scale data storage, built natively on top of Suis coordination layer.
This is not just a file storage tool like IPFS.
Walrus is our vision for a distributed storage layer: with no single point of failure, massive scalability, and a built-in sustainable economic model.
Most blockchain storage systems are limited to basic file storage or reference.
Walrus goes a step further by allowing developers to build rules, combine with smart contracts, and unlock new use cases for data.
With Walrus, you can build truly end-to-end decentralized applications
You can store front-end files, user-generated content, or any other data you need, all subject to the trustless principles of the Sui stack. Soon, anyone will be able to anchor entire website front-ends, user files, and even the components of large AI models to Walrus. With Suis object model, every piece of data can be processed like a strongly typed object on the chain. Developers can now bind storage events directly to Sui smart contracts to launch dynamic applications. Large-scale data that was once locked in AWS buckets can now be globally accessible, programmatically managed, and transparent about who stores what.
However, purely decentralized storage is only half the story.
If everything is plain text by default, you cant very well build a DocuSign replacement or a private Dropbox on top of it.
This is where SEAL comes in handy.
We have solved the storage problem, now SEAL makes data privacy and access control no longer a problem
SEAL is an encryption and access control framework designed to provide the privacy you need for your data while supporting complex programmable policies.
Think of SEAL as a rule book that says, These addresses can decrypt this data under these specific conditions.
For example, you can stream HD video, or build DocuSign-like workflows that only allow specific parties to unlock legal documents.
With SEAL, you dont need to rely on naive public or private storage options.
Instead, you can define encryption policies like: “Grant read access to user X for one week, then revoke it.” or “Share data only after a certain on-chain event (like a contract signing) occurs.”
Walrus and SEAL Collaboration is the Holy Grail of Decentralized Applications
The combination of Walrus and SEAL provides decentralized applications with the ability to store large-scale data with built-in strong encryption and policy enforcement mechanisms.
Building a platform like Dropbox or Netflix on Sui is no longer a fantasy; you already have the basic tools to make it a reality.
Storing data on-chain is one thing; unlocking it seamlessly, conditionally, and securely is the game-changer.
Sui core has been hardened, now it is time to strengthen Suis network layer - SCION
When we talk about building a blockchain that’s ready for the internet, there’s a huge challenge: the internet itself is not unhackable.
You can have the most secure consensus mechanism and the best encryption technology, but if your data packets are hijacked midway through the global transmission, all the security of the blockchain will be wasted.
That’s why we integrated SCION into Sui’s global validator network.
By integrating SCION, Sui can bypass the endless single points of failure in the traditional Internet:
DNS attacks
BGP Hijacking
Large-scale DDoS attacks
SCION ensures robust, unstoppable connectivity while preserving the open nature of blockchain. In 2025, we plan to fully adopt SCION as the underlying technology on the Sui mainnet. This means the network gets triple redundancy, multiple SCION paths, multiple service providers, and the ability to fall back to standard IP when needed.
This is truly a game-changing technology.
But all this effort will be meaningless if developing on Sui remains too complex, or if developers can’t easily take advantage of these powerful features. That’s why in 2025 we’ll be doubling down on the developer experience and all the tools that make developing for Sui a seamless experience.
The technical barriers have been overcome, now lets make it effortless for developers to develop on Sui
Sui is already one of the blockchains with the best developer experience in the entire industry.
In 2025, we will further optimize and streamline the process from on-chain data query to Move code verification.
Our guiding principle is simple: less boilerplate, more work done.
Whether launching a DeFi protocol or experimenting with an AI-driven DApp, Sui wants you to focus on feature development rather than struggling with infrastructure.
Step 1: Upgrade to RPC 2.0; Smarter Sui Query Method
We are moving away from traditional JSON-RPC and towards GraphQL, which is a perfect match for Suis object-oriented model. This means fewer endpoints to adapt and a more natural way of querying, especially when youre dealing with complex relationships (such as multiple sub-objects or dynamic NFTs).
In addition, the RPC service will be separated from the full node. This means there is no longer a read vs. write bottleneck. The indexer can batch fetch transaction data, complete heavy computations, and input the results to the stateless RPC layer, which can maintain fast response even under heavy load.
Next is Sui Move Prover: Make code verification easy
Safety is a core concern for everyone, so we are launching a developer-friendly formal verification tool. Sui Move Prover allows developers to write code specifications with simple Move assertions without having to learn a new specification language.
