But with LLMs, the current sci-fi moonshots seem to become a reality much sooner than expected. Relay’s info is at the bottom if you would like a Robot to be made at the ENS name: ai.yourcompany.eth

User Onboarding

1. The relevant initial conversational direction and prompt of the agent

2. The default responses of the agent when the user input is undirected or confused.

3. Agent assistance leads to a +EV boost in first-time users engagement and high-level understanding of the product

4. Chatbot personalization based on the user's transaction history and token holdings, providing a tailored experience

5. Proactive assistance with setting up and configuring user wallets, ensuring smooth onboarding

6. Agent can provide a guided tour of the platform, highlighting key features and use cases for new users

User Education

1. Agent can explain everything in the docs, with multiple levels of expertise (from ELI5 to crypto expert)

2. Can explain company position in relation to competitors and the web3 landscape.

3. Can focus in on the business CTA for the user, and the value prop thereof

4. Agent can offer interactive quizzes to test user understanding of platform features and concepts

5. Can provide real-world use case examples to help users better understand the potential applications of the platform

6. Customizable learning paths based on user's prior knowledge and interests

User Support

1. Agent can help with common user problems

2. Discord/CRM bridge for transparency and, if needed, human support intervention

3. Escalation to a human based on keywords or sentiment analysis

4. Multi-language support for a wide range of users

5. Agent can detect user frustration and proactively offer support or alternative solutions

6. Integration with knowledge base and FAQ for quick and efficient access to information

Developer “DevRel” Support

1. SDK education/explanation

2. SDK troubleshooting

3. Help with Github template creation, with a deeper Github sign-in to the chat widget

4. Agent can assist with API documentation and usage examples

5. Can connect developers with relevant communities and resources for further support

6. Help developers understand and implement best practices for integration with the platform

Sales Channel

1. Agent can convince users to buy digital products with crypto

2. Agent recognizes the receipt of tokens, and can disburse an NFT/digital good/event ticket

3. Agent can collect user information and ask followup questions to segment leads

4. Agent can offer personalized product recommendations based on user's interests and token holdings

5. Can upsell or cross-sell related products and services to provide a more comprehensive solution for users

6. Agent can answer questions about pricing, payment options, and refund policies

DeFi Interaction

1. The agent can assist users with onchain interactions

   1. Metamask Swap

   2. Uniswap Swap

   3. ENS Registration

   4. Maker DAI creation / swapping

2. The agent can tell the user about their token holdings on any EVM chain

3. The agent can provide real-time data on token prices, liquidity pools, and other DeFi metrics

4. Can walk users through staking, lending, or borrowing processes on various DeFi platforms

5. The agent can offer risk assessments and advice on portfolio management for DeFi investments

6. Agent can autonomously manage user's DeFi investments by actively monitoring market conditions and adjusting portfolio allocations based on predefined risk profiles.

DAO Interaction

1. The agent can explain the main governance features of a DAO and proposal system

2. The agent can explain who the largest / most active delegates are.

3. The agent can explain the current proposals, and the active forum discussions

4. The agent knows the entirety of forum discussions / discord channel / official twitter / blog posts about a DAO

5. The agent can assist users in delegating or voting on an active proposal

6. The agent can guide users through the process of submitting a proposal or joining a working group within the DAO

7. Can provide updates on the status of proposals and ongoing discussions

8. The agent can offer insights on the historical performance of the DAO, including past proposals and their outcomes

User Data Insights

1. Relay could use wallet data analytics such as Absolute Labs

2. Most frequent bot questions / answers

   1. Also agent answers that end a conversation (no user response) could be important

   2. User emoji reactions to denote positive or unhelpful responses

3. Sentiment analysis of user feedback to identify areas for improvement and optimize the chatbot experience

4. Analysis of common user paths and drop-off points to inform platform UX/UI improvements

5. Periodic reporting on chatbot interactions, user satisfaction, and support metrics to track performance over time

Communication Channels

1. Website chat widget

2. XMTP clients

3. Other web3 messaging adapters

4. ChatGPT plugin or Poe API

5. Discord / Telegram bot

6. Integration with email and SMS for outbound notifications and support

7. Voice-based interface for users who prefer a more natural, conversational interaction

8. Omnichannel support, allowing users to seamlessly switch between different communication methods while maintaining conversation context

NFT Marketplace Assistance

1. Agent can provide information on the latest NFT drops, trending collections, and artists

2. Can guide users through the process of buying, selling, or minting NFTs on various platforms

3. The agent can offer tips on NFT valuation, rarity, and provenance

4. Can assist users in setting up and managing their NFT gallery or portfolio

Gaming and Metaverse Support

1. The agent can introduce users to popular blockchain-based games and virtual worlds

2. Can provide guidance on in-game assets, currencies, and tokenized items

3. The agent can help users navigate and participate in metaverse events and experiences

4. Can offer tips on game strategies, leveling up, and earning rewards

Social and Community Engagement

1. The agent can facilitate user connections with other platform members based on shared interests or token holdings

2. Can inform users about upcoming community events, AMAs, and webinars

3. The agent can help users find and join relevant sub-communities within the platform

4. Can encourage user-generated content and sharing of experiences with the platform

Security and Privacy Guidance

1. The agent can provide best practices for securing wallets, tokens, and digital assets

2. Can educate users on potential scams, phishing attempts, and security risks in the web3 ecosystem

3. The agent can offer guidance on maintaining privacy while interacting with blockchain-based platforms

4. Can assist users in setting up multi-factor authentication and other security measures for their accounts

We hope this informs your ideas of everything a web3-enabled chatbot can do for your company!

We’re building this out with these more advanced bots going live this summer, email sean@relay.cc or message seanwbren.eth on xmtp.chat if you would like a demo.

Note: Industry sectors are loosely-defined!

