Progressive decentralization is the idea that a crypto protocol should not launch as a fully-formed DAO, nor stay a company forever, but move from one to the other in stages – a founding team relinquishing control "by degrees, over time." The term was coined by Jesse Walden of a16z in January 2020, and it has since become the default mental model for how a startup becomes a community-owned network. It is less a single mechanism than a sequencing discipline: which control to give up, and crucially, when.
Walden's playbook is three ingredients delivered in a strict order. First, product/market fit – built by a small core team with full control, because a protocol nobody uses has nothing worth decentralizing. Second, community participation – economic incentives that turn users into contributors who maintain and extend the protocol. Only third, once the social and economic dynamics have stabilized, does the team pursue sufficient decentralization: distributing tokens and governance rights widely enough that the network runs without them.
The ordering is the whole point. Walden argues that attempting the three at once, or decentralizing before there is a product worth governing, leaves a project "stuck" – a treasury and a token-holder base arguing over a thing that has not yet proven it should exist. The canonical worked example is Compound, which shipped a lending product, then a governance token and on-chain voting, and only then let the team fade into one voice among many. The reverse – governance before product – is one of the recurring patterns in how DAOs fail.
Progressive decentralization is not only product advice; it is a regulatory strategy. Under the US Howey test, an asset is an investment contract (a security) when buyers expect profit "from the efforts of others." A token sold by a small team that controls the roadmap looks a great deal like that. The industry's counter-argument traces to former SEC official William Hinman's 2018 speech, which floated the notion that a token can start as a security and later cease to be one once the network is “sufficiently decentralized” – at which point "the efforts of others" no longer meaningfully drive its value.
That single idea does a lot of work. It reframes "hand over control" from a nice-to-have into the event that (in theory) transmutes a token from a security into a commodity, and it is why so many teams pair a progressive-decentralization roadmap with a legal wrapper – a foundation or association that can hold the keys during the awkward middle. The obvious hazard is optics-as-substance: Walden's own warning is that "faking autonomy is a quick way to undermine trust," and regulators have been increasingly willing to look past a governance forum to ask who actually holds the upgrade key.
Congress came close to writing that idea into law. The CLARITY Act (H.R. 3633), which passed the House 294 to 134 on 17 July 2025 (CoinDesk), would have let a token's issuer, a registered exchange or the network's own decentralized governance system certify to the SEC that the network is a mature blockchain system: one not controlled by any person or group of persons under common control. The House text draws the lines in numbers. Nobody may have unilateral authority to alter the system's rules or direct 20% or more of its voting power, and the issuer and its affiliates together may not own 20% or more of the token.
A decentralized governance system and its participants count as separate persons, and so does a legal entity that implements one without centralized management. A certification would have taken effect after 60 days unless the SEC rebutted it (House-passed text, section 205). Governance vendors such as Tally advised clients to design for the test before it passed (Tally). It never became law: on 15 September 2026 the Senate fell short of the 60 votes needed to open debate, 49 to 50, which ended the bill for 2026 (CoinDesk).
A useful refinement came from Jad Esber and Scott Duke Kominers (a16z, 2023), who point out that decentralization "needn't be all-or-nothing." They break a protocol into minimum decentralizable units – the core team, external contributors, the technology stack, finance, and internal processes – each of which can be slid toward community control at its own pace. A team might open-source its contracts and hand the treasury to a token-holder vote At scale the same logic runs the other way: a mature DAO can spin a unit out into a SubDAO, working group, or pod with its own scope and budget.
The honest reading of the framework is also its sharpest critique. If every dimension has its own slider, a team can leave the load-bearing ones – contract upgradeability, the admin multisig – pulled firmly toward "centralized" while advertising the token vote as proof of decentralization. Critics call the result regressive or theatrical decentralization, and it is the gap between a governance forum and a credible exit right that usually reveals it: if holders who lose a vote cannot leave with their share, and the team can still upgrade the contract, control never actually moved.
Progressive decentralization assumes a protocol needs an ongoing governance layer and the only question is who runs it. A rival school argues the opposite: the safest governance surface is a small one, and the destination is not a vibrant DAO but a protocol with almost nothing left to govern. Liquity shipped its V1 as immutable contracts with no admin keys at all – there was never a control key to progressively hand over. Reflexer's RAI went further and removed governance in staged "ungovernance" levels until no new collateral types could ever be added.
Read side by side, the two roads answer the same anxiety – that a token vote is a large, permanently exploitable attack surface – in mirror image. Progressive decentralization distributes the keys; governance minimization destroys them. Most real protocols land somewhere in between, keeping a guarded lever or two (a security council, a parameter steward) behind timelocks and vetoes. The design question every DAO eventually faces is not "how decentralized are we" but "how much of this should be governable at all."
Caper sidesteps the progressive-decentralization journey by not starting centralized. A caper's balances live in an immutable Scrypto state component with no founder key, no pause, and no treasury wallet to relinquish later – so there is no founder handover to sequence, and none of the securities-optics of a "we'll decentralize eventually" roadmap. The founder's only lasting privilege is economic (a skim on every buy and sell, in XRD and in tokens, front-loaded because it shrinks as the supply distributes), not a control lever. The platform tier is a different matter: the shared logic every caper runs can be swapped, either by a $CAPER UPGRADE proposal or by the protocol operator's own propose-and-activate path, behind a registry delay fixed at deployment that reads zero on today's testnet (security councils has the detail).
Its governance surface is deliberately small and fixed at genesis. The treasury lives in a protocol-controlled vault that moves only through five typed, on-chain proposal paths fixed at genesis – PAYOUT, INVEST, DIVEST, METADATA, and UPGRADE, the last of which the contract asserts may only be raised on the $CAPER caper itself – and never through arbitrary calldata, and anyone can trigger a passed proposal's execution (the execute entrypoints are public; the treasury badge is presented as a proof inside the call and never leaves its vault).
In the vocabulary of this page, that is closer to governance minimization from day one than to progressive decentralization: rather than promising future decentralization, a caper simply never concentrates the load-bearing keys in the first place. What weight a holder does carry is earned, not merely bought – the same earned weight sets both a member's vote and their exit share, and it grows with a non-transferable, soulbound proof-of-vote record. Holdings still count, but the decisive part has no secondary market: the record is earned only by casting a ballot and can never move, so a bag cannot buy one at all – not its own, and not anybody else’s.