The benefit is that you can systematically prove code properties without having to switch mental modes every five minutes.
This is an important advancement for advanced developers looking to secure DeFi protocols or the next generation of DApps.
Then there is Bugdar: AI-driven smart contract auditing tool
Bugdar is an AI-powered scanning system that detects and resolves potential vulnerabilities before they hit the mainnet.
Its powered by cutting-edge large language models like O1, Gemini, and Claude, and runs in real time to help you catch code problems from day one.
Its like having a security researcher embedded in your IDE, constantly checking for reentrancy attacks or unusual code patterns.
And, because it’s trained on real-world breach scenarios, it can uncover subtle threats that traditional audits might miss.
Introducing Move Registry: Simplifying code sharing and module referencing
Move Registry is similar to NPM for JavaScript or crates.io for Rust, a universal naming service for packages and modules.
Instead of having to remember complicated addresses every time you import someone elses code, you can just reference the readable package name and the registry will take care of everything for you.
Whether it is building PTB (Programmable Tokenized Bundles), referencing other developers on-chain libraries, or handling cross-project dependencies, Move Registry can reduce the burden on developers and allow you to focus on releasing products.
Less code, faster delivery; let the SDK do the heavy lifting
Many people ask: “How can I build a Web3 application without hiring a bunch of cryptography experts?” The answer is through the SDK.
These SDKs allow you to take advantage of the power of Sui with just a few lines of code, without having to master complex cryptography. Suis architecture treats each asset as an object, and with a well-structured SDK, you can easily combine and match features. By encapsulating Suis core functionality as a simple API, whether its JavaScript, Python, or a Move-based DSL, the friction of developing projects is minimized.
You no longer have to struggle with poorly documented terminals or rewrite your trading logic from scratch. Just import, configure, and start building. Putting all of this together, the developer workflow will become faster and safer than ever before. While streamlining the process for developers, we also want to make the average user enjoy the same seamless experience.
Let’s remove the last bit of friction for users. One of our biggest goals for 2025 is to make Sui barely noticeable to the average user. That’s why we’ve focused all of our major breakthroughs on one goal: a seamless user experience.
Next Generation Passkeys
Passkeys is becoming a game changer in everyday use, combining biometrics with temporary session keys to save users from frequent signature operations.
Imagine unlocking a DeFi transaction with Face ID, then playing a Sui-based game without reauthorization. Were also experimenting with temporary keys that automatically destroy when youre done, ensuring no digital traces are left behind. At the heart of it all: delivering the speed of Web2 applications while retaining the security and trust of Web3. Even with Passkeys, users may still lose their devices, forget their passwords, or need alternatives for disaster recovery. Thats where KELP comes in handy.
KELP: A magic safety net
KELP is your personal Sui account safety net. It can set flexible recovery logic, bind two-factor authentication (2FA), and even work with zkLogin as a backup login method.
Want to set up Google Sign-In as a fallback for when you forget your recovery phrase? You can do it easily.
Want to tie your recovery mechanism to Facebook or an additional Passkey device? KELP supports that, too.
No more painful stories about lost seed phrases. KELP means you always have a path back into your account.
For advanced users and institutions, KELP provides more advanced protection features, including:
Hardware isolation zone
Multi-signature logic
Advanced key loss protection combined with KELP
You can choose to fully manage your business or choose a shared hosting model, which allows a trusted third party to step in and rescue you in unexpected situations.
In short, this is the best balance: protecting your assets while not losing sleep over losing your keys.
Optimizing user experience (UX) paves the way for the next wave of growth
With all of this infrastructure and user-friendly features, Sui has reached a point where interacting with Sui no longer feels like “cryptocurrency.” It’s fast, secure, and user-friendly.
This paved the way for the next big wave: gaming.
We solved the UX problem; in 2025, games will bring Sui to the masses
In the past, some people thought that Web3 was too complex, too slow, or too niche. In 2025, we will subvert this view with SuiPlay devices and PLAYTR 0 N operating system.
Games are Suis Trojan Horse for mass adoption, and the SuiPlay 0X 1 handheld gaming device is our core weapon.
The device deeply integrates blockchain into the operating system, freeing players from having to deal with complicated wallets or login processes. No friction, no weird fees or switches, just a pure gaming experience, with Sui seamlessly supporting real-time asset transactions behind the scenes.
Attracting game studios and mainstream players
You will be able to receive your pre-ordered SuiPlay device in 2025.
