Reconnecting the Dots

BABE (Blind Assignment for Blockchain Extension) is Polkadot’s block production protocol. The Web3 Foundation’s own research page describes it:

And in the security analysis:

Ouroboros Praos was published by Bernardo David, Peter Gazi, Aggelos Kiayias, and Alexander Russell in 2017. It was developed for Cardano’s consensus layer. BABE’s VRF-based slot assignment, its epoch randomness accumulation, and its threshold calculation (p = 1 - (1-c)^(1/n)) are from this work. The two modifications BABE introduces, substituting GRANDPA finality for the Praos chain selection rule and replacing NTP clock sync with a median algorithm, are engineering adaptations, not new theory.

In the polkadot-sdk monorepo, a search across all 8,694 files returns zero references to "Ouroboros" or "Praos." There are zero citations to Kiayias et al. The only external reference is a link to the Web3 Foundation’s own research page. All copyright headers in the BABE implementation read "Parity Technologies (UK) Ltd."

Polkadot’s validator election algorithm is seq_phragmen, named in the code for Lars Edvard Phragmén, a Swedish mathematician who developed the method in 1894 for proportional representation in Swedish parliamentary elections. The sequential variant has been part of Swedish election law for over a century.

The Web3 Foundation’s contribution here is the Proportional Justified Representation (PJR) property, formalized in their 2020 paper (arxiv 2004.12990). The PJR property extends Phragmén’s algorithm with formal guarantees about fair stake distribution. The phragmms variant, a hybrid that achieves constant-factor approximation of the Maximin problem, represents further original optimization.

In the codebase, the file phragmen.rs contains no reference to Phragmén beyond the function name. The only paper cited is the W3F publication.

Parachain data availability uses reed-solomon-novelpoly, an implementation of the algorithm from "Novel Polynomial Basis and Its Application to Reed-Solomon Erasure Codes" by Lin, Chung, and Han, published at FOCS 2014. The paper is not cited in the erasure coding module.

GRANDPA (GHOST-based Recursive Ancestor Deriving Prefix Agreement), authored by Alistair Stewart and Eleftherios Kokoris-Kogia, is a finality gadget that separates finality from block production. It can finalize multiple blocks in a single round through chain-based voting. The paper is published on arxiv, the design carries formal proofs, and the separation of finality from production is a contribution the field has recognized.

Storing the blockchain runtime as WebAssembly bytecode on-chain under a :code storage key, enabling governance-driven upgrades without hard forks, appears to be a Substrate innovation (circa 2018). The execution engine is Wasmtime, but the architectural pattern of on-chain runtime storage is original and has influenced other projects.

The FRAME pallet system is a modular runtime framework for composing blockchain logic through macros that define storage, events, errors, and dispatchable functions. It is well-designed engineering. The shared security model, which secures multiple chains from a single validator set through a multi-stage validation pipeline (backing, approval voting via VRF, dispute resolution), is architecturally original. The sr25519 signature scheme (Schnorrkel/Ristretto) is a sound cryptographic choice with better multi-signature support and side-channel resistance than Ed25519.

In academic computer science, citing prior work is not a courtesy. It is a professional standard that serves a specific function: it allows others to evaluate the novelty of a contribution and to trace the intellectual lineage of a design. When a codebase implements an algorithm that its own organization’s research arm describes as "the same as Ouroboros Praos except for" two modifications, and the codebase contains zero references to Ouroboros Praos, there is a gap between what the research acknowledges and what the engineering presents.

This is not unique to Polkadot. Many engineering organizations cite their own publications rather than the underlying academic work. But in a project that draws significant credibility from its research pedigree (the Web3 Foundation maintains an active research division and publishes academic papers), the selective application of attribution norms is worth observing. The standard applied to the organization’s own publications (proper citation, formal proofs, arxiv hosting) is not the same standard applied to the codebase that implements prior work.

The governance of the Web3 Foundation’s 30% allocation (the restrictions on token sales, the accountability mechanisms, the decision-making authority over those tokens) is not publicly detailed at the granularity one would expect for an allocation of this scale in a project that emphasizes transparency and community governance.

Polkadot’s parachain slot auctions, launched in November 2021, used a crowdloan mechanism. The design: retail participants bond DOT tokens to a parachain candidate for the duration of a lease period (96 weeks). If the candidate wins the auction, the DOT remains locked for the full lease. Contributors receive the winning project’s native tokens as compensation.

The economic structure is worth examining in detail, because it illustrates a class of mechanism common across the industry.

The project receives locked capital for 96 weeks at zero interest cost. The contributor bears the full opportunity cost: staking rewards forgone (approximately 11% annualized at the time), complete illiquidity for nearly two years, and full market exposure to DOT’s price movement over that period. The compensation is denominated in the project’s native token, an asset created by the beneficiary of the lock, typically more volatile than DOT itself, and whose value is correlated with the broader crypto market.

This compensation structure differs from how locked capital is treated in traditional finance. Bonds and certificates of deposit compensate holders with yield denominated in a stable or senior asset. The compensation is priced relative to what the holder forgoes. In crowdloans, compensation is denominated in a junior asset issued by the party that benefits from the lock. The contributor provides capital certainty to the project and receives valuation uncertainty in return.

