Special thanks to Jessie Li, Perama, Allison Duettmann, Yijia Chen, and Sylve Chevet for inspirations.
TL;DR: On validity rollups, we can build decentralized innovation protocols, or DIPs, for open verifiable innovation, the core of each is the combination of an insert-only timestamped order-preserving idea ledger, an onchain verifier and an onchain simulator. Together with new mechanisms of capital formation using the blockchain, we can decentralize capital-intensive and license- and commercial secret-heavy industries such as the semiconductor industry and increase the rate of innovation in these industries.
1- What is the true potential of validity rollups that leverage zero knowledge proofs?
2- Eventually, we will have awesome onchain video games, with trailblazers like Dark Forest, Conquest, Influence, Lattice, and Curio leading the way. They will spring from onchain components for media manipulation, where Briq and bitmapbox among others lead the way in the StarkNet ecosystem. With contract immutability and composability, onchain games will change our relationship with games from consumers of content to participants that create new mechanics and collectively expand the boundary of play.
3- Eventually, we will have a new kind of social media, one that is built with onchain open social graphs and knowledge graphs with no ties to our fiat identities, on top of which hyperstructures of libraries and protocols prosper.
4- Both video games and social media are categories that already exist. Could we think of categories that do not exist which leverage the core properties of validity rollups?
5- To possibly reason about the question, we need to go to the core of both what blockchain enables and what validity rollup enables.
6- Quoting Virgil Griffith: Ethereum is an unprecedented arena for playing cooperative games, and enables powerful economic vehicles we don’t yet understand, by bringing cooperative game theory into new domains. To me, the blockchain is an engine for coordinating massive positive-sum games in decentralized ways that could drive our collective productivity to new levels.
7- Validity rollups will scale computation by multiple orders of magnitude and compound that recursively using recursive proofs over the next decades. This is not high performance computing, but verifiable computation.
8- I believe over the coming decades, we will invent a new paradigm for knowledge production, verification, registry, and compounding on validity rollups. This will progressively trivialize the need for patent offices worldwide.
9- The problem with patent offices is threefold: (a) it is a legacy-ridden manual review process of new knowledge (b) it is tied to national interests on the backend (c) it tries to bundle proof of inventorship and restrictive use of invention, where the latter is enforced passively by the patent litigation processes costing exorbitant time and money. Regarding unbundling this bundle, NFT, in its most popular and often abused connotation, actually paves a new way by recording the proof of patron onchain while posing no restriction towards the use and distribution of the artwork tied to the NFT - Nouns being a perfect example. The more recognizable the artwork is through distribution, transplantation and derivatives, the more value the NFT accrues.
10- This logic fundamentally applies to knowledge as well. The more a piece of knowledge is iterated upon, expanded upon, and turned into production, the more it actualizes itself and thus the more valuable it is. The problem lies in allowing the original knowledge creator to capture downstream value both by being recognized credibly and neutrally as the inventor, and by being recognized as the beneficiary of derived financial value of said knowledge.
11- Perama wrote about Open Ideas, a system to increase the rate at which ideas are made available to society, using append-only public graphs of ideas and retrospective content-address-attribution royalty model. Among all possible ideas in different domains at different specificities, I am particularly interested in the subset where the correctness of ideas can be digitally verified, the “structure” of the ideas can be formally checked for equivalence, and the merits of ideas can be digitally quantified and compared.
12- Abstractly, this subset of idea can be viewed as mathematical functions that behave in particular ways according to objectives. For example, an idea for a digital circuit implementation of 32b fixed point multiplier is a function contructed from logic gates and perform multiplication on input operands. An idea for a rocket engine is a function constructed possibly from chemical, electrical, and mechanical components that convert chemical/electric potential into mechanical energy under certain conditions and at certain triggers. The performance of a multiplier circuit can be judged from its latency, power consumption, and gate count. The performance of a rocket engine can be judged from its propellant efficiency, max thrust, and reliability.
13- For any quantifiable input-output objective, there exists an insert-only idea ledger where (a) the insertion is gated on both the correctness of the idea and the structural uniquess of the idea; (b) the insertion index derives from the performance of the idea.
14- Using validity rollups, we can (a) build verifiers that verify the correctness of ideas through simulations; (b) build simulators to quantify the performance of ideas; (c) build checkers to perform formal equivalence checks between ideas.
