2023 WEB3 Startup Direction Guide

2022.12.20

2023 WEB3 Startup Direction Guide

From infrastructure to applications, from AI to ZK... 14 major赛道, over 60细分 fields.

2022.12.20 - 16:19:53

WEB3DAODeFiAI

Navigating Web3 tides with focused insights

From infrastructure to applications, from AI to ZK... 14 major赛道, over 60细分 fields.

2023 WEB3 Startup Direction Guide

Author: Qiao Wang

As 2022 draws to a close amid dancing black swans and deep bears, we welcome 2023.

What new waves and surprises will unfold in the WEB3 world in the coming year? For entrepreneurs and developers, perhaps the most important question is: Which WEB3 startup areas and directions are worth exploring and building in 2023?

This article covers some of the crypto/Web3 startup ideas and themes that AllianceDAO, the largest accelerator DAO in the WEB3 space, finds particularly interesting. If you're researching these areas, we encourage you to apply to join the community and receive funding (https://alliance.xyz/apply). If you’re unsure what to build, this piece may spark your inspiration.

This article is published in Chinese by TechFlow with authorization from AllianceDAO.

Decentralized Infrastructure
Developer Experience
Gaming
Consumer-Facing Applications
Proof of Physical Work (PoPW)
Zero-Knowledge Proofs
Appchains (as Rollups or Cosmos Zones)
Decentralized Science (DeSci)
DeFi
Crypto B2B
AI and Web3 Convergence
Verticalization of Category-Defining Products
DAOs as Risk-Sharing Tools
NFT Communities as Digital Nations

1. Decentralized Infrastructure

Decentralizing Every Layer of the Tech Stack

In recent years, Web3 seems to have lost interest in true decentralization. Memes like “users don’t care about decentralization” are everywhere, yet real decentralization remains the primary feature that makes Web3 compelling.

Decentralization only matters when it’s real. With the recent collapse of centralized crypto institutions and broader financial and social censorship globally, decentralization has never been more critical.

An application is only as decentralized as its most centralized technical layer. Therefore, for applications that require decentralization, the entire tech stack must be decentralized.

The layers of the tech stack include:

DNS (e.g., Handshake, a decentralized domain protocol)
General-purpose computation layer
Data storage (e.g., Filecoin)
Data indexing (e.g., The Graph, a decentralized data indexing protocol)
Oracles (e.g., Chainlink)
Cross-chain bridges
Node infrastructure
Fiat on/off ramps
Wallets
ISP (e.g., Helium)
CDN
Version control

It may seem many layers already have dominant decentralized players. However, most existing projects remain at best proof-of-concept prototypes. Even those that are production-ready can still face legitimate competition after weighing trade-offs. Solana is an example.

Nevertheless, these represent some of the most challenging engineering and computer science problems in Web3.

Decentralized Postgres

There's currently no way to easily build a sleek Web3 client app like in Web2. In Web2, you launch a database on an AWS instance and let your frontend call it for reads and writes. In Web3, nothing like this exists.

You can't just write data directly onto L1 blockchains—it's too expensive for most use cases. Storage protocols like Filecoin are mainly for archival, lacking enterprise-grade performance guarantees (e.g., low latency) and familiar interfaces (e.g., SQL) for data retrieval.

Application-Specific L2-as-a-Service

Projects like Axie, Bored Apes, and dYdX have launched or are considering launching their own appchains. This sacrifices shared security, atomic composability, and rapid prototyping in favor of customization, sovereignty, value accrual, and predictable gas fees. These appchains are essentially application-specific L2s.

An L2-as-a-service product tailored for specific applications would make it extremely easy for dapp developers to launch their own L2. It would handle behind-the-scenes tasks such as recruiting sequencers and validators, and launching application-specific indexers, oracles, bridges, and block explorers.

Oracles for Static Web2 Information

Currently, oracles in Web3 primarily pull high-frequency data like prices from centralized exchanges, enabling lending and derivatives applications.

We also need oracles for relatively static information from:

Web2 social platforms
Gaming platforms like Steam
KYC providers
Credit rating agencies
Banks

Semi-Custodial Wallets

Wallet infrastructure remains the biggest bottleneck for mass adoption of Web3.