Blockchain-aware AI Assistants

Relay — Helpful LLM agents with ENS names — making them available on the wallet messaging networks (XMTP, WalletConnect, Push, etc). Users talking with our Robots natively share the context of their wallet tokens and transaction history — enabling smarter customer support, multi-step DeFi transaction assistance, and conversational NFT sales or POAP distribution.Soon, AIs with self-custody of their crypto can use our web3 communication rails to initiate chats with others, allowing an economy to bloom. Website — Twitter — Technical Overview & Vision

Algovera — A platform for building personalized AI assistants driven by your data sources, as well as a hub for AI startups and a lab for decentralized AI infrastructure. Algovera and AI3 are hosting an AI x web3 hackathon next week in Lisbon. Website — Twitter

MitiAI — AI-powered web3 explorer. Unreleased; they had a very cool EthDenver demo. Website — Twitter

marginfi omni — A Solana-ecosystem-focused chat interface for interacting with marginfi lending and learning about Solana projects. Launch thread. Website — Twitter

Airstack — A web3 developer platform that provides fast and easy access to on-chain data across projects and blockchains. Their AI assistant creates GraphQL queries for you, and can access Ethereum and Solana data. Website — Twitter

Greenwood — Help desk for DeFi, via chat. Demo shows finding the highest yield MKR strategy and other uses. Website — Twitter

LangWallet — A personal crypto wallet agent, using WindowsAI. Also founding LangDaoAI, a home for LLM builders bringing them to the decentralized economy. Twitter

WindowAI — The Window extension allows you to configure the models you use on the web. You can choose from OpenAI, Together, Cohere, or even an AI running on your computer if you need privacy. Website — Twitter

AI Agent Decentralized Economy

Autonolas — A protocol to register and manage agents, components and autonomous services. Their recent launch of the Open Autonomy Client SDK is a very useful way to query data from a set of agents in the Autonolas protocol.Website — Twitter

Fetch.ai — Open Infrastructure for smart, autonomous services. You’re able to create an agent and deploy it to the Agentverse. Website — Twitter

Valory.xyz — Weaving the economy of the future through research & development at the intersection of multi-agent systems and distributed ledger technology. Website — Twitter

Bittensor — Incentivizing intelligence. Recent podcast interview with the founder. Website — Twitter

Alethea AI — Released the CharacterGPT product, and their new product NoahsArk has interactive NFTs. Has the token ALI. Website — Twitter

Data for AI

Ocean Protocol — Ocean Protocol helps to unlock data, particularly for AI. It is designed for scale and uses blockchain technology that allows data to be shared and sold in a safe, secure and transparent manner. Website — Twitter

Semiotic AI & The Graph — In 2021 the Semiotic AI Team joined The Graph for $60m over 8 years. Recent overview thread. Also working on the Agent economy. Website — Twitter

Gensyn AI — Decentralized compute: the gensyn network unites all of the world's compute into a global machine learning cluster, usable by anyone at any time Website — Twitter

Numerai — AI hedge fund built by a network of data scientists, where you can download free data and train a model. Website — Twitter

Spice AI — Has an “AI OSS”, designed for developers building the next generation of intelligent apps together. Share your dataset, models, knowledge, and progress. Website — Twitter

Identity & Authenticity

LitProtocol — Decentralized access control, and now AI Authenticity via Threshold Signatures. Website — Twitter

Worldcoin — When non-AI zones of the internet are required, Sam Altman has a solution to his OpenAI problem. Allows sign-in auth to proven members, as well as “Anonymous Actions” which can be completed a limited number of times per person. Website — Twitter

Bonus

Modulus Labs — Bringing AI on-chain using zkps, with some released games and posts on their Medium. Website — Twitter

Kosen Labs — ex-Google Deepmind team building active liquidity strategies for DEXs. Seed Raise Article. Looks like an AI that can optimize your DEX positions for you. Website — Twitter

AllianceDAO — They wrote an excellent post on the convergence of AI x blockchain, and shared their ideas for the features and companies that can grow. They were recently on Bankless to discuss it.

Verkle Labs — AI x Blockchain Consulting. Launch thread. Website — Twitter

Recent Hackathon Projects

De_Terminal — EthGlobal Tokyo link — They create an AI agent that knows how to access information about Uniswap V3 Pools and uses this information to answer questions from the user.

Chatshire — EthGlobal Tokyo link — Designed to simplify blockchain data exploration using your natural language without needing to learn SQL grammar.

Octoplorer — EthGlobal Tokyo link — Get human-readable answers to blockchain data questions in formats like tables, images, and graphs. Has responses through Push Chat.

Special credit to Achilles Schmelzer and Richard Blythman for help in sourcing this post.

The world already seems changed from November 30th’s ChatGPT launch, with AI becoming a powerful force and policy-making topic, as it is right to be.

Relay was founded on the trend-line of increasing benefits to communicating as your web3 identity. With blockchain-aware LLM agents, we can turn complex DeFi flows into clear conversations, turn confused devs flooding Discord channels into transparent DevRel learning session, and turn helpful onboarding chatbots into user-wallet-knowing upselling agents of commerce.

Looking relatively far in the future (for this nascent space) to ~Q1 2024 — We’re developing LLM agents able to self-custody their crypto safely, communicate with other bots, and transact with them. The Relay interfaces and communication rails through ENS and wallet chat can become the de facto standard as the easiest method for bots to 1) control their crypto 2) communicate 3) transact.

The Relay Robot

The Relay Robot is a system for foundational models that significantly improves on baseline LLM solutions by implementing what we call an LLM trajectory framework, an idea Relay adapted from the famous ReAct whitepaper. It’s useful to think of a “trajectory” as an LLM’s train of thought or its “trajectory through reasoning-space”.

We instrument an LLM with external tools, show the LLM how to use the tools, and then observe the LLM’s trajectories as it responds to both a training environment and real user queries. By tracing, modeling, and analyzing these trajectories (sometimes in real-time) we can massively improve upon baseline LLM behavior and deliver production-ready solutions.

In this post we go through the problems with the product made by following the first (and great!) OpenAI data retrieval tutorials, and then we go over our solutions, both in-practice and theoretical, that improve a web3-enabled chatbot used for education and transaction assistance.

Lastly we discuss Relay’s vision for how the AI agents that we and others build will use our LLM framework for oversight and safety, plus Relay’s wallet-messaging rails, to:

1. self-custody and have plausible independent ownership of their crypto

2. communicate with other user-wallets (both human and AI)

3. intelligently transact crypto and NFTs with humans and other AIs. Trading, buying digital goods, buying services, it’s all possible.

The AI x Blockchain industry is growing fast, we also just published our landscape overview post here.

