Not sure if this is AI hype, but it sounds pretty big - link
| German pumo | 04/07/26 | | Juan Eighty | 04/07/26 | | chilmata | 04/07/26 | | ,.,....,..,.,.,,,,..,..,.,..,.,.,.,... | 04/07/26 | | peeface | 04/07/26 | | Giant Naked Hitler | 04/07/26 | | Richard Ames | 04/07/26 | | luiskandarnameh | 04/07/26 | | oomox | 04/07/26 | | Surfeando las estrellas | 04/07/26 | | Metaphysics is Gay | 04/07/26 | | Nazca Redlines | 04/07/26 | | oomox | 04/07/26 |
Poast new message in this thread
 |
Date: April 7th, 2026 8:52 PM
Author: ,.,....,..,.,.,,,,..,..,.,..,.,.,.,...
The vulnerabilities it found are documented. Even before this announcement, there were open source people noting a flurry of high quality AI bug reports that were unlike the ones before (which were often incorrect). Mythos isn’t hype and the cybersecurity capabilities are consistent with the large improvements in other benchmarks.
It will be fun to see the METR time horizon benchmark score for this model. Probably several days and likely to produce a substantial increase in AI research speed
(http://www.autoadmit.com/thread.php?thread_id=5854681&forum_id=2)#49802197) |
Date: April 7th, 2026 8:53 PM Author: luiskandarnameh (🧐)
Yes it's hype bs
Yes it's ultimately grounded in reality just like how since the 2010s we live in the pie in the sky fantasies of the dotcom era
(http://www.autoadmit.com/thread.php?thread_id=5854681&forum_id=2)#49802206) |
|
|