By design, centralized custodial wallets sacrifice flexibility for convenience. Beyond transactions and transfers, you can’t do much with something like a Coinbase custodial wallet. We need self-custodied wallets to access dapps without restrictions.

We envision wallets that are "semi-custodial":

Users hold one part of the key, while the company holds another. Both parties sign transactions via multi-party computation (MPC). Funds remain secure even if one party is compromised, as no single entity ever holds the full key.
If users lose their keys, MPC allows recovery using a third shard stored in an independent key management system (e.g., AWS KMS).

Ideally, wallets should offer a WeChat-like "super app" experience, allowing direct access to various other apps within.

Finally, wallets should have some or all of the following features:

Support for native fiat on/off ramps;
Easy integration for dapp developers;
Multi-chain support;
Batch transactions (sign once to execute multiple on-chain transactions);
Warning users before signing suspicious transactions.

Like many verticals dominated by mature leaders (e.g., MetaMask), building a successful wallet requires identifying underserved market segments overlooked by current players.

Uncensorable GitHub

On August 8, 2022, the U.S. Treasury sanctioned Tornado Cash. GitHub immediately censored its code repository.

2. Developer Experience

Vercel/Firebase for Web3

Achieve a developer experience similar to deploying a Web2 site on Vercel—launching a static frontend, deploying to IPFS, and setting up ENS domains.

Cosmos/Solana Developer Tools

Make development on these chains feel as smooth as on EVM. Think HardHat/Foundry for Cosmos/Solana.

Smart Contract "Copilot"

Large language models (LLMs) can serve as companions for dapp developers, offering real-time autocomplete suggestions or generating boilerplate code. They can also help developers understand existing code more easily by providing human-readable summaries.

3. Gaming

Fully On-Chain Games

On-chain games store not just in-game assets like currency and NFTs, but the entire game state and logic on the blockchain.

Consider tic-tac-toe. Concepts like turns, the 3x3 board, players, and move order must be immutably recorded as rules. The history of player moves must also be logged on-chain.

Meanwhile, visual elements like board and piece colors, shapes (X and O could become Y and Z), animations, and sounds can stay on the client side.

Putting all state and logic on-chain unlocks novel behaviors:

Smart contract-based players (vs. human players) who play according to intelligent, evolving rules. Consider composable interactions between games and players. One of the earliest examples was actually MEV in DeFi. MEV bots are smart contract-based players; MEV is the game. Note that here, contract-based players compete not only against each other but also against humans.
Trustless collaboration among smart contract-based players. Consider player composability. Take the game Dark Forest. At launch, it supported informal alliances—players had to trust others wouldn’t betray them. Then someone created a smart contract acting as a competitor. This contract allowed players to donate points permissionlessly and eventually win prizes. It tracked everything transparently, enabling proportional reward distribution upon victory.
Game mods built by third-party developers without fear of platform bans. Consider game composability. Again, using Dark Forest, someone could create a single-player version derived from the multiplayer game. It shares the same core logic but changes map, UX, and scoring. Instead of competing to dominate the same map, players race to solve the same puzzle faster than others.

Game Engines

While most game engines today aim to add Web3 features to existing Web2 games, we’re excited about tools designed specifically for fully on-chain games. What if we could create interaction or physics standards for on-chain games? Imagine Dungeons & Dragons rules being on-chain, reusable and remixable by anyone.

If history is any guide, the first successful on-chain game engine will likely come from a successful on-chain game studio. Think Unreal Engine and Epic Games.

For the same reason, the first successful Web3 game publisher might emerge from a successful Web3 game studio. Think Steam and Valve.

Potential key features for a Web3 game engine include wallet integration and in-game NFT markets.

4. Consumer-Facing Applications

A Web3 Twitter That Doesn’t Want to Be Web3 Twitter

The winning “Web3 Twitter” will ultimately deliver what we’ve always wanted: user-owned data, free from control by any single entity or individual; permissionless multi-client development; and customizable content curation algorithms.