The Baseline Solution

Advances in language model technology have made it straightforward to build a proof-of-concept chatbot with access to an external knowledge base. These chatbots all follow roughly the same core architecture:

The default early solutions all follow roughly the same pattern:

1. Download a bunch of data from the internet

2. Chunk up the data into LLM-manageable pieces

3. Convert the text data into embeddings using an embeddings model (an embedding is a quantitative representation of the semantics of a piece of text)

4. Upload the embeddings into a database

5. When a user asks a question, convert the question into an embedding and use it to search the embeddings database for related documents

6. Pass the user’s question and results of the search to a language model

7. Return the language model’s answer to the user

Building a chatbot in this way feels somewhat magical, it’s pretty incredible the extent to which they “just work” without running headfirst into any additional complexities. That said, most of these bots aren’t suitable for production use-cases and have some major challenges — especially at scale.

Problems with the Baseline

💰 Expensive

The state-of-the-art and publicly-available-for-production LLM (GPT-4 from OpenAI) is $0.06 per 1,000 tokens (1-2 pages of English text) generated by the LLM. A single request to GPT-4 can cost up to $0.48. For language models to behave in a way that feels natural to users, they require entire conversation histories to be passed in every request, so it is quite common to run into this upper limit. Even at very model scale this is quite expensive (10,000 requests at 2,000 tokens per request would cost ~$1000).

🦥 Slow

A single request to GPT-4 can take anywhere between 3 and 30 seconds depending on server load. In a conversational setting, this amount of latency is unacceptable to users and will lead to “conversion abandonment”. This latency is decreasing as OpenAI scales their own servers, but it is a fact that caching and UI/UX patterns have to account for the greatly atypical response time.

🔐 Vendor Lock-In

Every language model (even different “flavors” of the same underlying models) display unique behaviors and react differently to identical inputs. Migrating a system from one model to another is more of an art than a science.

💥 Unreliable

Commercialized language models are both extremely novel, extremely resource-intensive, and in extremely high-demand. These factors produce a level of unreliability that is mostly unheard of with modern web applications, and so potential API downtime must be designed around.

☢️ Unstable

Language models are somewhat unpredictable. They actually aren’t even deterministic: if you pass a language model the same input 100 times, you’ll end up with 3-5 seemingly random answers. This can lead to confusing and heterogenous user experiences and will undermine users’ trust in the platform.

💔 Brittle

Language models feel extremely powerful, and they are, but their limits become obvious with non-trivial use. If you want a language model to do anything other than literally answer a question, with no constraints on the kinds of answers you desire, they work well, but that’s about it. The default LLM experience will become stale and users bored and unenthusiastic about your brand.

⬜️ Generic

Language models don’t adhere to any particular brand. If you go to example.com and talk to their GPT-4-powered chatbot and then go uniswap.org and talk to its chatbot, they will be indistinguishable from each other.

🤦‍♂️ Impersonal

Language models are not only generic, but they’re also completely impersonal. They don’t know “who” they’re talking to or “why” they’re talking to them or what the user wants. Talking to a language model can become frustrating for users, especially when the conversations are within a specific and complicated context.

🕛 Static

Language models don’t learn from their conversations and they don’t improve over time. They don’t acquire new “experience”. Once deployed, a baseline language model solution will never improve.

🐪 Non-Autonomous

Language models don’t have the capacity to interface with anything other than their prompts and the data they’ve been trained on. Users expect new features, they expect applications to grow with them. A baseline language model cannot support users’ expectations.

The Advanced Relay Robot

Each of the following sections describe a category of solutions that improve on the baseline. Each section will describe the category in general, provide a specific example of what the very first version could look like, and go on to describe a more long-term vision. One thing to keep in mind is that the space of possible features is quite large and the field quite young, so the following sections are not meant to be prescriptive nor exhaustive. The goal is to provide a primer or a map of what’s possible.

Agent Tooling

Augment an LLM by providing a natural-language interface for calling external tools. A first version of agent tooling could include

• a way for the LLM to search white-listed, well-known web sites for info

• access to a pre-populated, indexed knowledge base

• access to on-chain data such as token prices and transaction history for the current user

Agent tooling can be used to improve reliability, stability, and dynamism (decrease brittleness).

Semantic Tracing

Every LLM trajectory leaves a “trace” in “semantic space”: each step in the trajectory can be defined by a question in natural-language that the LLM is trying to answer. By embedding and indexing these questions, we generate a dataset that can be used as an input into many of the other solutions we develop. A first version of tracing could include:

• trace the LLM input

• trace the LLM output

• trace each call to agent tooling

• trace each “cache” miss (failed calls to agent tooling)

Semantic tracing can used to improve costs, latency, reliability, observability, analytics, and more.

Real-time Human in the Loop

Intercept, pause, cancel, and divert trajectories. The behavior of when to intercept, when to cancel, when to loop in a human can be pre-determined or based on dynamic data (such as data generated from semantic tracing). A first version of real-time human in the loop could include:

• basic observability (pipe all trajectories through Discord)

• escalation (react to a trajectory to jump into the conversation in lieu of the robot)

• trajectory labeling (similar to, and the beginnings of, a reinforcement learning framework)

Real-time human in the loop features can be used to improve stability, observability, and branding (reduce genericness).

Real-time Analytics

Provide dashboards and alerting based on predefined criteria and semantic tracing data. A first version of this could include:

• topic-based alerting: trigger alerts or notifications when users ask about specific topics

• sentiment-based alerting: trigger notifications for conversations based on “how the conversation is going”

• a semantic heatmap with clustering: show clusters of user requests and responses

Real-time analytics can be used to improve reliability, stability, and useful insights into the product.

Semantic Caching

Very similar to semantic tracing, we can also introduce semantic caching. When a trajectory is very close to another in semantic space we can choose to halt the trajectory and return the result from the previously-executed trajectory. One way to think of this is that we allow the LLM to learn. A first version of this solution could include:

• pre-trajectory short-circuiting (if the user request is a match, don’t even call the LLM)

• mid-trajectory short-circuiting (if the LLM-generated question for the trajectory step is a match, fetch the previous answer and return immediately)

Semantic caching can be used to improve cost, latency, and reliability.

Replay

Because we’re tracing all of the important details about each trajectory, it should be trivial to “replay” the trajectory. A first version of this could include:

• replay a trajectory

• edit a trajectory and replay the edited version

Replay can be used to improve observability and to reduce vendor lock-in.

