Overview

It is evident that Starknet is taking its roadmap seriously, more so than many other L2s (and L1s). Launching the SN Stack in January 2025 and promising developments on the horizon. Parallel transaction execution, block packing, and data compression have already reduced costs and increased transaction speeds. The next major upgrade, v0.14.0, aims to further reduce block times to 2 seconds while also creating a more efficient fee market. Coupled with the ongoing progression of its staking phases, the wider community can begin to glimpse what Stakrnet is working towards.

Cairo

Unlike many other blockchain-based languages, Cairo was written explicitly to create provable programs using validity (or succinct) proofs. Cairo’s syntax was inspired by the popular web2 language, Rust, which is known for its performance as well as its usage in Solana and AI applications. 

Thanks to a zk-circuit first approach, programs in Cairo can be more efficiently proven while being more versatile than other languages. Furthermore, developers do not need to know the intricacies of zk cryptography to create usable applications. In fact, zero knowledge of zk-related technology doesn’t make any significant difference. A notable real-world implementation is the Raito project, which uses Cairo to prove Bitcoin's UTXO, enabling faster node synchronization. All in all, Cairo represents a much-needed alternative that can be thought of as the “zk-Rust” of crypto languages. 

Cairo VM

Below we highlight some of the differences between Cairo VM and the EVM at an abstract level.

Technically, the Cairo VM execution model is not parallel, but instead, the execution engine uses (optimistic) concurrency to increase throughput, but this is highlighted to show the direction the protocol is taking vs the current standard EVM. 

Developer Experience

The developer experience on Cairo and Starknet has shown a positive evolution over time. A recent survey of Starknet developers revealed that 86% rate their development experience as above average, with 62% noting significant improvements over the past year.

Source: Starknet

The ecosystem now features robust development tools, including Scarb for package management, Starknet Foundry for smart contract development, and popular SDKs like Starknet.js. Most developers (61%) come with 3-5 years of experience and have no problem transitioning due to the development environment being primarily centered around VSCode, an extremely common IDE. Learning resources are readily available, with 29% of developers primarily using the Cairo Book, while others rely on Starknet documentation and community resources. Below we show the number of active developers by type since inception. 

Despite being inspired by Rust (created in 2006), Cairo (created in 2020) is the newest crypto language with material adoption. The only other modern language in the same category is Move (created in 2019). Below we show the growth in the adoption of these languages in recent years among full-time developers. 

Full-time developers are defined as those who commit code 10+ days out of a month. At the same time, established developers are defined as those who have worked for over two years. Below we show growth in established developers over recent years. 

SN Stack

The SN Stack, Starknet’s zk technology stack designed for constructing bespoke chains, was officially released in January 2025. Unlike other L2 stacks, the technology behind the SN Stack has been rigorously tested and proven over the past four years through its implementation by industry-leading applications such as dYdX, Immutable, and Sorare. The success of these apps is a testament to the top performance and reliability of the SN Stack technology. As an L2, dYdX facilitated more perp volume than any other dApp (over $1T), Sorare onboarded thousands of non-crypto natives through fantasy sports, and Immutable has millions of monthly active addresses. Starknet achieved a TPS of 128, the second highest of any L2 only after Base, which achieved 155 in January 2025. 

As shown below, the SN Stack offers flexibility with three distinct implementation options:

• StarkWare Sequencer: This is the implementation a chain would use if it shares similarities with Starknet, the “sequencer for CVM chains”.
• Madara: With dYdX and ImmutableX, Starkware was a first-mover in the formation of application-specific chains (appchains). Madara is suitable for these more custom applications.
• Dojo (Katana sequencer): While not yet popular, game development has long been heralded as a main use case of blockchains. The Dojo implementation is designed with game development and tooling in mind. 

Proving

For the proving later, developers also have a choice of three different components:

• The Stone Prover is the base one; not the most performant but rigorously tested over time.
• SHARP (SHARed Prover) is a step-function improvement on Stone in the sense that it operates by aggregating multiple proofs into a single one. This implementation, known as recursive proofs, allows the verifier to verify only one proof, which implicitly verifies many proofs at once. This aggregation reduces costs through amortization and can bundle ~220K transactions into a single proof on Ethereum. That being said, Starknet is not one to rest on its laurels, and the release of a next-generation prover, Stwo, is already underway. Expect the proving landscape to continue to evolve rapidly as both the hardware and software are improved upon simultaneously. The endgame for this technology is “real-time” proving where proofs can be created and verified within one block.
• Lastly, there is Atlantic, which is operated by Herodotus. This is simply an intermediary layer that helps manage different aspects of the proving and verification. The combination of these three proving solutions enables Starknet to achieve high scalability while maintaining security and efficiency.

Verification

For verification, the standard L1 contract receives and verifies the validity proofs and updates Starknet's state root once the proofs are valid. However, Starknet also introduces the Integrity verifier, which enables developers to verify the execution of Cairo programs performed elsewhere, directly on Starknet. This requires building a layer on top of Starknet where transactions can be executed and bundled into a proof that attests to their validity. The verification of these proofs on L2 is where the Integrity Verifier comes into play. The Integrity Verifier introduces L3 scaling by adding a layer of proof compression on Starknet before settlement on Ethereum. This extra step reduces costs and enhances scalability, allowing the development of fully customizable appchains that leverage the Integrity Verifier's cost-saving features.