But there’s one catch. If history is any guide, the first successful Web3 Twitter is unlikely to start as something aiming to replace or resemble Twitter. Perhaps Web3 Twitter will emerge from a game, a DeFi super app, or some novel idea so seemingly trivial it gets mocked.

Web3 WhatsApp

A messaging app should be:

Privacy-preserving: Strong end-to-end encryption.
User-owned: Encrypted data stored on IPFS, controlled solely by user-held keys.
Uncensorable: Users cannot be banned by entities like Meta or governments.
Portable: Multiple independent clients can be built.
Natively commercializable: Supports native crypto payments.

Web3 WhatsApp should gradually decentralize. The first step might involve decentralized storage with centralized compute, migrating to fully decentralized compute later when the product is ready.

Intuitively, “messaging app” is easier to build and enter the market than “social.”

Start by capturing crypto-native users who appreciate the above five qualities and can onboard easily.

Habit-Forming Apps

StepN demonstrated that token incentives can effectively encourage healthy habits—running, sleeping, meditating, eating well, or learning new skills. Studies suggest forming a habit takes 18 to 254 days, averaging around 66. Overcoming this initial hurdle is hard for many, but economic incentives via tokens could provide the motivation needed to persist and make habits permanent.

A major challenge lies in designing an effective oracle system, as habit-forming apps must link off-chain actions to on-chain events.

Fan Engagement

NFTs enable creators across fields—musicians, athletes, writers, video creators—to monetize their personal brands in new ways. We’re interested in startups building fan engagement platforms across industries and geographies. We’re also open to supporting individual brands.

A key challenge for these startups is creating utility for NFTs. At the base level, NFTs are collectibles and/or digital identities. But ideally, they should be more—exclusive creator access, governance rights, royalty participation in derivative works. If NFT collections are merely collectibles or PFPs, they’re essentially indivisible versions of 2017 ICO tokens.

NFTs as Digital-Native Intellectual Property

Consider intellectual property (IP) as an asset class. To grasp its importance, intangible assets’ share of S&P 500 market cap has grown from ~1/5 to ~4/5 over the past five years.

NFTs representing individual IP can unlock liquidity across various assets:

Music, videos, writing, and art;
Proprietary technology and scientific discoveries;
Brands.

The idea is to build primary and secondary markets focused on specific asset types.

Several NFT markets exist today, but NFTs’ true potential isn’t speculation—it’s enabling derivative works, turning them into productive rather than just tradable assets.

One approach is attaching IP legally to specific NFTs, effectively tokenizing IP through traditional legal systems.

Ultimately, however, NFTs may become digital-native IP independent of legal systems, where original works gain value naturally through social consensus. In many popular PFP projects, “right-click saving” hasn’t harmed value—in fact, it acts as free marketing. Images spread widely, increasing exposure and sparking interest in originals.

5. Proof of Physical Work (PoPW)

Tokens may prove more effective than even the most capable centralized entities—governments and large corporations—at coordinating large-scale human activities. See our related research.

Health Data Sharing

A major challenge for public health researchers is insufficient datasets to test hypotheses. A possible solution: decentralized contribution of personal health data for research and drug development.

For example, individuals could share DNA data and related health info, creating a decentralized version of 23andMe, rewarding contributors. Universities, hospitals, and pharma companies could access this data via research and commercial marketplaces.

Another example: sharing wearable-collected activity data—heart rate, sleep patterns, etc. Health-focused businesses could use this data to improve products. These contributions are simple and standardized, ideal for PoPW networks. Privacy-enhancing tech like zero-knowledge proofs could further benefit health data use cases.

AI Dataset Creation

Creating large, complex datasets to train advanced AI models (e.g., in computer vision) requires massive human input. For autonomous driving models, photos of traffic conditions must be captured and accurately labeled—both steps needing significant manual effort. PoPW networks can aggregate human contributions to build datasets for AI and other uses. These datasets can then be used by companies developing and training large AI models. As model complexity increases, PoPW networks can continuously update and expand datasets to improve performance.