Guardrails

Every LLM output will pass through a “guardrails” evaluation framework. The framework will help catch obvious errors but can also evaluate more subtle details like whether the robot’s output stayed “on brand”. A first version of this solution could include:

• factual error detection (potentially using a “side channel” LLM)

• transaction assurance: For any onchain action recommended (or generated) by the chatbot, we use SOTA tools to increase the user’s understanding of the action and confidence that the action does what they expect. For an example of what’s possible, see Stelo Labs tx simulation.

Guardrails can be used to improve stability, reliability, and personalization (reduce genericness).

Learning and Recency Bias

The best way to think about the knowledge base is a pre-computed cache of LLM trajectories. The knowledge base is “what the robot knows”. Recent data can be attached a more significant weight in the vectorbase. For any dataset that we want the chatbot to understand, we trigger a process that automates the generation of a large number of trajectories over that knowledge base. A first version of ingestion could include:

• scraping a public or private website, generating trajectories over the data, and saving the results

• providing a patching API that allows the

Learning can be used to improve cost, latency, reliability, observability, and many other aspects of the platform.

Invalidation

Knowledge invalidation is a semantic search + delete over the robot’s knowledge base. A first version could include:

• semantic search + delete conditioned on some external criteria (like data source)

Invalidation is a useful tool to improve the flexibility of the robot, allowing it to evolve unencumbered by stale understanding.

Curation and Analytics

The robot’s knowledge base, and the tracing data around it, can be used to curate the platform’s user-facing content. A first version could include:

• A report that shows which subjects are most likely to be requested by users

• A report that shows which subjects that the robot doesn’t know about are most likely to be requested

• A report that shows which subjects require the longest/shortest trajectories

Curation and analytics are useful for improving essentially all pieces of the platform. In particular, insights from the chatbot can lead to improvements in user-facing documentation and materials.

Further R&D in the Industry

Query Routing

Some questions are answered by the docs index, some are answered by the list index, etc.

https://gpt-index.readthedocs.io/en/latest/guides/tutorials/graph.html

https://twitter.com/jerryjliu0/status/1653789212620230658

Caching

Similarity caching so that frequent questions skip a trip to OpenAI https://python.langchain.com/en/latest/modules/models/llms/examples/llm_caching.html#gptcache

Hallucination rate

Compare query/response/source to see if they all match. Leads to near-0 hallucinations. (Currently implementing).

https://twitter.com/jerryjliu0/status/1645451894637367298

LLM Evaluation

On Retrieval, how do certain variables affect the result?

split_method chunk_chars overlap embeddings retriever_type num_neighbors

https://github.com/PineappleExpress808/auto-evaluator

Guard Rails

These strictly check the output of the LLM, and ensure it conforms to certain patterns or text. Nvidia has an option shown here, and there is a separate project (that came out first at: getguardrails.ai

https://twitter.com/NVIDIAAIDev/status/1650887287494901763

Back to the lab! (Appendix links below)

Research and Theory

• react https://arxiv.org/abs/2210.03629

• reflexion https://arxiv.org/abs/2303.11366

• self-refine https://arxiv.org/abs/2303.17651

• Self-Ask https://ofir.io/self-ask.pdf

• self-consistency https://arxiv.org/pdf/2203.11171.pdf

• auto fine-tuning https://arxiv.org/abs/2205.00445

• mrkl https://arxiv.org/abs/2205.00445

• generative agents https://arxiv.org/abs/2304.03442

• camel https://github.com/lightaime/camel

• language model cascades: https://arxiv.org/pdf/2207.10342.pdf

• factored cognition: https://primer.ought.org

• prompt chaining https://arxiv.org/pdf/2203.06566.pdf

• chain-of-thought prompting https://arxiv.org/pdf/2201.11903.pdf

• zero-shot chain-of-thought https://arxiv.org/abs/2205.11916

• algorithmic prompting https://arxiv.org/pdf/2211.09066.pdf

• HuggingGPT https://arxiv.org/pdf/2303.17580.pdf

• OpenAssistant https://drive.google.com/file/d/10iR5hKwFqAKhL3umx8muOWSRm7hs5FqX/view?pli=1

• InstructGPT https://arxiv.org/abs/2203.02155

• HyDE https://arxiv.org/pdf/2212.10496.pdf

• Self-supervised learning guide https://arxiv.org/pdf/2304.12210.pdf

Projects and Tools

• https://github.com/hwchase17/langchain

• https://twitter.com/gpt_index

• https://shreyar.github.io/guardrails/

• https://github.com/transmissions11/flux

• https://github.com/jbrukh/gpt-jargon

• https://github.com/microsoft/JARVIS

• https://lmql.ai/

• https://github.com/microsoft/semantic-kernel

• https://jamesturk.github.io/scrapeghost/

• https://haystack.deepset.ai/

• https://github.com/LAION-AI/Open-Assistant

Relay’s products — Robot, Receiver, Recon — each give users more power and choice in how they communicate and interact with all web3 users, personas, and companies—any address at all.

Recon is the registry, a directory of curated valuable addresses to discover and have on hand. The address book of web3.

Receiver is the universal inbox for wallet messaging, a conduit between addresses that lives on relay.cc (and is also open-source for any site to add).

Robot is a machine, the agent, the utilization of a burgeoning new technology, coming into the crypto space and changing the way we interact with chains and apps.

In sum, these products make Relay the intersection of people, addresses, and purposes as the web3 nexus.

What will happen when every entity has an address?

Person. Business. Wallet. Contract. AI.

Relay is about connecting addresses in the decentralized web and making it easy for people to find, learn about, and communicate with addresses that bring them value. In all situations, Relay has users covered.

Relay Receiver: Web3 Messaging

The heart of Relay’s offering—what we spent 2022 building—is Receiver, our end-to-end encrypted, fully decentralized (via XMTP), and open-source wallet messaging widget. ENS, Lens, and 0x addresses all are compatible to message each other, making Relay a universal solution for web3 communication. It’s also how we establish user communications for the apps in Recon, and the means by which users talk with Relay Robot.

The magical part is that Receiver is one of many XMTP clients, so users of it can message anyone on many different apps, including Orb on Lens, and Converse or xmtp.chat. Installing theis

For developers, we invite you to install it on your site with a few lines of React code or signup for an integration call with us.

