LLMs Play Zachtronics' TIS-100
What happened
Sixteen large language models were given the programming game TIS-100 and its manual, then asked to produce and improve solutions for each puzzle. In a single pass per puzzle they discovered 133 new frontier points across 18 of the 55 puzzles and set three outright cycle records, while some models solved all puzzles and matched community performance on many fronts.
Why it matters
The results demonstrate that LLMs can autonomously generate and refine low‑level code in a constrained environment, suggesting they could serve as useful benchmarks for programming‑oriented AI research.
Is this a big deal?
Get stories like this every morning
The day's AI news in plain English, in your inbox at 8:30 IST. Free.
More AI news
Every Way to Get 192GB for Local AI (Without a $9,500 Mac) [video]Industry news · 14 Nov 2026DeepSeek V4 Pro 0813DeepSeek · 14 Nov 2026DeepSeek V4 Pro 0813DeepSeek · 14 Nov 2026Show HN: Portal2Bench – Can AI agents solve Portal 2?Industry news · 14 Nov 2026Anthropic Upgrades Claude Code Projects to Run Parallel AI Workers Hands-FreeAnthropic · 14 Nov 2026This page was collected from tis-100.wincon.dev and summarised automatically by AIPulsee's pipeline. See how it runs.