« Back
DAI: Decentralised Autonomous Intelligence
  • Crypto / Web3
Updated 393 days ago
Ethereum · Polygon · IPFS · Harmony · Arbitrum · Optimism · Chainlink
Governance, bounty & membership protocol to allow DAOs to actually be decentralised, autonomous and operationally efficient.

DAI: Decentralised Autonomous Intelligence



DAOs don't work (currently). Operations are non efficient, processes are not automated, and decisions are not decentralised.

DAOs are fundamentally composed of departments or interests such as development, design, legal, operations and other interest sections with ideologically aligned individuals.

These groups are constrained by the broad democratic approach a DAO takes to all decision making and operations. Unless the DAO adopts a significantly more centralised approach where a centralised entity such as a “core team” has the power to push these decisions to product.

Not only is this approach a security risk, it is not decentralised or autonomous in any way.

How ?

Naturally a DAO breaks into entities of interests and ideas that push decisions and contact the entity in “control” of the DAO to push proposals or an idea.

By using these entities as the foundation of the DAO, we have the ability to create groups of experienced localised representative democracy.

Entities in a DAO, or Guilds, already run to create ideas and proposals to push towards the wider DAO, with one of two outcomes, a successful snapshot proposal, or a centralised decision by the “core team”.

Why not use guilds to decentralise and automate processes within the DAO. The fundamentals are in place but the organisational infrastructure and governance to allow these Guilds to do so is complex and underdeveloped.

By creating a charter or proposal for each Guild, containing scope, budget and governance rights along with a separate multi-sig wallet, we can allow guilds within a DAO to operate and function within these constraints. This will establish a significantly more automated and decentralised process for successful and efficient operations in a DAO.

Each guild's scope and budget is voted on by the wider DAO as well as the ability for the DAO to propose changes to this at any stage , meaning you still have a democratic approach to governance.

This model is also still decentralised as there is the ability for any member to join a guild onboarding process.



Guilds are effectively sub-DAOs. These sub-DAOs are token gated using the native guilds ERC-721 token. This token is non-transferrable. Guilds operate in “seasons” which can be quarters or certain periods that are defined by the guilds scope. Each season, the guild has the opportunity to remove bad actors from a system that can be executed through a vote.

When a user holds one of these “soulbound” NFT’s, they are a member of the respective guild. E.g. if a user mints a non-transferable Development Guild, they are a member of the Development Guild.

When users are members of Guilds, they can partake in voting, bounties, quality assurance of bounties and other operations within the guild.


In order to become a member of a guild, users must complete open bounties or tasks to enter the guild. This is to promote experienced or enthusiastic members, curating a sub-community of ideologically aligned driven individuals.

Guilds and their members will have the duty of creating open bounties and tasks to be published to the community. These bounties can assist the DAO or a product in some way, or just be a single purpose test to allow new member onboarding. Therefore, this can be a once off creation and/or consistent product development opportunity.

When a guild member is interested in a guild, they can view the open or closed bounties using Trello, DeWork, Wonder or similar.

Multiple members can submit to open bounties and upon quality assurance, gain access to the guilds. Closed bounties work similar but have greater rewards and can only be submitted by one member.

This access is granted through the minting of a ‘non-transferrable’ ERC-1155 token.

When the bounty is submitted, a quality assurance transaction is automatically submitted. This transaction can be signed by the holders of these ‘Guild’ NFTs (non-transferrable 1155). Upon successful submission of the transaction, a NFT is minted to the submitter.

Quality Assurance

In order to pass the onboarding phase, users must go through a “competency screening” process. Members in guilds should have some experience or enthusiasm towards operations within the guild. In order to automate this process, DAI: has built a bounty quality assurance mechanism.

When a user submits a bounty, the quality assurance transaction is initiated. Users holding the specific guild NFTs can vouch or vote for bounties they believe to be committed to a good standard. This removes the need for centralised and non-autonomous decision making for Guild onboarding. Addressing the scalability problem of growing guilds.

When a guild member vouches for a transaction, they must provide logical reasoning for the quality assurance. This voting decision and the reasoning, are minted in the metadata of the NFT. Using AI and/or IEML to analyse large datasets of proposals and reasoning, we can identify outliers and bad actors within the system.

Each time a quality check is performed, guild members that voted with identified logical reasoning gain credibility and share of a quality assurance prize pool based on the proposal rewards amount.

When the bounty quality check tx passes a specified quorum (voted on by the guild members), the transaction mints the submitter a ‘soulbound’ NFT representing their newly acquired membership to that guild.


Guilds will have a scope defining the extent of their operations. However, in order for these operations to be carried out, guilds must adopt a democratic voting system to make these decisions. Members holding Guild NFTs have the ability to vote on proposals within guilds.

All users can view these proposals, however in order to vote, you MUST have the respective Guild NFT.

When a user votes, they must provide logical reasoning as to why they have taken the position on that vote. Once provided, an NFT is minted with that reasoning and vote.

All proposal reasoning is run through text classification to determine bad actors in the system. E.g. A user that enters lorem ipsum for their 50 char limit will be an outlier in the model and identified.

