◀AI LabsA DeFoor Innovations publication · AdvertisePodcasts▶

@DEFOORREPORT ON X · BLUESKY · TIKTOK

BUILDERS & RESEARCH

THE PEOPLE BUILDING WITH IT | 11 STORIES | UPDATED OCT 5, 2026 12:44 PM ET

HakemBench: A Turkish Benchmark of Typed Decisions

arXiv: Computation and Language · 13 hr ago

Finding the Move Is Not Winning the Game: XiangqiBench for Closed-Loop Evaluation of LLM Agents

arXiv: Computation and Language · 13 hr ago

Counterexample Generation via Per-Theorem Symbolic Verifiers: When Imitation Hurts and Reinforcement Repairs

arXiv: Computation and Language · 13 hr ago

FinDialogLens: Event Extraction over Multi-Party Dialogue for Missed-Trade Identification in Financial Chatrooms

arXiv: Computation and Language · 13 hr ago

CUEing User Simulators: Calibrated User Embeddings for Multi-Turn Benchmarking

arXiv: Computation and Language · 13 hr ago

Qwen3.8 27B addition in words

Simon Willison · 17 hr ago

We're going to need default hard budget caps on pretty much everything

Simon Willison · Oct 3

September sponsors-only newsletter

Simon Willison · Oct 3

Rex's Dino Store

Simon Willison · Oct 2

Quoting Matthew Green

Simon Willison · Oct 1

He Built This City

Simon Willison · Sep 30