Regenerative Finance (ReFi)

PoPW networks can promote sustainable activities. Participants earn tokens, which environmentally conscious institutions buy and burn. The system works like carbon credits—demand stems from social signaling. Use cases include cleaning waterways, recycling, funding better industrial filters, etc.

Decentralized Energy Grids

Rapid growth in renewable energy is positive for the environment, but intermittent sources like wind and solar make energy generation unpredictable. Batteries help by storing excess energy for later use. The more batteries connected to the grid, the more useful they become. PoPW networks use token incentives to encourage households to deploy batteries and pool them for emergency use.

Talent Markets via Crypto Networks

Web2 talent markets are diverse, but hourly and contract workers face issues:

As platforms gain bargaining power, they tend to increase fees charged to talent.
Users lack ability to transfer reputation across different markets.
Using fiat payment systems for international transactions has limitations.
PII requirements during onboarding exclude many global talents.

Crypto may help address all these issues.

Decentralized VoIP International Calling

Voice over IP (VoIP) technology has drastically reduced international calling costs by routing calls over the internet. Decentralization could cut these costs by another order of magnitude.

A PoPW network of users connecting local phone lines to the internet could create a global network, allowing users to make international calls at local rates. This decentralized network could offer a far more cost-effective alternative to traditional international calling services.

Oracles for PoPW Networks

In PoPW networks, participants' off-chain contributions must be proven on-chain to earn native tokens—a challenge known as the “oracle problem,” one of the biggest hurdles for PoPW networks. Malicious actors may try to manipulate oracles to extract maximum value. Examples exist in the Helium network: since rewards depend on hotspot geographic coverage, users have incentive to fake hotspots or locations. Despite Helium’s efforts—like denylists and challenge systems—proving hotspot authenticity and location remains difficult.

Other PoPW platforms, like Hivemapper, attempt hardware authentication to prevent oracle manipulation.

Hivemapper dashcams use GPS and Helium hotspot connections as part of their proof-of-location protocol, and they employ humans to verify image authenticity. While effective, this adds complexity and opens doors to collusion between contributors and reviewers.

Efficient PoPW oracles remain an open problem—no general solution exists yet. Hardware authentication can offer protection for specific use cases, as hardware is harder to spoof. Anti-spoofing GPS modules, for example, could secure location-sensitive PoPW contributions. But more robust and universal oracles are needed to support broader applications.

6. Zero-Knowledge Proofs

See our research for details.

High-Level ZK Development Frameworks

Like TensorFlow and PyTorch in AI, high-level zero-knowledge (ZK) development frameworks are crucial for unlocking innovation at the application level. These frameworks should:

Abstract away the complexity of underlying ZK backends;
Support various ZK backends and hardware environments (CPU, GPU);
Enable efficient debugging and testing;
Provide a rich development environment with examples and tutorials.

The closest equivalents in the Ethereum ecosystem are Hardhat and Foundry, but they’re unlikely to support zkEVM or ZKP soon. Instead, existing abstraction efforts (e.g., Cairo) may evolve to fill this gap.

ZK Cross-Chain Bridges

Existing cross-chain bridges rely on strong trust assumptions and are relatively expensive. zk-bridges can offer strong trustlessness and significantly reduce on-chain verification costs—similar to how zk-rollups validate on base L1s. However, cross-chain message passing is more complex due to differing signature schemes and cryptographic functions between source and destination chains. Technical details.

ZK-Rollup SDKs

zk-rollups are increasingly popular, enabling application-specific L2s for gaming or high-throughput DeFi protocols. Here, zk-rollups handle execution and settlement, while consensus and data availability are managed by L1. However, launching custom zk-rollups remains highly complex. We believe startups offering developer-friendly SDKs for custom zk-rollups will meet real business needs and can become valuable enterprises by providing toolkits, developer services, sequencing, and supporting infrastructure.

ZK Hardware Accelerators

Specialized hardware companies targeting specific use cases and establishing early market leadership can become extremely valuable.

This has been validated in AI. Nvidia became North America’s most valuable semiconductor company by focusing on AI hardware. Bitmain, Canaan, and WhatsMiner became unicorns by specializing in ASIC miners—similar dynamics exist in Bitcoin mining. Companies designing and building efficient ZK hardware accelerators may follow the same trajectory.