Voting will be hidden from other guild members using zero knowledge proofs. This will ensure that people do not vote in accordance with the majority and strive for a more accurate and fair voting system.


Bounties within guilds are used only for product development. These bounties require proven credibility (Guild NFT Status) and will often result in higher rewards than “onboarding Bounties”.

These bounties are only accessible by guild members and upon completion, enter similar quality assurance to the onboarding bounties. Upon successful quality assurance of a bounty, the submission address is minted an NFT with the bounty details.


Traditioal financial markets evloved around order books specific to two scalar units. Buyers and sellers get matched when they value these units the same relative to themselve (i.e. both accept one price).

All innovation in DeFi largely follows the same approach. While most exchanges retired the order book in favor of more innovative matching, they still facilitate trading for fungible tokens only.

We feel the approach of facilitating coordination through scalar units does not tap into the full potential modern ledgers have to offer. That's why we propose a semantic matching engine to facilitate trading of goods and services directly.

How it works

A SemanticSwap exchange instance consists of a single generalized order book:

  • Offers: (I am able to do someting)
  • Asks (I want to have something done)

The key invention is to express those statements in IEML, the Information Economy MetaLanguage. IEML is a regular language with the expressive power of a natural language. This allows to be very precise with expressions. And more importantly, it is making those expressions computable so algorithms can automatically reason about things. This allows us to direct all our computing resources towards exploring the humongously large space of possible coordination, and look for the the ones we desire most. Humans wouldn't have the computational capacity to do that and they wouldn't like to do it anyways. This solution allows to split the work between humans and computers: Humans express what they want and computers tell us how to get there, verifiably.

Merger DAO and Defi

Current governance structures for DAOs try to solve two distinct problems at once:

  1. Deciding on what we want to happen
  2. Figuring out how to get there

SemanticSwap is very focussed on the latter. Instead of using money for coodination it utilizes automatic reasoning. We foresee a non-monetary economy emerging here and propose a name change from financial markets to coordination markets. Basically removing the abstraction of money and have a market directly for the coordination hiding behind it.

The first problem however is the mission of the DAO and governance movement. By seperating those two concerns all DAOs can now focus on figuring out what the comunity wants by utilizing various types of voting mechanisms and pooling of resources. Then that collective consensus about shared values can be expressed in IEML and forwarded to SemanticSwap. With that the whole DAO participates in the coordination market. SemanticSwap facilitates and optimizes the social coordinaton (matching of resources), both within the DAO and between the DAO and the external economies.

Benefits over monetary economies

This approach to coordination (as opposed to monetary economies) unlocks a couple of things:

Improved flow of information between buyer and seller

Market participants precisely express their asks and offers instead of putting up just a number. Price discovery can be skipped allowing for efficent matching even in what traditionally was considered illiquid markets (option products, ride sharing, brick and mortar stores).

Cognitive load is taken care of by machines

Expressing all offers and asks in IEML allows us to have computing systems search the huge space of possible coordination as they are able to understand the situation. There are no humans required to painstakinly figure out what to match with whom. They wouldn't be capable of finding highly efficient coordination proposals anyways due to the limited computational power and precision of human thought.

Long term reduction of inequality

When goods and services are matched directly the need of pairwise transactions fade away. Pairwise transactions really are just two gifts in the opposite direction with instantaneous, debt-free settlement. This model of facilitating trade however hase inequaity fundamentally built in. Even through one of the two gifts aims to be of the same value as the other gift, mispricings are very real for various reasons (information asymetry, subjective value, price controls, emergencies). The sheer possibility of having mispriced goods introduces inequality even in perfectly equal systems which is then further reinforced by the compounding of capital over time.

Our matching engine solves that by getting rid of pairwise transactions and match a multitude of market participants at once. This is similar in the way bartering economies did not use money, but different in that the matching process is facilitated by a computer system and can span many market participants spread out accross space and time.

Further Development

There is a lot of work to do after the hackathon in order to unfold the full potential of coordination markets.

  • Continuous matching: Accept short term debt instead of using a fungible delta settlement.

  • Add a corpus of common knowledge to verifiably account for constraints of shared reality.

  • On-chain, verifiable IEML reasoning: Have an objective function to find the most optimal way of coordination.

  • Speculation: Integrate first class prediction markets to allow speculators to predict the future. But in addition to state what they think will happen speculators also need to commit why it will happen in order to qualify for a payout. This enforces information sharing among market participants.

  • Enhance privacy: Use a commit-reveal scheme together with zero knowledge Proof of Work variant.

  • Towards a shared computation: Building a distributed operating system that allows for everyone to contribute with both their data and computational resources and become a market maker in the coordination market.

  • Introduce a dispute process: Such a distributed operating system also is your attorney. It helps you to verifiably record all data you need to proof that you behaved correctly, just in case there is a dispute. Following that personal, automated attorney is your free insurance for legal battles

  • Coordination ASICs: Build special purpose hardware to do zero knowledge semantic computing. This is the device that allows you to mine social coordination.

  • Grant Participation
  • Hackathon Participation
  • Contribution History
  • Upvote History