weights
- trend score
- 5.75
- sources
- 19
- mentions
- 35
35 mentions in these 90 days, up from 0 in the 90 before, across 19 tracked sources. The trend score compares the two; it measures coverage, not importance. How trends work
Flash Anthropic releases FlashProve, a prototype using cryptography to prove an AI response came from the claimed model without exposing its weights.
Daily mentions evolution
- Nov 4: Anthropic releases FlashProve, a prototype using cryptography to prove an AI response came from the claimed model without exposing its weights.
- Nov 3: MIT's GPT-6 Astra boosts small Qwen model math score from 7.9% to 48.1% by directly editing 0.016% of its weights without training.