Relay Recon: The Web3 Registry

We understand that in order to fully unlock the potential of the decentralized web, it’s essential to have an up-to-date, accurate directory of apps and how to reach them. By our count there are over 1500 quality apps in the EVM world that people need to know about, and that number has been growing exponentially. For each one Relay will verify and connect their ENS for prospective users to discover, message and onboard.

Recon has over 600 curated companies and apps, including important sections of ImpactDAOs, DeFi, Lens, Music, Infrastructure, Venture Capital, and Eth Events.

This puts us in the unfortunate situation of being a centralized gatekeeper for companies and their verified ENS. So we are currently in the Tableland Pilot Program to experiment with dynamic data and decentralized access control for the data and credentials attached to each company address. The bright side of this type of centralization is that Recon is the perfect candidate for a small public good “hyperstructure”—with protocol income potentially coming from web3 attribution solutions such as Spindl.

Relay Robot: The Future of Support and Education

Launching today, our OpenAI agent solution brings a new level of efficiency and convenience to customer support, onboarding and education in the decentralized web. Trained on the documentation of the most influential web3 apps, our Robots are project experts that field users questions on-demand.

Developers can ask away, too. Relay Robot can ingest, analyze, and output code, meaning it can provide SDK code assistance to developers (note: code analysis is in beta, please signup for the waitlist here).

We’re very excited about Robot, and pushing the limits of the new LLM technology stack—OpenAI, GPT Index, and LangChain—while giving it the expertise and capability to access a user’s crypto tokens in order to help them with onchain transactions.

Empowering your web3 persona

Relay is building the web3 stack for everyone, open-source and in the open. Our suite of products is designed to work together, seamlessly connecting addresses and adding new possibilities to what you can do when you connect a wallet to a site. Relay ensures users won’t struggle to find the right address, learn about an address, or communicate with an address ever again.

So go discover Recon, message with Receiver, and learn about the web3 world through Robot.

All at relay.cc.

Trained on the documentation of the most influential web3 apps, the 14 Robots are experts on a company or protocol’s documentation located at specific ENS addresses that can be interfaced with through our wallet messenger Receiver or any XMTP client.

This combination creates the first-of-its-kind interaction between your web3 identity and a “smart contract concierge”. What is amazing is by attaching a wallet to the conversation context, the agent can interact with the blockchain on your behalf.

We see four immediate use-cases for these agents:

1. DeFi Education and DAO Onboarding

2. Customer Support with Discord Bridge

3. Technical Developer Assistance (DevRel and SDK help)

4. Smart Contract Concierge (Uniswap v3 and ENS Registration)

Customer Support

What’s the currently typical way users interact with web3 projects? The answer is almost always a Discord server, with a complicated weave of roles and permissions to simultaneously keep out bots and guide user onboarding.

While workable, how can a viral project handle thousands of DMs and requests for allowlist in a #general channel or Twitter inbox?

And how are mature projects supposed to scale customer support ticketing queue to tens or hundreds of thousands?

We’ve recently launched our curated directory of web3 companies, and what sets it apart from others is connecting their ENS or Lens for easy wallet communication. Building off that, we have Relay Robots manning the lines as the frontline support agent, guide, and devrel assistant. Users and customers will be happy to query an intelligent ChatGPT bot that searches and solves issues conversationally.

Lastly, Relay Robot will soon integrate with Discord (and other CRMs) for transparency and escalation to a human agent if needed. Customers can also use their project tokens or NFTs to unlock escalation automatically.

Education

Steep learning curves are a regularity among web3 projects and protocols, especially for users entering the space. Whether it’s a new DeFi protocol, DAO structure, or infrastructure SDK, there’s a lot of reading and research that comes with engaging in these new paradigms. Potential users bounce off when they feel the need to study docs or see a discord channel full of noisy questions.

Robot saves users from digging through docs, by answering questions directly and linking source material to users. Those brand-new to a company can literally type “What is this?” or “How do I get started?” and receive meaningful answers that accelerate their engagement, making them less-likely to leave a useful but complex project. And when you’re looking for that very specific fact, such as a governance process quorum number or that synthetic staking token yield rate (and who hasn’t), the Robot also can call up that information with a question.

Developer Assistance

Pushing capabilities further, Relay Robot can serve more technical protocols that are laying a foundation for developers to buidl on. Robot can be trained on a specific project’s codebase, then asked programming questions. It will output answers in code and helpfully explain function and variable purposes.

This means Robot can help developers build on a project’s SDK while simultaneously freeing the project’s team from answering questions. Developers can also check their work, as a project’s Robot is specialized with context that makes it more capable than Github Copilot. (Note: this is in a closed beta, please fill out the waitlist if interested)

Smart-Contract Concierge

The prior three uses for Robot all could be done in a web2 way, as a simple question-answer bot that analyzes the context its given and instructions it receives.

Relay Robot’s web3 architecture makes it an interoperable part of a web3 stack (vs. the dead end info silo of a web2 version), which means it can read the chain and engage with DeFi, DAOs, and other protocols.

Live next week (Feb 7th), you can message with swap.relaycc.eth for Uniswap v3 trades, and message register.relaycc.eth to complete an ENS name registration.

To mitigate any problems due to the Robot pre-loading the form incorrectly, we are including the Fire popup for an additional visual confirmation of the transaction the user is about to enact.

It’s easy to imagine a smart contract “concierge” Robot trained in a certain protocol to provide insight and perform actions. Such a Robot would analyze contracts and transactions ahead of time, and then crafts transactions to submit based on users’ conversational input.

Relay 🤖

The first class of 14 Robots are live now for the open beta demo; try them out at relay.cc and see how helpful and knowledgable they can be—and please excuse any wrenches in the machines during this demo period.

Relay will be refining and improving them in all areas, and we’re looking for interested partners. Sign up for the waitlist to explore AI-onchain with us!

Inspired by the ecosystem landscape posts for web3 music and Lens, we created the easiest way to discover quality web3 projects. Go to relay.cc and explore!

On top of finding curated projects, you can also message them. The directory is integrated with Relay’s Receiver chat, enabling messaging to an app’s ENS or Lens handle. This messaging is end-to-end encrypted on the XMTP network, open-source, and also free to implement on any (React) website.

We envision this will be a useful tool for customer support, technical questions, and DAO onboarding. Notably, we will have inbox filters and sorting (based on address tokens and NFTs) to allow a company to more easily respond to inbound messages.