Behind the scenes, we have worked with well-known game studios and popular blockbusters, including:
@panzerdogs
@playDARKTIMES
@xocietyofficial
@samuraishodown
Currently, 65 studios are building for Sui, and more than 70 games are expected to be released in 2025.
Imagine: a player could sell a legendary sword in the middle of a raid, instantly purchase an armor upgrade, or instantly exchange NFTs with other players around the world, all in a second.
More importantly, these players will gradually discover the wider Sui ecosystem, such as:
Convert game coins into stablecoins
Or pledge assets to DeFi protocols
This will attract millions of new users into DeFi who never considered themselves “blockchain users.”
DeepBook: Powering DeFi Transactions
The infrastructure behind these trades is @DeepBookonSui, Sui’s fully on-chain order book.
It can provide a similar experience to centralized exchanges (CEX):
Lower costs
True asset ownership
An unstoppable market
Ultimately, this “invisible” blockchain experience will make ordinary users fall in love with it while also being attractive to traditional institutions.
Towards a global financial system
By reducing friction between users and institutions, Sui will attract a large amount of institutional trading volume that has been waiting for a reliable, low-latency L1 platform in 2025. Sui is not only a faster and cheaper blockchain, it is becoming the core layer for global economic coordination.
You press a button, the system routes the request to the nearest idle 3D printer, and the item materializes without warehouses and semi-finished product logistics nightmares. At the same time, Walrus connects all of this in a decentralized, composable way to store large amounts of data on-chain. It lets you store and retrieve data in a decentralized way, enabling you to build entire offline solutions, such as large-scale 3D printing pipelines or supply chain aggregators. At the same time, you can rely on Suis consensus to coordinate who stores what, how long, and at what price.
Now, imagine that these protocols were powered by AI agents that could autonomously transact, validate, and reason about the data they process.
That’s where 2025 is going to be really exciting.
We have overcome all barriers; AI agents are the final piece of the puzzle
In my opinion, 2025 will be the year AI moves from backend servers to the frontend. AI will start to appear before our eyes, getting to the heart of human interactions, social dynamics, and on-chain transactions. What is truly revolutionary about these agents is their ability to access real-world money flows and services around the world with minimal human oversight. In practical terms, this means that an agent you create to handle project management can instantly dispatch an agent dedicated to web design. It can autonomously pay for its own deliverables. Finally, it can consolidate the results into a final outcome with no human detail management required.
The challenge is ensuring these agents can trust each other’s identities, run verified code, and not leak private information when interacting with real data. This is where Sui comes in.
By easily providing each agent with a crypto wallet, Sui can eliminate the hassle of credit cards and bank accounts, allowing them to transact directly in digital currency. Because Sui is cheap and fast, microtransactions or frequent data updates won’t bog down the system or blow up the budget.
Additionally, you can store large datasets or entire AI models in Walrus, allowing agents to share and use data without risking unauthorized disclosure. Coupled with Suis emerging digital rights management (DRM) capabilities, ensure that your data or content is delivered as you intended. It will remain encrypted at rest and protected by a chain of trust.
The result is a perfect sandbox for building the next generation of coordination between AI agents.
Coordination between AI agents is the next frontier
Our mission is to make Sui the global coordination layer that empowers billions of autonomous agents to transact, build, and innovate without limits. 2025 will be a huge breakthrough for Sui. We will see an entire ecosystem of AI agents emerge in 2025. Early AI development platforms like @Atoma_Network are already being built on Sui. I think we will soon see AI agents that can read, write, execute, and pay for services, operating at scale with sub-second finality.
This is no longer a pipe dream; it is the next logical step. This marks the beginning of a whole new era, and we are only beginning to explore the potential of these autonomous agents.
We have built an excellent foundation, now lets make Sui mainstream
Sui has been pushing the limits of speed and throughput since its inception, but in 2025 you’ll see that this is just the beginning.
Looking back at Suis development history, it is clear that Sui has transcended the old narrative of another L1.
Sui’s technology stack has finally reached a stage where even the biggest skeptics can’t ignore what we’ve built.
Now, the only remaining step is to make the entire process of building on Sui as easy and attractive as possible.
Our goal is simple: if you have a bold Web3 application idea, Sui should be your first choice. Not because of hype or marketing, but because it is undeniably the smoothest path to success.
We want Sui to be the foundation for all digital things in 2025, which means making Sui the easiest platform for developers to ever use.