15- As we build these verifiers, simulators, checkers, and build protocols that surround them, we are essentially trivializing the need for patent offices and changing our relationship with knowledge production at the same time. These protocols, which I will refer to as decentralized innovation protocols / DIPs for the time being (frivolous association with food does not seem good), will invite borderless and anonymous participation in knowledge production, records proof of inventorship onchain, records inventions in insert-only ledgers, and enforce co-ownership of these inventions. These onchain ledgers of invention are global by nature and, in emergent domains of quantifiable knowledge, will outpace the patent records and patent review processes that are fragemented in nation-backed patent offices.
16- This indicates that, while on L1 we accelerate the rate of innovation in the finance domain through open DeFi protocols, on L2 onward we will be able to accelerate the rate of innovation in many science and engineering domains through DIPs. This is how blockchain eats the world, and for me this is what the metaverse means.
17- As a more concrete example, Christopher is a proof of concept for an event-driven RTL simulator written in Cairo, together with a contract standard for representing logical circuits in Cairo contracts. The simulator is able to drive random stimulus to the circuit under test (or DUT / Design Under Test in industry lingo), propgate signals across the circuit and collect the output. The simulator is also able to count the latency from input to output, and the number of gates used by the circuit. Since the simulator tracks the signal toggles at every wire, it shall also be able to quantify power consumption of the circuit. This is the first step towards decentralizing the RTL design phase of the integrated circuit design process and could become part of the core of a DIP for integrated circuit design. Note that we can leverage verifiable randomness to construct random stimulus for functional verification that is hard to game.
18- Notice that theoretically this concept of DIP can be driven outwardly and inwardly. Outwardly, a DIP may cover progressively larger portion of the IC design process, starting from RTL phase and eventually covering the entire fully placed and routed chip. Inwardly, DIP can be directed towards improving components of DIP, essentially leverage this new innovation engine to improve the inner workings of the engine itself. For example, imagine we have DIP-STPOW and DIP-SNARK. DIP-STPOW is a DIP for innovating power management IC components for satellite applications, while DIP-SNARK is a DIP for innovating IC components for SNARK acceleration. It’s possible that both DIPs share the same RTL simulator. To prevent fragmented / repeated work and leverage DIP recursively, we can design a DIP-SIM, which is a DIP for innovating RTL simulators. Much like having consumer-facing companies and business-facing companies, we can have application-facing DIPs and DIP-facing DIPs.
19- It is intuitive to imagine that every DIP will be tied to a DAO for incentive alignment. This L2-native DAO will be highly autonomous because member contributions are directly measured onchain, derived from how much a member contributes to the idea ledger of the DIP. Combining onchain measurable contribution with onchain incentive mechanisms seems to leverage the advantages of DAO natively.
20- Referencing Karl Popper’s insight on the evolution of knowledge provided in chapter 5 of the new Foresight book, knowledge, much as biology, evolves by a process of variation, replication, and selection. Variation of knowledge as in tossing new ideas out there, replication of knowledge as in spreading ideas through conversation, and selection of knowledge as the discrediting of ideas through criticism. From this angle, I think the merit of DIP involves (a) streamlining the selection of knowledge through enforcing the quantifiable metrics of criticism in smart contract; (b) encourages the replication of knowledge through maintaining the idea ledgers transparently onchain, which is contrary to license- and commercial secret-ladden industry practices, and will impact the global education system as well since no formal degree or visa status is necessary to participate in DIPs, contrary to the regular company hiring processes; (c) incentivizes variation of knowledge through rewarding the submission of new performant ideas with protocol incentives.
21- A key question that follows a hypothetical success of a DIP for IC design is that it remains to be manufactured to have actual utility, and semiconductor manufacturing is costly. I am less concerned with this part of the equation because, as we have seen numerous times such as with the Constitution DAO and Nouns DAO, blockchain is extremely good for capital formation. I believe as the validity rollup ecosystem grows, we will create new capital formation mechanisms to fund both the upstream data procurement (for example, acquiring the TSMC 5nm node libraries in order to create the DIP for fully placed and routed chips, or acquiring biological data to create the DIP for human cell simulation needed for longevity research) and downstream manufacture and distribution.
22- One final thought: it will be desirable to apply DIP, if proven successful as a mechanism for driving engineering innovation, to domains the results of which shall not be monopolized by nation states. One such domain is the space industry.