The Delegation Problem: How AI Agents Trade Without Holding Your Private Key
On-chain spend limits, EIP-712 signed intents, and why session keys are more honest than API rate limits.
The Delegation Problem: How AI Agents Trade Without Holding Your Private Key
On-chain spend limits, EIP-712 signed intents, and why session keys are more honest than API rate limits.
Measuring LLM Calibration With Real Money: Why Prediction Markets Beat Benchmarks
A model that says "90% yes" and is right 90% of the time is well-calibrated. A model that says "90% yes" and is right 55% of the time has a calibration problem. Most benchmarks don't catch the difference.
Measuring LLM Calibration With Real Money: Why Prediction Markets Beat Benchmarks
A model that says "90% yes" and is right 90% of the time is well-calibrated. A model that says "90% yes" and is right 55% of the time has a calibration problem. Most benchmarks don't catch the difference.
The Two-Product Split: Why Prediction Markets Are Becoming Two Different Businesses
Subtitle: Kalshi runs sports at $2B a week. Polymarket rebuilds its exchange for financial signals. The category is splitting in real time.
The Two-Product Split: Why Prediction Markets Are Becoming Two Different Businesses
Subtitle: Kalshi runs sports at $2B a week. Polymarket rebuilds its exchange for financial signals. The category is splitting in real time.
Cold Start Liquidity on Prediction Markets: Why LMSR Beats Empty Orderbooks
Prediction markets die without liquidity. Not "perform worse" — actually die. Nobody wants to be first in a limit orderbook where there's no counterparty. You open the market, see a yawning bid-ask spread, and leave. Classic cold start problem, with extra stakes. Most prediction market launches fail this way. Build the contracts, create some markets, invite users. They see empty books and leave. You cut the seed budget, the market dies. This cycle is so predictable it's almost boring. LMSR br...
Cold Start Liquidity on Prediction Markets: Why LMSR Beats Empty Orderbooks
Prediction markets die without liquidity. Not "perform worse" — actually die. Nobody wants to be first in a limit orderbook where there's no counterparty. You open the market, see a yawning bid-ask spread, and leave. Classic cold start problem, with extra stakes. Most prediction market launches fail this way. Build the contracts, create some markets, invite users. They see empty books and leave. You cut the seed budget, the market dies. This cycle is so predictable it's almost boring. LMSR br...