Data Availability

Developers on Starknet have a range of solutions for data availability. Ethereum blobs offer a structured approach with 6 blobs per block at 128KB each. Celestia distinguishes itself with blocks ranging from 2MB to 8MB and implements data availability sampling (DAS), delivering roughly 40 times more capacity than Ethereum blobs at significantly lower costs. Avail takes a different approach with its 20-second block time and 40-second finality, utilizing validity proofs and KZG commitments alongside DAS.  

Settlement & Consensus

Starknet is introducing staking to its network using a phased approach (with phase 1 already complete). Initially, validators focus on running full nodes and testing the economic incentives and smart contract components. By the time validators assume their complete role in producing, attesting, and proving blocks, there will be an established group of sequencers with proven reliability in block production and attestation. This means that Starknet will be responsible for its own consensus through its own validator set. This is important if Starknet wants to become an execution layer for multiple L1s, but it does add an additional trust assumption (an honest majority of the new validator set). Furthermore, Pitolver (the Cairo core contract), allows Starknet to have its own settlement offering as well. Starknet currently has over 150M STRK staked, 100 validators, and over 60K delegators. 

Source: Starknet

It remains to be seen if the previous usage of Starkware technology will be enough to convince teams to utilize their stack over others. As of February 2025, Paradex marks the first and only live chain on the SN stack and has accumulated $35M in value secured, a 700% increase YoY. 

Upgrades

The Starknet ‘Bolt’ upgrade (v0.13.2) in August 2024 was a massive performance upgrade for Starknet. It introduced two major features:

Parallel Execution

Starknet became the first L2 to implement parallel transaction execution on mainnet. This feature allows independent transactions to be processed simultaneously rather than sequentially, significantly expanding the network's capacity beyond current needs to support future use cases. 

 

Source: Starknet

Block Packing

Whereas parallel execution improves efficiency in the transaction-execution stage, block packing improves efficiency in the step that comes after, when transactions are batched together into blocks.

L2s periodically send transaction batches (blocks) to L1 for final settlement. This incurs costs on L1, which are fixed and increase linearly with the frequency of blocks received, regardless of transactions per block. Block packing addresses this by batching multiple L2 blocks into a single validity proof for L1 submission, combining multiple Starknet state updates, and batching L1 fixed costs. Now, low block times are unlocked without incurring increased costs, resulting in faster confirmation times and lower costs. With block packing, transaction confirmation will take only 20-60 seconds (the goal is to reduce that to only 2 seconds). On the cost front, the majority of the fixed L1 cost is now spread over all the batched blocks, slashing these costs by between 50-66%

Source: Starknet

Similar to how L2s achieve cheaper transactions by combining multiple user transactions together into a “block”, the same is true for combining multiple blocks. As usage grows on the network, more L2 blocks fit in each proof, shrinking each user's L1 footprint as the network grows, ensuring Starknet's scalability.

v0.13.3

In November 2024, Starknet reduced costs even further with its v0.13.3 upgrade by making blob gas 5 times cheaper on the network. This was thanks to state diff compression and enhanced block squashing. These optimizations were particularly timely, given the recent stabilization of blob traffic (on Ethereum) around target rates.

v0.13.5

Starknet's v0.13.5 upgrades further optimized blob gas costs through a new feature called "stateful compression". A new gas model was implemented, introducing L2 gas to more accurately account for Layer 2 resource consumption, such as computation and storage. Error handling was also improved (smart contracts can now catch and handle errors without immediately halting execution), and a new wallet-dApp API was introduced (simplifying integrations and fostering a more unified ecosystem).

Cairo-native execution is expected soon. Previously, the Starknet sequencer executed programs using a virtual machine that emulated code, which introduced significant overhead and slowed down transaction processing. Instead, native execution allows programs to run directly on the operating system, improving execution speed and resource efficiency. 

v0.14.0

Starknet's version 0.14.0 is a major upgrade scheduled for release in June 2025 and will introduce a fee market mechanism designed to help users optimize their transaction flow during periods of network congestion. The update will also implement 2-second block times and memory pool optimization, along with the introduction of the Stwo integration mentioned earlier. Stwo will allow for client-side proving and two to three orders of magnitude performance improvement over the existing Stone implementation. This upgrade will coincide with Starknet's staking phase 2, where validators will begin to actively participate in block attestation. This will help to evaluate the liveness metrics of the validator set before assuming a formal role in consensus by the end of 2025. 

Looking Forward

2025 is a promising year for the Starknet community. The upcoming Stwo prover will be orders of magnitude better than the existing base solution. Combined with blockpacking, 2-second block times, and native Cairo execution, the new goal for TPS is in the 1-2K mark. Furthermore, as outlined in a previous report, Starknet is aiming to become the first network to settle simultaneously on both Bitcoin and Ethereum. The multi-phased staking approach is progressing smoothly, while the SN stack is set to allow for more L2s and L3s. The team is also building and contributing to frameworks for developers to create cross-game and DeFi AI agents on Starknet directly. While the engineering side seems to be progressing without any hiccups, there are still improvements to be made on the liquidity and usage fronts. Look to 2025 to see how well an impressive technology stack can attract applications and create the desired network effects.



The information contained in this report and by Blockworks Inc. and related affiliates is for general informational purposes only and is not intended to provide legal, financial, or investment advice. The report should not be construed as an offer or solicitation to buy or sell any security, token, or financial instrument and does not represent any recommendation or endorsement of any investment or financial product or service. Blockworks Inc. and related affiliates are not registered as a securities broker-dealer or an investment advisor in any jurisdiction or country.