PHASE 16 Specialize · ongoing

Research & Specialization

At this point you've shipped a production-grade dApp. The field is too big to master all of it — pick a lane that matches your taste, and go deep.

Goal — pick a specialization, commit to reading real papers/codebases, and make visible contributions.

1. The five honest lanes in 2026

Lane	Day-to-day	Background fit
DeFi protocol engineer	Mechanism design, Solidity, market math	Backend + trading / finance
Security / auditor	Break things for a living	Adversarial mindset, deep EVM
Infrastructure	Nodes, RPC, indexers, rollup stacks	Distsys / SRE background
ZK engineer	Circuits, proving systems	Math-curious, Rust/C
Wallet / account abstraction	UX, EIP-4337, passkeys, key management	Frontend + security

2. DeFi track

Read: Uniswap v2 & v3 whitepapers, Curve StableSwap, Aave v3 docs, GMX v2, Pendle.
Build: a lending market with oracle, liquidation bot, partial-liquidation logic.
Follow: t11s, Dan Robinson (Paradigm), Hasu.
Jobs: Paradigm, Uniswap Labs, Aave Companies, Synthetix, Frax.

3. Security / auditor track

CTFs: Ethernaut, Damn Vulnerable DeFi, Paradigm CTF, Secureum bootcamp.
Tools: Foundry (forge test --fuzz-runs 1m), Echidna, Slither, Halmos, Certora.
Read post-mortems on rekt.news; cross-reference with the actual on-chain tx in Tenderly.
Start on Code4rena / Sherlock / Cantina contests; top finders get offers.

4. Infrastructure track

Run a validator on testnet. Sync Erigon, Nethermind, Reth. Compare.
Dig into OP Stack, Arbitrum Orbit, zkSync Hyperchains — launch your own app-chain.
Data: Dune, Flipside dashboards; contribute queries.
Jobs: Optimism, Arbitrum, Polygon, Conduit, Caldera, Gelato.

5. ZK track

Start: zkHack, Circom docs, Proofs, Arguments, and Zero-Knowledge (Justin Thaler).
Build: zk-password, zk-Sudoku, zk-Merkle airdrop, zk-KYC.
Graduate: Halo2, Plonky2, RISC Zero, SP1.
Jobs: Matter Labs, Scroll, Polygon Labs, =nil;, Risc Zero, Succinct.

6. Wallet / account-abstraction track

Read EIP-4337 end to end. Understand bundlers, paymasters, entrypoints.
Build: a smart wallet with passkey (WebAuthn) auth, session keys, gas sponsorship.
Integrate: ZeroDev, Pimlico, Biconomy, Alchemy AA.
Jobs: Coinbase Wallet, Argent, Safe, Privy, Dynamic.

7. Staying current — a low-noise diet

Weekly — Week in Ethereum News, Bankless podcast, rekt.news.
Research — ethresear.ch, Paradigm blog, a16z crypto, Vitalik's blog.
Code — subscribe to OpenZeppelin, Foundry, ethers, viem release notes.
Twitter — narrow follow list; mute the rest.

8. How to contribute to OSS (for real)

Pick one repo you've used (e.g., ethers.js, foundry-rs, openzeppelin-contracts).
Read the CONTRIBUTING.md. Fix a "good first issue".
Write one good bug report per month — reproducible, minimal, with a PR if possible.
Participate in governance forums where you're a user (Compound, Uniswap). Temperature-check proposals.

9. Reading list — the canonical 15

Bitcoin whitepaper (Nakamoto, 2008)
Ethereum Yellow Paper (Wood)
"Ethereum is a Dark Forest" (Dan Robinson) — MEV origin story
Uniswap v2 whitepaper
Uniswap v3 whitepaper (concentrated liquidity)
EIP-1559 (fee market)
EIP-4844 (blob transactions)
EIP-4337 (account abstraction)
Optimism: Cannon & fault proof design docs
zkSync Era Litepaper
"An Incomplete Guide to Rollups" (Vitalik)
"Flash Boys 2.0" (MEV paper, Daian et al.)
Paradigm: "The Dark Forest Revisited" & MEV-Boost docs
"SoK: Decentralized Finance" (Werner et al.)
Danksharding & PBS design docs (ethresear.ch)

10. Final project — pick one

Capstone (choose one, ship publicly)

A lending market on L2 with a liquidation keeper bot.
A zk-airdrop with private claim eligibility (Phase 13 extended).
A cross-chain NFT marketplace with LayerZero/CCIP.
An indexer + analytics site for a specific protocol, on Dune or your own stack.
An ERC-4337 smart wallet with session keys for games.

Write a teardown post. Tweet it. Put it on your CV.

11. Parting principle

Web3 rewards people who can hold two ideas at once: (a) software engineering is the same everywhere — correctness, observability, tests — and (b) the adversary is real and well-funded. Most fixes come from (a). Most losses come from forgetting (b).