What does a Web3 marketing agency actually do?

We turn protocols and tokens into categories. That means strategy (what bet are you really making?), narrative (what do you stand for?), community (who shows up for you?), and launch motion (TGE, mainnet, product). For AI startups, the same playbook adapts to product launches and developer adoption.

How is this different from a crypto KOL agency?

KOL deals are a tactic, not a strategy. We sometimes use them, but they're maybe 10% of what good Web3 marketing looks like. The other 90% is positioning, owned content, community design, and earned distribution that compounds.

Do you work with pre-launch projects?

Yes. Most of our best work happens before token launch — the strategy and narrative decisions you make pre-TGE are the ones that compound (or haunt you) for the next two years.

Web3 + AI Marketing Glossary

Plain-English definitions of 25 terms commonly used in Web3, AI, DeFi, and crypto marketing. Every term is anchor-linkable as /glossary#term-slug so you can deep-link from anywhere.

TGE (Token Generation Event)

The point at which a project's token is created and distributed. Usually paired with mainnet launch or a public sale. The TGE is the single highest-stakes event in a token project's life — get it right and you have years of momentum, get it wrong and you spend years recovering.

Tokenomics

The economic design of a token: supply schedule, distribution, utility, vesting, burn mechanics, governance rights. Good tokenomics align the interests of founders, investors, contributors, and users. Bad tokenomics enrich one group at the expense of others, which usually shows up as a death spiral in the first 12 months.

MEV (Maximal Extractable Value)

The profit a block producer can extract by reordering, including, or excluding transactions. MEV is morally neutral but economically significant — on Ethereum and Solana, MEV redistribution shapes which validators win, which apps work, and how user transactions get priced.

LP (Liquidity Provider)

An entity that deposits assets into a DEX or money market to earn fees. Marketing to LPs requires different messaging than marketing to traders: LPs care about capital efficiency, impermanent loss protection, and yield stability. Traders care about slippage, depth, and execution speed.

KOL (Key Opinion Leader)

A creator with audience and credibility in a specific niche. Crypto KOL deals are a tactical channel — useful occasionally, never strategic. The bigger the KOL, the more expensive and the lower the conversion. We rarely recommend KOL spend unless the project has specific narrative they want validated.

DAU / MAU (Daily / Monthly Active Users)

The headline metric for product engagement. In Web3, DAU is more meaningful than follower count or token holder count — it measures whether anyone uses the thing.

TVL (Total Value Locked)

The dollar value of assets deposited into a DeFi protocol. TVL is a useful proxy for adoption, but it's gamed by mercenary capital that chases yield. Sticky TVL — money that stays after incentives end — is the real signal.

RWA (Real-World Assets)

Assets from outside crypto (treasuries, real estate, credit, commodities) brought onchain. RWA is one of the largest unlock vectors for Web3 — a $500T+ offline asset market potentially migrating to onchain rails.

DePIN (Decentralized Physical Infrastructure Networks)

Networks that coordinate real-world infrastructure — wireless (Helium), storage (Filecoin), compute (io.net) — using token incentives to coordinate physical hardware. DePIN is a category bet: that token incentives can coordinate physical work better than centralized companies.

ZK (Zero-Knowledge)

Cryptographic technique for proving a statement is true without revealing the underlying data. ZK is foundational to scaling (zk rollups), privacy (zk proofs of identity), and interoperability (zk bridges). For marketing: most users will never understand how ZK works, but they'll feel its effects (faster, cheaper, private).

L1 / L2 (Layer 1 / Layer 2)

Layer 1 (Ethereum, Solana, Bitcoin) — the base settlement layer. Layer 2 (Arbitrum, Optimism, Base) — execution layers built on top of an L1 for speed and cost.

Mainnet

The production version of a blockchain or protocol. Distinguished from testnet (where developers experiment without real money). Mainnet launch is usually paired with TGE for token projects.

Restaking

Using staked assets as economic security for additional services beyond their original chain. EigenLayer pioneered the category on Ethereum.

Airdrop

A distribution of tokens to users based on past activity. Airdrops can build communities (early Optimism, Arbitrum) or destroy them (badly designed ones invite Sybil attacks). Airdrop design is its own subdiscipline.

Sybil

An attacker who creates many fake identities to claim disproportionate airdrops, governance votes, or rewards. Sybil resistance is the central technical challenge of every Web3 incentive design.

Narrative

In crypto: the dominant story explaining why a category will grow. Narratives are the units of mindshare. 'L2 summer,' 'RWA narrative,' 'AI agents narrative' — projects that align with active narratives outperform those that don't, even with worse fundamentals.

Mindshare

The percentage of attention in a niche that a project commands. Mindshare is tracked qualitatively (who do people mention when asked?) and quantitatively (X follower growth, search volume, podcast mentions). Mindshare leads price.

Vampire attack

Marketing strategy where a new protocol incentivizes users to migrate from an incumbent (SushiSwap vs. Uniswap is the canonical example). Effective in the short term, often loses to the incumbent in the long term.

Foundation model

A large pretrained AI model (GPT-4, Claude, Gemini, Llama) that serves as a base for downstream applications. Foundation model marketing is dominated by a handful of labs; the bigger market is application-layer AI products built on top.

RAG (Retrieval-Augmented Generation)

AI architecture where a model retrieves relevant documents at query time and uses them to generate responses. RAG is how AI products keep up-to-date information without retraining the base model.

Inference

The process of running an AI model to generate output for a user. Inference cost is the largest variable cost for AI products at scale. Inference optimization (smaller models, distillation, quantization) is a core technical strategy.

Open-source AI

AI models with publicly available weights (Llama, Mistral, DeepSeek). Open-source AI marketing has different mechanics than closed-source — the audience is developers and researchers, distribution is HuggingFace and GitHub, and reputation comes from benchmark performance and ecosystem adoption.

Agent

In AI: a system that can take autonomous actions to accomplish a goal, typically using a foundation model to plan and tools to execute. Agent marketing is nascent — most 'agent' products in 2026 are workflow automation rebadged.

Embeddings

Numerical representations of text/images/audio that capture semantic meaning. Embeddings power search, recommendations, and RAG. For marketing: embeddings are why AI products can do things keyword-based products can't (semantic search, similarity matching).

AMA (Ask Me Anything)

A live or written Q&A format where founders or contributors answer community questions. AMAs are a high-leverage launch tool when done well — direct founder access without filter. Done badly (canned answers, dodging hard questions) they erode trust faster than no AMA at all.

Apply these in practice

We use every term on this page in client work. Free marketing audit, our services, recent case study with Plume Network.

External: CoinDesk, a16z crypto, Dragonfly.