Aptos: Solving the Layer 1 blockchain trilemma

I‘ve just managed to get some down time this weekend so wanted to share a project I’m excited about — Aptos. I caveat that some of the figures used in the analysis are outdated but provide a directional sense which is still nonetheless insightful.

Recap of smart contract market dynamics

The current situation in the L1 landscape: The total market capitalization of smart contract L1 blockchains was valued at ~$400bn as of 2021 (likely declined by at least ~50% in aggregate to date). Ethereum is the largest smart contract L1 blockchain in terms of users, activity, total value locked (“TVL”). Continued demand for block space has led to network congestion and users are faced with high gas fees due to Ethereum’s capacity constraints — this is the underlying problem statement.

As of today, the current solutions to resolve Ethereum’s lack of scalability include:

1. Increase block size: Short-term band-aid solution; may compromise security as it makes blockchain node software more resource intensive.
2. L2 roll-ups: Optimistic (fraud proofs — e.g. Optimism, Arbitrum), Zero-Knowledge (“ZK”; validity proofs — Starkware, Aztec, zkSync).
3. L2 sidechains: Polygon (fka Matic Network; note uses Plasma as well as Optimistic and ZK roll-ups but technically operates in parallel to the mainnet and has its own consensus mechanism and security framework is not dependent on the L1, hence distinct to aforementioned L2 roll-ups).
4. ETH 2.0: Changing Ethereum’s architecture from a single chain to multichain, initially through execution sharding; now data sharing is being considered. Ethereum’s consensus mechanism will also change from Proof-of-Work to PoS. More on ETH 2.0 here.

It is still unclear which of the above will dictate Ethereum’s future shape and form e.g. if Ethereum were to be roll-up centric, then the L1 will only be used for data availability and security while transaction execution takes place on the L2 protocols. This technical debt has therefore seen the emergence of alternate L1 blockchains. Notably Binance Smart Chain (“BSC”), Solana, Avalanche, Algorand, Cosmos, Fantom, Near— Aptos being the newest to enter the market.

Diving into Aptos…

Aptos overview

Aptos is a Layer 1 (“L1”) Proof-of-Stake (“PoS”) blockchain designed by the original creators, researchers, engineers, designers of the Diem blockchain incubated at Meta (fka Facebook). While this is being built largely by the team behind Diem, it is not Diem itself, nor does it have any affiliation with Meta whatsoever.

Aptos aims to solve the trilemma (decentralization, security, scalability), capable for mass adoption, including enterprises. Diem was originally intended to serve Meta’s >2bn userbase — Aptos will scale beyond Meta’s network (note Aptos is “for the people” in Ohlone, California). The modernity of Aptos’ tech stack incorporates recent developments in consensus mechanisms built on a flexible, safety-first programming language — Move (discussed later).

At the time of the seed round, Aptos had initiated the devnet, with a number of companies contributing code and providing feedback — including Anchorage, Binance, Coinbase, Livepeer, Moonclave, Paxos, Paymagic, Rarible and Streaming Fast.

Team: >30 in size, all playing a critical role in Aptos, led by co-founders:

• Mo Shaikh, CEO: Formerly Head of Partnerships at Meta; previously Consensys, BlackRock, Boston Consulting Group
• Avery Ching, CTO: Formerly Principal Software Engineer having led Meta’s internal Blockchain solutions; previously Yahoo

The team benefits from their organizational continuity, efficiency and maturity — hence, is expected to execute per the stipulated plan / timeline (their testnet roadmap has delivered per planned).

Transaction

• Date: March 2022
• Fundraise, round: $200m, seed
• Tokenomics: To be announced (“TBA”) upon mainnet in Q3–2022
• Notable investors: Led by a16z alongside Multicoin, Tiger Global, ParaFi, IronGrey, Hashed, Variant, BlockTower, 3AC, Coinbase Ventures, FTX Venture, Paxos, others

Merits

Language: Move is purpose-built for safe resource management and modularity (also used to develop Diem). The language enables:

1. Easy customization of properties for assets which allows assets to flow through smart contracts as arguments and returned by functions — Solidity is less dynamic as it stores assets in hash maps that are permanently locked in a contract.
2. Referenceable transparency for immutable references — Solidity is immutable without native access control, making it relatively difficult to resolve bugs and vulnerabilities once smart contracts are deployed.
3. Memory safety by preventing dangling references and memory leaks. Move also offers a formal tool “Move Prover” which allows developers to quickly verify and test their code is executing as intended.
4. Secure storage of important information (tokens, smart contracts) through its native data type “resources”. Resources have high status in Move’s code architecture which prevents it from being copied or accidentally destroyed.
5. Multichain deployment of dApps by design. Pontem has developed a fork of the Move Virtual Machine (“VM”) which can be readily deployed to other chains: Avalanche, Cosmos, Polkadot, etc. Its roadmap includes a potential Ethereum VM that will be compatible with Move VM to facilitate deployment, interoperability and migration of dApps across multiple ecosystems.