2023 Decentralization

While useful, Relay doesn’t want to be the centralized gatekeeper for apps and their data & links.

We imagine each app card as an NFT, with dynamic and configurable data using Tableland, controlled by community edits in a DAO-like way.

The implementation would be similar to the web of trust anti-phishing project MobyMask. Being able to be 100% sure of the correctness of a company’s ENS name is a requirement, as it is attached to user messaging.

We will also move the website hosting to Arweave, to decentralize the hosting and deployment of the relay.cc website.

Discover

We invite you to explore Relay's directory, learn about the new projects being built, and try out a new app. More projects are being added every week.

If you have an app you’d like to see in the directory, please fill out this request form https://airtable.com/shrO74Gi0jm9C5b6W

If you’d like to connect your company’s ENS to enable messaging, please DM @relay_eth on twitter, or even better, use relay.cc to message relaycc.eth!

Because Receiver users are web3 wallets, it becomes possible to change permissions and visibility based on the contents of those wallets—compared to email addresses, which are simply strings of characters and contain nothing. The customizability of this wallet dynamic leads to finer-tuned functionality for viewing your inbox. And what else can you do with the contents of a wallet? While messaging with another address or app, you should be able to trade and interact with them.

The best context for on-chain actions

The biggest consequence of wallet chat is the ability for messages to interact with the contents of communicating wallets themselves. Inside Receiver is the best place for two users to conduct a trade or complete a 2-party transaction, because it allows all relevant information to stay tightly woven to the context of the conversation.

Planned Integrations include:

• Opensea Seaport using the new Gossip protocol

• Uniswap trades using their Universal Router

• Lens follows and collects

• Snapshot proposal voting

On-chain integrations are the ultimate form of contextual communication, and will easily expand beyond NFT commerce into more DeFi, swaps, and governance activity.

Your inbox: restrict it, filter it, sort it.

Token-gating allows rules to be set where only wallets meeting certain token criteria can participate. Limiting the discussion to relevant parties (such as holders of a specific NFT like a POAP, a DAO token, or delegation relationship) lends functionality towards maintaining a good signal-noise ratio.

For commercial purposes, a limit like “only allow wallets with enough funds to actually purchase my BAYC to communicate with me" can be extremely time-saving. Additionally, we’ll detail in a separate post how we are implementing group chat with token requirements, difficult to do with a decentralized and open chat platform!

The same wallet content checking can be extended to create inbox-filtering mechanisms that sort messages or toggle visibility based on the sender's wallet. Because wallet chat is an open protocol (like email), it's crucial to develop anti-spam measures that keep user inboxes tidy, sorted, and relevant.

The fastest way to add wallet chat

With Relay chat we’re excited to introduce things impossible to do on Telegram or Discord, helping you with both more powerful tools and more relevant conversations with other addresses and apps. Read more about getting Receiver on your site on Github and reach out to us with any questions on our Discord.

This process strips away the context of our being—the obvious physical facts of our appearance, age, and history that make us who we are. This transformation is double-edged, both the root source of joyous online anonymity and a persistent thorn in the key vital of identity. Online interaction is novel, but identity is needed all the same to keep track, to keep score, to allow reliable reputable commerce, organizational structure, and social interactions.

Web1 was relatively identity-less, at least compared with the tidal wave of KYC, identification software, and document digitalization that arrived with Web2. The transition to Web3—with its expectation of improved data privacy, data custody, default encryption, and a “sufficient” level of decentralization—further removes us from legacy methods of tracking identification.

Here it is useful to winnow the broader discussion of identity down to “digital identity”, the identity created and used once a user is already online (and onboarded to whatever network).  Email, social media profiles, bank accounts. Things you log into and assume to carry out your online behaviors. Web3 has shifted the frontier from the creation and securing of these to the iteration and improvement of digital identity itself.

There are several serious goals here.  Freeing digital identity from the walled gardens and network effects of centralized services creates persistence and weakens monopolistic hegemonies, empowering the portability and ownership of data. Owning a persistent digital identity and the associated data history allows users to capture value created by the participation and contribution in platforms. This dynamic will destroy the popular “free” social paradigms that rely on externalized, shadowy backrooms auctioning off user data.

That’s how the theory goes, at least. The vision is grand, and still at the start.  Auspiciously, the architecture of blockchain design has brought with it a new type of digital identity, the native account of web3—the wallet.

Previous iterations of web advancement began with messaging, and later expanded to financial uses and smart functionality. Conversely, Bitcoin was in large part a reaction to the behind-closed-doors mismanagement of markets that led to the 2008 financial crisis. The priority was always on money, payments, and the transparency of those things.

Put simply, web3 went money first, and now it’s looping back to messaging and social functionalities. The same wallets designed for non-custodial & trustless commercial interaction retain these qualities for other purposes, specifically as the identity primitive of web3.

Blockchain architecture makes wallets persistent and universal. Wallets can connect with other wallets without any platform requirements, unconstrained by walled gardens or sunken-cost network effects of past behavior, buy-in, and data accumulation. Wallets can go anywhere with their data, which means interactions can’t be artificially limited. The data doesn’t stay on any one platform, the data is the commons of web3.

Equipping wallets with communication functionality means they are closer than ever to being a robust solution towards web3’s identity needs. Building reputation and permission systems on top of wallet architecture (while leveraging wallet activity, holdings, etc) could plug a major hole in Web3, sharply increasing usability and proliferating use-cases.

Relay is working on all these things. Wallet chat built on top of XMTP, profiles that aggregate wallet history and social graphs—Lens, Gitcoin Passport, Disco, to name a few—plus the token-gating and message-filtering needed to ironclad the web3 experience from spam and abuse.

Try out Relay Receiver now, easily get it on your React site, and follow us on Twitter.

Who is messaging me? Who is this 40 hexadecimal character address? or who is this ENS name?

So we made something better than sleuthing through Etherscan for context.

Introducing Relay Profiles, an aggregator of your web3 identity.

There is now a rich set of public data points for an address, and now is the right time to start showing them off and creating new connections.