Decentralized Intelligence Organizations

ZKPs could spawn decentralized intelligence organizations. In such networks, intelligence operatives, data detectives, and spies could participate without interacting or knowing each other. Participants could use ZKPs to prove knowledge of certain intelligence data, then receive private payments in exchange. Such systems could also facilitate collaborative, composable enrichment or interpretation of collected data while preserving participant privacy.

ZK On-Chain Game Engines

Incomplete information is key to some of humanity’s best games—poker and StarCraft (examples of games with complete information include chess and Go).

Dark Forest proved ZKPs can enable incomplete-information on-chain games—crucial for designing more interactive experiences where player actions remain private until they choose to reveal them. As on-chain gaming matures, we expect ZKPs to become integral to game execution engines. The opportunity is huge for startups successfully integrating privacy into high-throughput on-chain game engines.

ZK DeFi

See Privacy DeFi and Auction Markets section.

Personal Data Markets

Much of our personal data could drive positive social change—if shared with the right entities. Personal health data could be crowdsourced to help researchers develop new drugs. Consumer behavior data could be shared with brands to improve products. Yet privacy remains a persistent concern. ZKPs can enable private sharing and monetization of such data.

7. Appchains (as Rollups or Cosmos Zones)

See our research for details.

High-Performance DeFi Protocols

DeFi protocols aiming to compete with Web2 on performance will need appchains to achieve their goals. Central Limit Order Book (CLOB) exchanges are prime candidates. dYdX started this trend—we hope to see spot and derivatives exchanges built as appchains to benefit from low fees and latency. Key enablers include customizable tech stacks tailored to each DeFi protocol’s needs.

Appchain Game Engines

Limited options remain a barrier to widespread appchain adoption in performance-constrained applications like gaming. StarkEx is a popular choice. We’d like to see startups build new, efficient architectures supporting 100K+ TPS for on-chain gaming.

Games may require less composability than DeFi, making the appchain path especially viable.

Application-Specific L2-as-a-Service

Refer to Section 3 in Chapter 1.

Enabling Shared Cross-Chain Security

Products enabling shared security can mitigate appchain security challenges. Similar to merged mining in PoW chains, we envision methods allowing shared security across PoS chains—e.g., validators securing appchains by staking ETH instead of the appchain’s native token. Liquid staking protocols may play a key role in such mechanisms.

8. Decentralized Science (DeSci)

Research Funding

One of science’s biggest problems is lack of funding—rooted in the speculative, risky nature of scientific discovery. The idea: channel crypto’s “degen” energy into funding experiments. Perhaps DAOs as risk-pooling tools are the answer.

Decentralized 23andMe

Another obstacle to scientific progress is lack of data. Humans have tens of thousands of interdependent genes. Gaining deep insights from genetic data may require vastly more data points (otherwise datasets are too sparse). The idea: use token incentives to crowdsource a massive genetic dataset.

Additionally, centralized companies like 23andMe face major privacy concerns. ZKPs can shine here.

More broadly, we want to see startups use token incentives to bootstrap valuable datasets for scientific research—something previously impossible.

Tokenized Carbon Offset Markets

Despite blockchain’s promise of greater transparency and liquidity, real-world asset tokenization has historically underdelivered. One reason: for most real-world assets, legacy markets already exist—clunky but “good enough” and deeply entrenched. For example, if existing MLS* works well, what’s the point of tokenizing real estate as NFTs?

Carbon offsets may be the exception—an emerging asset class without a dominant, unassailable legacy market. Building Web3 infrastructure to tokenize carbon offsets on-chain could be the first real success story for RWAs.

MLS (Multiple Listing Service): A centralized platform standardizing cooperation and commissions among real estate brokers while listing properties.
Carbon offset: Supporting or funding projects that reduce greenhouse gas emissions to compensate for emissions caused by daily activities and their impact on climate change.

9. DeFi

Market-Driven Lending Protocols

Price discovery in most existing lending protocols isn’t fully market-driven—it’s governed by decentralized governance or rigid mathematical formulas.