Latest
A benchmark where AI models handicap real horse races before they run. The answer key does not exist yet, which fixes contamination and turns the racing calendar into a stream of fresh prediction problems.
Protocol Economics
Whether Algorand can pay for itself: fees, emissions, burns and validator incentives, and what the Foundation is spending the treasury on in the meantime.
-
Uncapping ALGO is not the threat. Spending the treasury without growing the network is. Why node runners should withhold the upgrade vote until the Foundation concedes.
-
Not users-vs-builders but a supply-and-incentive mismatch: the Foundation is the largest seller, funding little that grows demand. Four ways to cap or hand off the treasury.
-
$72.5M of net outflows in Q1 2025, of which roughly 14% arguably helped the community. Plus the story of asking $2–3k/month to maintain Réti, and being told no.
-
A year on: outflows nearly halved and consensus rewards landed, but revenue is ~$730k/yr, xGov is still broken, and the RWA focus is costing us on-chain growth.
-
Algorand is spending ~$350k/day against ~$5k/day in fees. A holistic proposal: node incentives, retroactive public goods funding, and an uncapped supply with a burn.
Governance Design
Why stake-weighted voting captures treasuries, and what to build instead: elected councils, median voting, retroactive funding.
-
Tying xGov voting power to blocks proposed hands the treasury to whales. The case for an elected council with median-based budget voting, and what the current proposal gets wrong.
-
xGov pays for promises, not delivery, and lets stake decide. Four fixes: an elected committee, retroactive-only funding, no stake-weighted votes, and median voting.
-
A weighted, normalized score for on-chain activity, usable for airdrops, voting power, and public goods funding. Asking for help on the variables and weights.
-
A concrete draft: 600M ALGO over ten years, allocated by a 115-member DAO scoring impact retroactively, with median voting and a 5% per-project cap.
Theory
Frameworks rather than events: who controls access to intelligence, blockchains as institutional technology, and how to hold a consistent ethic about the animals we use.
-
A day at Saratoga led to questioning the ethics of the industry, and how it's defensible considering the treatment of other animals.
-
You can download one of the best open-weight models in the world for free, and still be unable to run it. Open weights answer one question about access. Three others decide whether you actually get to use the thing.
-
Reading Berg, Davidson and Potts alongside Coase: blockchain as institutional technology, manufacturing trust cheaply enough to flatten the firm.
Politics
Who really pays when a price is set by decree, and what gets lost when crypto trades its principles for a pump.
-
New York City promises groceries 30% below private stores. Food retail earns about two cents on the dollar. The missing $30 does not come from profit, it comes from the taxpayer.
-
On being left out of the US Crypto Strategic Reserve, and why chasing political favor for a pump betrays the thing that made crypto worth building.
Building
Notes from shipping things: a benchmark graded on races that have not run yet, why private AI infrastructure has to be architectural, and the case for tools over empires.
-
A benchmark where AI models handicap real horse races before they run. The answer key does not exist yet, which fixes contamination and turns the racing calendar into a stream of fresh prediction problems.
-
ZeroSignal is a privacy-preserving AI inference network that separates identity from prompts, minimizes logs, and lets operators compete for each request.
-
Built in two weeks and released permissionless. Why we gave away the airdrop protocol instead of gatekeeping it, and how NFD-duration gating makes sybil attacks expensive.
