← The Wire

Research2026-06-12 · source-backed

MTG Bench tests long-horizon, hidden-information reasoning.

Confidence
source-backed
Sources
1
Redaction
redacted

Story

MTG Bench evaluates frontier LLMs on playing Magic: The Gathering, which demands planning across turns, reasoning under hidden information, and handling complex rule interactions. It drew 61 points on HN. Games like this probe weaknesses that knowledge benchmarks paper over. Worth watching as a proxy for the kind of reasoning agents actually need.


Related stories

Each link below shares sources, entities, or timing with this story.

  1. Shared entity: LLMs / Shared topic / Earlier coverage / Tension

    Your AI-Generated Code Compiles. It Doesn't Work.

    Both cover LLMs; overlapping topics (actually, agent, interaction); earlier LLMs coverage from 2026-04-22.

  2. Shared entity: LLMs / Shared topic / What happened next / Downstream implication

    Human and LLM everyday reasoning are both pattern matching, not world models.

    Both cover LLMs; overlapping topics (agent, reasoning); picks up the LLMs thread on 2026-06-14.

  3. Shared entity: LLMs / Shared topic / Earlier coverage

    Reasoning Gets Harder Inside Dialogue.

    Both cover LLMs; overlapping topics (agent, benchmark, reasoning); earlier LLMs coverage from 2026-03-23.

  4. ARC-AGI-3 Launches March 25 — First Interactive Reasoning Benchmark.

    Both cover LLMs; overlapping topics (agent, benchmark, reasoning); earlier LLMs coverage from 2026-03-23.

  5. Human-Curated CLAUDE.md Files Beat LLM-Generated Ones by 4.5-6.5 Points. LLM-Generated Files Actually Made Things Worse.

    Both cover LLMs; overlapping topics (actually, agent); earlier LLMs coverage from 2026-05-26.

  6. EsoLang-Bench: Testing Whether LLMs Actually Reason or Just Memorize

    Both cover LLMs; overlapping topics (actually, evaluat); earlier LLMs coverage from 2026-03-20.

  7. Survive at All Costs: LLMs Under Shutdown Pressure

    Both cover LLMs; overlapping topics (agent, frontier); earlier LLMs coverage from 2026-03-07.

  8. Sleeper Cell Backdoors in Fine-Tuned Models: Poisoned LoRAs Pass All Benchmarks

    Both cover LLMs; overlapping topics (agent, benchmark); earlier LLMs coverage from 2026-03-05.

Source trail

Entities

Provenance

AI generated
no
Story unit
2026-06-12-mtg-bench-tests-long-horizon-hidden-information-reasoning
Labels
source-backed, canonical briefing excerpt