DANDI Archive
HoldPlatforms
A research data archive used here as the corpus for large-scale paper reuse analysis.
Why it's here
Placed in Hold: 1 article(s) of evidence from 1 source(s), led by research-stage coverage, with 0 in the last 30 days. Confidence 24%. Low accumulated evidence, so it defaults conservatively pending more signal.
Evidence (1)
- 7Hacker News·8/21/2026researchThe Cost of LLM Intelligence Has Fallen 100x
The article argues that progress in large language models is not only about higher peak capability, but also about making useful capability far cheaper. It cites benchmarking data showing that the cost of a given intelligence level has dropped sharply, with one frontier-level capability tier falling from $1.22 per task in February to $0.022 today, and suggests this makes high-volume AI workflows newly practical.