Humans Still Beat AI in the Long Horizon: Revisiting Test-Time Scaling in the Agent Era

Agents can spend test-time compute by trying, observing, and revising. We derive an Elo reference for repeated sampling, then show that in a 2022 two-week coding marathon, current agents plateau within 24 hours while top humans keep improving.

11 min read   ·   June 16, 2026   ·   test-time-scaling   LLM-agents   scaling-laws     ·   research  

We Scored 100% on AI Benchmarks Without Solving a Single Problem

AI benchmarks decide which models get funded, deployed, and trusted. We hacked 13 of them. 45 working exploits. Every benchmark rated critical. If the scores are fake, so is everything built on them — including your training data.

11 min read   ·   April 2, 2026   ·   benchmark   evaluation   reward-hacking   AI safety   trustworthy     ·   research  

Argus: Automated Discovery of Test Oracles for Database Management Systems Using LLMs

We present Argus, a novel framework that uses LLMs to automatically discover test oracles for DBMS testing — finding 41 previously unknown bugs across 5 widely-used databases. Accepted at SIGMOD 2026.

18 min read   ·   February 23, 2026   ·   database   testing   LLM   SQL   Auto Discovery     ·   research