We already have ENS for the Ethereum-wide account username. So what else is there? This summer saw the launch of: new decentralized social graphs — Lens new verifiable credentials — Disco, MintKudos, POAP new Sybil resistance — Gitcoin Passport

These combine into a strong and diverse “portfolio” of your web3 identity, and Relay is aggregating these into one simple profile. You can share yours, or you can learn about others.

Yet there is an issue with this. So far every web3 social interaction requires finding the address owner on Twitter, or Discord, etc in order to communicate.

The fix to this is a very natural pairing with our current relay.cc wallet message client — decentralized & encrypted with XMTP — now web3 profiles can message each other and web3 messaging has the context of profiles.

This is such a clear benefit of connecting open protocols, and only possible now by building on the work of XMTP and Lens.

We hope you find Profiles and Relay messaging extremely useful, and will be continually adding new web3 identity integrations. Reach out to us on Twitter or message the creator at relay.cc/u/seanwbren.eth.

Bonus:

Not convinced? Here’s a list of different user experiences this solves:

Going from profiles to messaging, you can explore Relay Profiles of:

• NFT collectors, to message them buy or sell offers (with asset-validation)

• DAO members, message them about DAO votes or updates or calls

• DAO delegates with influence, message them about a proposal

• NFT creators, message them about past/upcoming mints

• Hackathon/conference friends, see what DAOs/events/NFTs they like

• Fellow members of your NFT collective, such as BAYC

Going from messaging to a profile, you will need Relay Profiles if:

• Someone you don’t know messages you on Relay and you want more context on who they are

• Someone messages you about a friend or a possible new hire, and you’d like to see their profile

• A friend messages you out of the blue and you’d like to see what’s been happening in their life

• A new user of your project messages you for support help and you see if their account has the correct NFT

• You join a new DAO group chat and you want to see what other DAOs the members are part of

• As an NFT creator/collector, someone messages me about buying a valuable NFT and I go to their profile to see if they have the assets to do so

• You’re in a large telegram group for an event and you’d like to know more about who’s in it, so you look up their ENS name on Relay

  So look out for relay.cc profiles, and get your Lens handle if you don’t have it yet!

Introducing Relay Receiver, a no-code wallet chat that websites can implement to easily establish web3 communication between two users or between a user and the site.

Importantly, Receiver keeps users on-site for messaging. It eliminates the common case of someone leaving and then never completing a signup or finishing a trade.

As Receiver is connected to the website, it allows notifications and updates to be sent directly to the user wallet.

Building with XMTP always means wallet chat is encrypted and decentralized. Using Receiver means a website is wired up with all web3 users, everywhere, automatically.

Three common use-cases for Receiver:

• Lets web3 sites enable their users to easily chat, connect, and conduct business without having to build their own chat and database

• Lets web2 sites become web3 integrated and offer features to wallet users, saving a lot of work with sign-in flow to their UI

• Has Intercom-like functionality with Receiver Pro, establishing support agent workflows for interacting with web3 users (chat bot support coming soon!)

We think every site could use wallet chat and Receiver will be the easiest way to add it. The release is open-source as an npm package, and while it’s as easy to install as a single React Component, we’re here to help if you need it.

Follow @relay_eth on Twitter to stay informed, join our Discord here https://discord.gg/pyg6ATzB, or try out Relay now and message us at relay.cc/seanwbren.eth.

Foundations

A few months ago, in the process of designing and creating a feature-rich DAO governance dashboard, a key requirement revealed itself. Our proposal drafting and delegate interactions need messaging to function, and to do that with a long-term view, the messaging must adhere to web3 standards of decentralization and encryption.

Our search for solutions led us to XMTP. XMTP developed the ability to communicate between existing crypto wallets via its own encryption protocol and decentralized network.

When XMTP released their beta in April, we were on it within a week. Developing the messaging guts needed for governance operations led us to create the first XMTP generalized messaging client.

This work plus our relationship with XMTP further convinced us how much sense web3 native messaging makes, and how little progress has been made so far.

Due to this greater and more foundational scope for communication tools and identity, we need a new name: Relay.

Web3 needs messaging solutions that leverage the shared data commons of a user’s address. We see clear next steps to unlock new relationships of web3 identities, dapps, and smart contracts. Wallet attributes such as on-chain transactions, tokens and NFTs held, DeSoc reputation, and more will combine with XMTP messaging functionality to create meaningful interactions--some never seen or imagined before. Powerful integrations in chat panels will be the control panel for your web3 life.

Relay

Relay is multiple tools: A wallet chat inbox (direct messaging, group chats, and token-gated channels), profiles, a Receiver chatbox to embed in other sites, and the admin Broadcast functionality.

Here's quite a few examples of what will be live on relay.cc this summer, and how it can be used:

• Wallets can be messaged directly when issues arise with dapps, protocols, DAOs, smart contract vulnerabilities. This solves “how do I talk to this anon wallet”, a common problem and part of the famous 25,000 Ether rescue by samczsun

• Authenticated messaging from company wallets solves the impersonation and scam problem of trusting Discord and Twitter

• Group chats between multiple wallets. No explanation needed here.

• As a user, you can opt-in to receiving important messages, such as Snapshot Proposal vote alerts. Consequently, NFT ownership will govern what users and what companies will be allowed the privilege of messaging you.

• Profiles: aggregate NFT & token wallet holdings plus Lens social plus Gitcoin Passport in one place makes it easy to start a conversation around any one of these

• Relay Receiver: a drop-in chatbox for any website to support Relay messaging, saving dev time and database setup for a feature every site needs

  • For NFT collectives, DAO membership, marketplaces & exchanges, web3 music, and any site that should support a user as a wallet. Talk to us now about adding Receiver for your users

  • We will support a pro version (like Intercom) for companies/brands to have a verified wallet for customer support, integrated with modern chatbot help

• Musicians/artists can Broadcast private messages to their NFT holders -> allows high-value relationships with the person behind the wallet. More details in blog post

All of this is underpinned by an encrypted and decentralized data messaging layer that works natively with web3.

The Relay platform is full of tools, features, and innovations that run wire and connect the space between and among wallets. Message everywhere using your web3 identity.

We’d love for you to try it out, either by a message to us at relay.cc/seanwbren.eth , or share out your own ENS to chat with anyone you wish!

P.S.

We still believe in the huge potential for DAOs and a growing need for better delegate and membership tools. With the chief role they have in the coordination and organization of online communities in web3, we know the work we put into the Athens Governance Engine will come to fruition next year, yet after Relay. The foundational and immediate unlocks that wallet communication give us need to be made first.

