Rosa Del Mar

Daily Brief

Issue 89 2026-03-30

Attribution-As-Moat And On-Chain Data Pipeline

Issue 89 Edition 2026-03-30 9 min read
General
Sources: 1 • Confidence: Medium • Updated: 2026-03-31 04:43

Key takeaways

  • Nansen has a data platform with more than 500 million labeled blockchain addresses and real-time on-chain flow analytics.
  • Nansen decided to go vertical on investors/traders rather than remain a horizontal analytics provider, aiming to become a full-stack on-chain agentic trading product.
  • Nansen believes agent behaviors will diverge across users due to differing intents, risk tolerances, portfolios, data access, tools, model choices, and LLM stochasticity.
  • Nansen describes a 'trust ladder' for agentic trading that starts with user-confirmed trading and progresses toward auto-execution and then strategy-level delegation once reliability is proven.
  • Alex Svanevik calls for broader participation in open-source AI experimentation and says there is an 'upswing' in open-source AI activity that he compares to early DeFi.

Sections

Attribution-As-Moat And On-Chain Data Pipeline

  • Nansen has a data platform with more than 500 million labeled blockchain addresses and real-time on-chain flow analytics.
  • Nansen outsources much node provisioning to providers such as QuickNode and Alchemy while indexing data in-house into ClickHouse.
  • Alex Svanevik argues that raw on-chain data is commoditized but that an enriched attribution layer is not, and he claims other analytics providers demand Nansen's labels.
  • Nansen describes its on-chain analytics stack as indexing raw chain data into ClickHouse, harmonizing schemas across chains/DEXs, and enriching the result with an address-attribution labeling layer.
  • Nansen states it labels identities only from public-domain information and will remove labels upon request for individuals (but not for projects/corporations), while noting some linkages (e.g., ENS) are immutable on-chain.
  • Nansen labels deterministic entities (e.g., Uniswap pools) via on-chain events and infers off-chain-controlled entities (e.g., Binance wallets) via behavioral flow analysis and hybrid heuristics.

Vertical Integration: Research-To-Execution Agentic Trading Surface

  • Nansen decided to go vertical on investors/traders rather than remain a horizontal analytics provider, aiming to become a full-stack on-chain agentic trading product.
  • Nansen relies on partners for trading infrastructure including Privy for embedded self-custodial wallets and DEX aggregators including LiFi, OKX DEX API, and Jupiter (on Solana) for swap routing.
  • Alex Svanevik claims the Nansen agent can answer a wide range of on-chain questions using near-real-time data from seconds ago.
  • In Nansen's current mobile app flow, each trade requires explicit user confirmation via a modal and a biometric MFA step.
  • Nansen's mobile app (Nansen AI) allows users to chat with an agent for on-chain analysis and execute token purchases within the same product.
  • Nansen aims to build trust by disclosing in the UI which aggregator provides the trade quote and surfacing routing information.

Strategy Abstraction And Simulation/Backtesting As Product Primitive

  • Nansen believes agent behaviors will diverge across users due to differing intents, risk tolerances, portfolios, data access, tools, model choices, and LLM stochasticity.
  • Nansen is pursuing a 'co-creation' workflow where users describe a strategy and the agent provides performance feedback by testing whether it would have worked over past periods.
  • Nansen Gym trains trading agents by replaying historical on-chain data as a fast-forwarded 'time travel' environment in which the agent is unaware it is seeing a replay.
  • Nansen is working on 'Smart Money 2.0' intended to predict which addresses are likely to make money in the future, and Alex Svanevik expects release in weeks rather than months.
  • Alex Svanevik expects users to shift from micromanaging individual trades to specifying higher-level strategies that agents execute under conditions.
  • Alex Svanevik expects retail investors to gain hedge-fund-like strategy-testing capabilities through AI agents, potentially leveling the investing playing field.

Trust And Safety Gating For Automation

  • Nansen describes a 'trust ladder' for agentic trading that starts with user-confirmed trading and progresses toward auto-execution and then strategy-level delegation once reliability is proven.
  • In Nansen's current mobile app flow, each trade requires explicit user confirmation via a modal and a biometric MFA step.
  • Nansen aims to build trust by disclosing in the UI which aggregator provides the trade quote and surfacing routing information.
  • Alex Svanevik claims a vanilla LLM fails for on-chain use because it lacks real-time blockchain data, cannot reliably execute/sign transactions without tool integrations, and often misjudges significance without domain tuning.

Developer Surface And Open-Source Ecosystem Positioning

  • Alex Svanevik calls for broader participation in open-source AI experimentation and says there is an 'upswing' in open-source AI activity that he compares to early DeFi.
  • Nansen directs users to nansen.ai and states the CLI can be installed via 'npm install nansen-cli' and that the team is on X at 'nansen_ai'.
  • Nansen's command-line interface (CLI) is described as open source and accepting contributions via GitHub.
  • Nansen plans improvements including more chain support, better liquidity via additional aggregators, and 'time travel' capabilities with feature parity to the core product for agent stacks.

Watchlist

  • Alex Svanevik calls for broader participation in open-source AI experimentation and says there is an 'upswing' in open-source AI activity that he compares to early DeFi.
  • Nansen is developing 'Nansen 3' aimed at mass-market consumers and Alex Svanevik claims it will be the best product Nansen has built with a futuristic but simple user experience.

Unknowns

  • What is the measured accuracy of Nansen's address/entity attribution (e.g., precision/recall) and how often do labels change or get corrected?
  • What is the real end-to-end latency from on-chain events to Nansen's indexed/served analytics and agent responses across supported chains?
  • How does execution quality compare against alternatives (slippage, price impact, failed transactions, and routing reliability) given reliance on multiple aggregators and an embedded wallet provider?
  • What are the actual monetization mechanics (fee schedules, staking revenue flows, and contribution margins) for the $49 subscription and Solana-free offering?
  • Does 'Smart Money 2.0' ship on the stated timeline, and what is its out-of-sample predictive performance versus simple baselines?

Investor overlay

Read-throughs

  • If attribution quality is durable, Nansen could sustain pricing power in on-chain analytics and use labeled-entity insights to differentiate agent-assisted trading workflows.
  • Vertical integration from research to execution could shift revenue mix toward higher-frequency engagement, but only if execution quality and safety gating earn user trust.
  • A data plus simulation plus execution loop could create a developer ecosystem moat if the open-source CLI and time travel capabilities become reliable primitives others build on.

What would confirm

  • Disclosed and improving attribution metrics such as precision and recall, with transparent label correction rates and QA evidence standards that remain stable at scale.
  • Published end-to-end latency figures from on-chain events to served analytics and agent responses across supported chains, consistently low and reliable in real usage.
  • Execution quality reporting versus alternatives including slippage, failed transactions, routing reliability, plus clear monetization mechanics for subscriptions and any fee flows.

What would kill

  • Attribution errors remain frequent or unquantified, labels change unpredictably, or QA processes fail to prevent high-impact mislabeling that undermines trust.
  • Material latency, outages, or degraded coverage across chains make real-time flows unreliable, reducing the usefulness of agent-assisted decisioning and time travel backtesting.
  • Execution stack underperforms with high slippage or failure rates, unclear economics, or delayed delivery and weak out-of-sample results for Smart Money 2.0 and strategy features.

Sources

  1. 2026-03-30 podcasters.spotify.com