Further, Solidity is susceptible to re-entrancy and other vulnerabilities. As a new language enhancing DX, it builds on legacy blockchain languages (Solidity) from security and flexibility perspectives. Move’s programming efficiency is understood to be comparable to Rust.

Network architecture (single shard): Multicoin, amongst others, have long been strong proponents of optimizing single shard performance. Key benefits (relative to sharding) include:

1. Reduced technical complexity and social coordination for developers;
2. Lower latency;
3. Improved composability of web3 primitives;
4. Stronger dApp communication and functionality, thus UI / UX;
5. Overall conducive to network effects, ceteris paribus, which would result in relatively concentrated liquidity.

These drove the L1 melt up in 2021 as alternatives to L2 scaling solutions, in addition to new dApps and overall ecosystem growth. Note that optimizing single shard performance and L2 scaling solutions are not mutually exclusive. Given a single shard’s performance is also a function of system capacity at a point in time, should performance reach its peak, then L2 scaling solutions can be considered.

Decentralization: Barriers to entry is used as a proxy to ascertain the degree of decentralization. The two are inversely related. This is measured through two dimensions — at minimum, the requirements are:

Hardware to run a node: Per Aptos’ recommendation:

• CPU of 4 cores (Intel Xeon Skylake or newer);
• Memory of 8gb RAM;
• Storage of 300gb.

Below summarizes requirements for other L1 peers:

Tokens for staking: Given tokenomics are still TBA and no live tokens exist today, there’s little value in comparing staking requirements, % staked and distribution of validators across key L1 blockchains. However, the below illustrates how lower staking requirements reduce barriers to entry for one to run a validator node for a network.

The lower the barrier to entry, the higher number of validators, ceteris paribus. When assessing decentralization, comparing the number of validators across networks provides a simplistic sense of proxies for independent decision makers in the ecosystem (though they may not always map to one distinct entity, assuming uniform or no delegation) — albeit less meaningful without the minimum staking requirements in $ terms.

BSC is a common example of a blockchain known to lack decentralization given its high staking requirements resulting in a total of 21 validators only, where 7 validators represent 33% of staked BSC, as of 2021. Note 33% is the threshold for PoS in assessing security and liveness as it represents the minimum stake required to censor transactions and halt finality in PoS networks. By comparison, Ethereum’s 33% is comprised of 23 validators out of 30,594; Avalanche is 25 validators out of 978). The risk of validators colluding in BSC is significantly higher than other blockchains.

Aptos is currently undergoing testnet so its validator system is still being refined. The testnet is staged as follows:

• Phase 1 had 100 validators; 18,000 full active nodes at its peak (the largest known PoS node community running today).
• Phase 2 had 200 validators.
• Phase 3 initially announced to have 1,000 validators, but will adjust as necessary (will be more than phase 2).
• Phase 4 to not limit the number of participants but this may change.

Using validator count as a novel litmus test of decentralization and assuming Aptos has 1,000 active validators (2nd after Ethereum), suggests the network is likely to be relatively decentralized (but this would also depend on how many validators represent 33%). The actual degree of decentralization will only be known once Aptos launches mainnet, and it will evolve over time as the network scales and matures.

Since April 21 2022, Aptos entered into partnerships to accelerate growth. Notably, Google Cloud enables Aptos nodes to be set up in <15 minutes.

Security: In addition to Move’s security features, other components to the Aptos’ security lie within its consensus mechanism as well as account protection for users.

Consensus mechanism: The Aptos Byzantine Fault Tolerance (“BFT”) protocol (“AptosBFT” v4) — a modern, low latency derivative of HotStuff (similar to Cosmos Hub’s Tendermint BFT). Key features include:

• The AptosBFT is designed to ensure network continuity, and has experienced no downtime when undergoing upgrades (more on mitigants to potential network outages later). Its consensus mechanism provides fault tolerance of up to one-third of malicious validator nodes before the network has been compromised. State synchronization allows validators that have crashed or fallen behind to catch up quickly. Low barriers to entry, diversification and decentralization of validators adds to Aptos’ security.
• The protocol separates liveness from safety, whereby the network will not fork so long as the AptosBFT integrity guarantees are maintained — even in light of a network outage. The safety of AptosBFT has been both audited and formally verified.
• Block verification is done through a reputation system which analyzes the on-chain state and automatically rotates leader nodes to adjust for unresponsive validators. Since validator management and configuration is managed with on-chain state — upgrades can be voted on by the community and executed transparently and efficiently.

