Hacker News · AI· Arno_Bighub·· 7 hours agoSelectedAI score64
First Silicon Hardware Benchmark Released: Frontier Models Detect Under 20% of Chip Flaws at Zero False Positives
Show HN: First Silicon, real chip bugs that AI models miss
AI guide
A research team has launched First Silicon, a hardware benchmark. Results show that top-performing Claude Opus 5.5 detected only 18.6% of real chip flaws under a strict zero-false-positive threshold, with two-thirds of its alerts being false alarms.
Why it matters
Evaluating frontier models across 86 real-world chip fixes, the study reveals critical shortcomings of LLMs in hardware formal verification, notably high false-positive rates and misjudged counterexamples.
Source:Hacker News · AI · umwelt.vision