# What You Build When the Data Doesn't Exist Yet *“Everyone has a plan until they get punched in the mouth.” — Mike Tyson* By [0x836c](https://paragraph.com/@0x836c370ab3ca72a823ec7f1a9c148305db3e5812) · 2026-05-26 defi, analytics, ai, mcp --- The first time I needed granular Liqwid pool data for an institutional pitch, I got nothing. Not lag. Not partial data. Nothing. The Liqwid contract ABIs weren’t in Dune yet — three years behind Ethereum’s data infrastructure timeline — and i was sitting there with $81M TVL growing organically, a pipeline of allocator conversations live, and no clean way to query the numbers that the people on the other end of those calls actually needed to see. That was the punch. WHAT YOU BUILD WHEN THE TOOLS DON’T EXIST YET --------------------------------------------- Most growth operators in DeFi assume the analytics infrastructure exists first and they slot into it. That holds at Aave. It held at Uniswap circa 2021. It didn’t hold at Liqwid in 2023, and it doesn’t hold in most non-Ethereum ecosystems today. You build your own observation layer or you’re presenting allocators confidence intervals you can’t actually back up. What started as a Python script and a Slack webhook — running every morning at 5:30 UTC because i needed to know what had shifted before i was caffeinated enough to find it myself — became defi-intel. I published it as an MCP server. It’s live on Smithery. WHAT IT ACTUALLY DOES --------------------- DefiLlama is brilliant. Everyone uses it. It has four core data tools. defi-intel has eleven, and the seven that aren’t in DefiLlama are the ones that matter most in an allocator conversation: **Governance proposals.** Live voting state for the protocols you’re tracking. Not the TVL — the actual governance posture. Who’s proposing what, at what quorum, and what happens to the protocol’s risk parameters if it passes. I got blindsided by a Liqwid parameter vote once that i should have caught two days earlier. That doesn’t happen if you’re running governance signal. **RWA attestation scores.** The defi-intel MCP integrates with rwa-attest directly — the same attestation engine i built to score BlackRock BUIDL, Superstate, and Maple on-chain. If you’re evaluating RWA collateral for a lending protocol or a fund allocation, you get verifiable scores, not vibes. **Live X/Twitter narrative.** Via Grok x\_search. Not sentiment analysis applied to day-old data. Live. The narrative velocity that precedes price action — in theory. In practice it's the signal that tells you whether the next week is noise or structure. I can't point to a Bloomberg story I beat just yet. But what i can tell you is that I stopped being surprised. **The one-call operator brief.** This is the output format that doesn’t exist anywhere else: TVL + yields + stablecoins + governance + market narrative compiled into a single NotebookLM-ready document on demand. Your fund analyst reads it in four minutes. Your founder reads it before the board call. It’s the DeFi equivalent of a Bloomberg terminal morning brief, without the $24,000/year subscription. THE PART I GOT WRONG -------------------- I thought the internal tool was the product. It’s not. The pattern is the product — the judgment about what to monitor, how to weight competing signals, and how to synthesise them into something a non-technical allocator can act on in a 15-minute meeting window. That judgment took two years of Cardano growth operations to calibrate. I published it as an MCP because the barrier to building your own version is real — Dune still doesn’t have every protocol, xAI’s x\_search requires a live API key the DeFi analytics world hasn’t caught up to yet, and the governance data infrastructure for anything outside Ethereum is still fragmented. defi-intel is a one-install solution for researchers, protocol contributors, and fund analysts who need operator-grade DeFi intelligence without building the stack themselves. It runs on no enterprise subscription. It costs the compute you’re already paying. Install it on Smithery: [smithery.ai/servers/defibabylon/defi-intel](https://smithery.ai/servers/defibabylon/defi-intel) Full tool list and self-host instructions at [github.com/defibabylon/defi-intel](https://github.com/defibabylon/defi-intel) --- *Originally published on [0x836c](https://paragraph.com/@0x836c370ab3ca72a823ec7f1a9c148305db3e5812/what-you-build-when-the-data-doesnt-exist-yet)*