Tools for questions I actually have as a DeFi user, including the ones I killed.
● Refreshed automatically from DefiLlama
"Where can I park my stables right now, and is that yield real?"
A trust filter, not a yield hunter. It screens every stablecoin pool DefiLlama tracks against a hand-kept list of trusted protocols and stablecoins, then ranks by durable, organic yield. A high APY counts as a red flag as often as an opportunity.
"I want to park euros or dollars in real-world yield. Show me what's real, what's risky and what's fake."
Classifies tokenized funds by what actually backs them (government debt, private credit, basis trades and more), gives each a trust tier, and flags products whose label doesn't match what they hold. New products above $5M TVL surface automatically as "unclassified" until reviewed.
"What am I actually buying?"
Type a ticker and see a token's fundamentals in one place: valuation against its peers, traction, value accrual, dilution and unlocks, holder concentration, treasury, security and development activity. Each number is rated from very low to very high by fixed, visible rules. Free public data, runs in your browser, no AI.
When the answer is a finding, not a tool
669 meaningful token unlocks, measured against BTC and against each token's own normal trend. The extra drop an unlock causes comes out at +0.2%, which is indistinguishable from zero. "Sell before the unlock" doesn't survive a baseline control.
Read it →Prototyped on real data, then stopped
The idea: find crypto assets that diversify away from BTC. On a year of real data, the only things that did were stablecoins and gold. Every risk asset sat between 0.66 and 0.92 correlation.
The tool's only honest answer was "hold cash", which isn't worth building a tool for.
The idea: a calendar of upcoming unlocks with the expected price reaction. On 22,168 historical unlock events, the reaction turned out to be the token's normal drift, not the unlock. What was left was about as predictive as a coin flip.
No tool, but the finding became the research piece above.
How I build
Before any code, I question the idea hard: who decides what with it, and what it has to beat to be worth using.
A throwaway probe on live data with a go/no-go test. Two of the four ideas so far stopped here.
A scheduled job pulls DefiLlama, a pure pipeline applies the opinion, and a static site publishes it. Everything is open source.