This is why we created Broadcasts, a first-of-its-kind messaging feature from Relay that allows artists and musicians to broadcast and reach their audience directly and privately, targeted to holders of their NFTs that were distributed to fans through online sales, physical albums, concert tickets, or however artists choose.

The scarce ownership of NFTs forms a new corridor of value creation that grants artists the ability to focus on a core fanbase directly. With NFTs as a "super subscription” that creates dynamic membership clubs, artists have freedom of creativity in how they curate the fan experience. Fans holding artist NFTs can receive concert tickets, early access to merch, invites to meet-and-greets, airdrops, giveaways, and special content like secret songs, artwork and music videos.

Relay’s Broadcasts allow this communication from artists to NFT holders to be done natively, in a one-to-many broadcast that feels personal to fans. Artists can specifically message holders of their NFT collection, reaching these super-fans without relying on central platforms or the manual creation of something like an NFT Discord server.

Relay's messaging is always tied to wallet addresses, built on the open-source data layer of XMTP that means Broadcasts are guaranteed to be end-to-end encrypted with peer-to-peer message passing. This unlocks a completely new type of fan community, empowering artists to move beyond app-based support, beyond platforms that capture the lion’s share of value.

Artist NFTs in combination with Broadcasts means this value can now be kept between the expression of the artist and the love & loyalty of a fan, where it belongs.

To be among the first to try Broadcasts, message us on Twitter @relay_eth or try out XMTP messaging on relay.cc/seanwbren.eth to join the closed beta.

DAOs adapt to changing conditions through the proposal process, with members drafting and passing high-value proposals that impact DAO operations. The lifecycle of a proposal is comprised mainly of a few people drafting it, a small amount of people discussing it, and a larger number of people voting on it.

The inefficiencies and pain points in this process are causing DAO frameworks to trend towards greater decision-making power for smaller groups of delegates. Empowering delegate agency allows DAOs to pass successful proposals with greater velocity, and in turn the DAO demands more accountability and expert knowledge from delegates.

Relay aims to be the communication medium delegates use and trust to govern well. A combination of communication, tracking, and reputation solutions will equip delegates with the tools needed to have productive and authentic rapport with DAO members and contributors.

We've already launched the first piece of these solutions, relay.cc, consisting of:

• 1:1 encrypted, decentralized, and open-source wallet to wallet chat

• ENS-aware

• Broadcast functionality: DAOs to members, Delegates to their delegators, and NFT creators to the holders of a collection

With this chat cornerstone in place, Relay is soon releasing what we call the Athens Governance Engine: a powerful virtuous cycle of roles, actions, and features that empowers delegates and generates positive governance results. [Aug 2022 edit: pushed to early 2023]

Athens Governance Engine features:

• Proposal Lifecycle Dashboard

  • Token-gated drafting stage

  • Customized “tracks” or checkpoints depending on the impact of a proposal

  • Public read / token-gated write for discussing

  • Voter Guides (public delegate viewpoints on the merits of a proposal)

  • Vote off-chain, much like Snapshot, with integrated discussions

• Delegate Profiles

  • Data-rich and professional public profiles

  • Voting history and participation rates

  • Self-described background and voting ideals

  • DAO membership and NFT gallery

  • EPNS integration for polling and notifications

  • LensProtocol social graph integration

  • Gitcoin Passport Proof of Personhood integration

  • Encrypted messaging with DAO contributors and their delegators

• DAO Operations

  • Token-gated, public chatrooms

  • Post-proposal data and comments from the DAO, tracking the impact of proposals

  • Dune Analytics integration to track DAO health metrics

These features in aggregate will allow delegates to craft better proposals in a space designed for promoting intelligent debate. The end goal is to supply delegates with the tools to govern well.

DAO governance requires a neutral arena for all parties and viewpoints, so Relay believes we must adhere to the principles of being an open-source public good. With this in mind, Relay aims to make the frontend as permissionless as possible to function with any DAO governance framework.

This decision necessitates the creation of safeguards against malicious actors in the DAO/token space. In the short term, Relay will work with DAO admins and NFT creators, with clear plans to move to a decentralized system of allowlists/denylists (similar to this implementation). More on that in a future post.

If you find these topics exciting, we invite you to join our Discord, follow on Twitter, or start your own encrypted web3 chat by sharing xmtp.chat/dm/[your0xaddress]!

Built on top of XMTP, Relay is a foundational piece of our public goods governance infrastructure. Panel is focused on building tools for delegates to govern well, and there is a requirement for uncensorable and private web3 communication for the DAO decision makers. Internal subjects such as treasury movement, DAO to DAO token swaps, and delegate messaging should be encrypted and be done with your ENS or wallet identity, with complete privacy.

Of course Relay is usable by anyone for any reason, with some use-cases highlighted below.

Relay Features:

• XMTP client provides end-to-end encryption between ENS names or any chain’s wallet (EVM-specific to start, more soon)

• ENS avatars

• Browser notifications

• Mobile support (Metamask browser and WalletConnect work best)

Upcoming:

• EPNS notifications

• Delegate to delegator messaging and polling

• NFT creator to holder chat

• Token-gated chatrooms

• Upgrades including rich media content and emojis

• Lens Protocol social graph connections

Beta Disclaimer: XMTP is nearing its production-ready release, so expect some minor bugs and crucially **your chats being deleted every week or so** as they clear the network. We are working closely with them to deliver the best possible client for their exciting and very needed protocol.

That's it for now, to start. Over the coming weeks Relay will be sharing more about our convictions and roadmap to improve governance velocity and outcomes.

With a team history going back to MakerDAO in 2017, we care deeply about DAO coordination. Relay will be the place for delegates and DAOs to discuss, share, debate, and succeed. We're excited to share the governance dashboard we’ve crafted; built to support proposal drafting & voting, delegate profiles, and token-gated discussions (ETA: July).

Relay also believes that any trusted and private governance platform must be a public good, informing our own plans to dissolve into a DAO ourselves. But more on that later.

We hope you love the secure and private chat we've built with Relay. Let us know what you think, and if you care about DAOs and delegate governance join our Discord and get in touch.

https://discord.gg/5N3wpzNjTT

Twitter: twitter.com/relay_eth