| Fable 5 – Median thinking declined in August(twitter.com) |
| 418 points by espeed 22 days ago | 291 comments |
| tl;dr: Summary not available. |
| HN Discussion:- ↑Speculation that AI companies deliberately degrade models to manufacture perceived improvement in new releases
- ↑AI companies should be regulated like other industries to prevent inconsistent product quality
- ↑Personal anecdotal confirmation of noticeable model degradation over weeks of use
- ~Reduced token usage or thinking isn't inherently bad if results remain comparable via optimization
- •Independent private benchmarking is needed but hard to sustain against lab gaming and monetization challenges
|
|