Two structural features are analytically significant. First, the auctions launched in November 2021, near DOT’s all-time high. The 96-week lock meant tokens unlocked between October 2023 and April 2024. The mechanism contained no market-cycle adjustment: no early exit provision, no partial unlock, no compensation adjustment for extended downturns. A fixed-duration lock across a bull-to-bear transition is not a novel situation in finance; the asymmetric outcomes for lockers versus liquid holders are well-understood by institutional participants who study market cycles professionally. Whether they were equally well-understood by the 81,000 individual wallets that participated in Acala’s crowdloan is a different question.

Second, the mechanism was optional for well-capitalized projects, which could self-fund their parachain bids without community contributions. The projects most dependent on crowdloans were those with strong community interest but limited capital reserves. The mechanism channeled retail capital toward projects that could not fund themselves, while projects with institutional backing retained full control of their capital allocation.

The crowdloan mechanism was deprecated in 2023, replaced by Agile Coretime, a marketplace model where projects purchase computation time at market prices. The transition is presented as technical evolution toward greater flexibility. It is also the case that crowdloan participation had declined significantly and that the mechanism had served its primary function during the period of highest market enthusiasm.

Before mainnet launch in August 2020, a 100x DOT redenomination expanded the token supply from 10 million to 1 billion. The Web3 Foundation initially shelved the proposal after noting "non-negligible opposition" within its own organization, then ran a binding referendum.

The redenomination is mathematically neutral per holder. Each holder’s proportional ownership is unchanged. It is also the case that lower nominal token prices have a documented effect on retail purchase behavior. This is well-studied in equity markets: stock splits do not change a company’s market capitalization, but they do increase retail trading volume and attract smaller investors who anchor on nominal price. The redenomination was executed before most retail participation in Polkadot began.

Polkadot uses OpenGov, a token-weighted voting system where one DOT equals one vote. Conviction voting allows holders to lock tokens for extended periods in exchange for multiplied voting weight, up to 6x for the longest lock duration. The system replaced an earlier three-body model (Council, Technical Committee, Referendum Chamber) in June 2023.

Token-weighted governance is common across the industry, and Polkadot is not unusual in choosing this model. The specific outcome, that a small number of addresses with large holdings make treasury decisions affecting all holders, is the expected result of a system where participation costs (research, active monitoring, capital lock-up through conviction voting) are borne individually while benefits are distributed collectively. This is a well-studied dynamic. Mancur Olson’s Logic of Collective Action (1965) describes how concentrated interests outperform diffuse interests in governance systems where participation is costly and the per-capita benefit of any single vote is small. Token-weighted systems amplify this effect because participation capacity scales directly with wealth.

The treasury decisions made through this governance structure are substantial. In 2024, the Polkadot treasury spent $133 million. The first-half breakdown: $87 million total, of which $37 million (42%) went to marketing, advertising, events, and influencer campaigns, including $10 million in sports and esports sponsorships. Network revenue over the same period: $250,000. Even the Web3 Foundation CEO, Fabian Gompf, acknowledged the treasury "has spent too much" on marketing with poor results. A former Polkadot developer, Seunlanlege, described the spending as "insane."

Polkadot’s mainnet launched with a sudo key, centralized administrative control held by the Web3 Foundation. This was removed in July 2020, two months after launch, through a governance vote. The phased approach has a reasonable security rationale. It also means that the network’s initial state (validator selection, parameter configuration, the initial governance framework) was determined by the Foundation before decentralized governance existed to influence those choices.

Polkadot’s defining technical promise is cross-chain communication between parachains. The mechanism designed for this is XCMP (Cross-Consensus Message Passing), which would enable direct peer-to-peer messaging without relay chain involvement.

XCMP has not shipped. As of 2026, Polkadot uses HRMP (Horizontal Relay-routed Message Passing), a relay-routed fallback described in the documentation as a "temporary alternative." It has been temporary for years. The core interoperability mechanism of a project designed specifically for interoperability remains undelivered.

For context: Cosmos IBC (Inter-Blockchain Communication) connects 117 chains and processed over $41 billion across its top 10 routes in 2024.

January 2020 (Kusama): A governance referendum deployed the wrong runtime, Polkadot mainnet code instead of Kusama code. BABE consensus failed. GRANDPA had already finalized the bad block, making rollback extremely difficult. Recovery required coordinating 130+ validators. The forkless upgrade mechanism deployed wrong code; the finality gadget prevented easy recovery.

May 2021 (Polkadot): The relay chain halted when the NPoS election algorithm exceeded the 64MB Wasm memory limit. Recovery required downgrading to native runtimes, which then produced different results from the Wasm runtimes due to compiler version differences, causing a second crash. The dual execution model was not deterministic.

Each major transition in Polkadot’s history, from sudo to governance, from parachains to coretime, from coretime to JAM, has a technical rationale. Several are improvements. Agile Coretime is a more flexible resource allocation model than fixed-term parachain leases. JAM’s service-oriented architecture may address scalability limitations the current design cannot.

What the timeline also shows is that participants in each era bear the transition costs of the next era’s pivot. Crowdloan contributors locked capital into a model that was deprecated. Parachain developers built on a framework that is being replaced. Coretime-era builders now face another transition. The institutional apparatus (Web3 Foundation, Parity Technologies, the core development team) persists across eras. The participants change.

The crypto industry has a well-documented structural incentive here: a project under active development retains narrative potential that a completed project does not. Venture capital and market attention flow toward new visions, not toward mature execution on old ones. When a project can perpetually redefine its roadmap, the promise is never falsifiable by present results, because the present is always framed as a transition toward a future that will deliver on the original ambition. Whether this describes Polkadot specifically or merely the industry it operates in is a distinction each reader can evaluate against the timeline above.