Transaction execution is deterministic, hermetic, and metered. As stated by Aptos, deterministic and hermetic means the output of transaction execution is predictable and based only on the information contained within the transaction and current ledger state. This is a common attribute of L1 blockchains, except for Ethereum, Avalanche. Metering is an important defense against denial of service attacks at the transaction execution level.

Account protection:

• Users (including validators) can rotate their private (consensus) keys to mitigate theft.
• Key recovery methods are currently being developed and will be integrated into the blockchain account model in order to avoid value becoming inaccessible due to lost keys. This is a new feature not yet available in other blockchains.

Scalability: Based on the testnet activity to date, Aptos outperforms its peers across two key facets — time to finality (“TTF”) and throughput. The blockchain is able to achieve <1 second TTF with max throughput of 160k transactions per second (“TPS”). While testnet results are derived from a controlled environment and tend to overstate performance, they are the only data available for Aptos today and nonetheless provide a moderate degree of visibility to the network’s capability in a production environment following mainnet. Below is a comparison of Aptos vs. other L1 chains.

Source: Pontem Network

High throughput blockchain implies low transaction costs for users. It is therefore worthwhile breaking down TTF / TPS for Aptos:

• Separation of transaction execution and the consensus mechanism. Protocols that integrate consensus and execution trade off throughput and latency due to their co-dependencies, despite being a simpler in design.
• For AptosBFT, transactions are validated in two network trips. Other L1 blockchains entail multiple rounds of voting on other consensus mechanisms.
• Authentication of state ledgers with optimized data structures with higher branching factors to Merkle trees (are only efficient at small scale), and using (optimized) tiered storage and state rent (instead of to persistent storage).
• State synchronization, enabling full node support without full computational intensity (CPU and network bandwidth). By allowing non-validators to distribute transaction data to directly update the ledger state ensures speed while the network stays synchronized. In contrast to other L1 blockchains, this data is typically replicated by network users as they are used, while validators execute transactions to add new blocks to the chain.
• Further, nodes can access the latest committed transaction from the top level transaction accumulator to sync network nodes, instead of downloading the full history of ledger transactions often required in other L1 blockchains.

Considerations & Risks

Potential network outages: A prominent risk inherent in all blockchains — network outages impacts UX and DX, e.g. Solana. For Aptos, the team continue to refine and upgrade its tech stack, including 4 iterations of AptosBFT . The protocol is understood to have had no downtime during several years of testing in a private mainnet environment. However, should an attack or periods of network outage eventuate, the on-chain reputation system minimizes the negative impact of down validators automatically.

Scaling an ecosystem organically: Aptos is still in its early innings. As a L1 blockchain starting from scratch, the key to ecosystem growth hinges on its ability to attract developers to build on top of it. Only then will dApps sprout to become the catalyst for user adoption for any blockchain. Move’s modernity, security and flexibility features discussed previously are expected to drive DX and facilitate the ideal developer environment to garner developer activity.

As of July 11, Aptos had 54.9k Discord members, 8.0k developers. A comparison with L1 peers below indicates Aptos has already started closing in the gap with other blockchains during its developer testnet.

Outlook

Aptos is undoubtedly well funded following their $200m seed investment. The team is seasoned and talented — and is strongly positioned to execute on Aptos’ vision, especially when considering 1) the resources and backing they’ve built; and 2) Move serving as an ideal language to foster development activity.

While the equity valuation of $1bn is a mouthful, it 1) reflects the maturity of the Aptos team and tech stack; and 2) pales in comparison to the total addressable market for smart contract L1 blockchains should Aptos propel into the top 5 positions. Aptos’ tech stack aims to solve the many pain points in the status quo — viz. the blockchain trilemma. Many have tried but few have made any respectable progress. Those that progressed struggled to sustain it. As such, it stands as a formidable contender to become the preeminent L1 blockchain.

I am optimistic about Aptos’ place in the web3 universe, and look forward to: 1) remaining testnet results; 2) it going live onto the mainnet in Q3 2022; and 3) when tokenomics have been defined and announced. I intend to follow up again with incremental updates.

Finally, if you are interested in working for an up-and-coming L1 blockchain for the people, then check out the positions available here.

Until next time…