Article does not mention the other reason: in the interview with Dwarkesh, Amodei remarked about how other organizations are copying or training off Opus for their models.
By delaying allowing others to train off Mythos, they hold their SWE-Bench Pro head start longer so among other things, the USG can't but notice Anthropic's lead when they're deliberating on whether to further substantiate the "supply chain risk".
Precise motives are hard to work out as a general rule. Ultimately, it often comes down to a decision that decision makers like or don't like for a confluence of reasons.
This is what I've been doing, it works great. I've tested up to 4GB video files transcoded in the browser, then uploading the resulting video data. Also extracts thumbnail images, then uploads them to the server. Then I just do a quick verification server-side to check that it is actually a video or jpeg file, but the user's computer does all the work, so there is essentially zero cost for the whole ffmpeg transcode operation. It's brilliant.
Since it kind of had an LLM flavor in the original anyway, I had an LLM redraft this for my particular non-Japanese intercultural situation, and it wasn't valueless. I had to tell it to "Make sure to keep the same technology focus of the original."
"We're not perfectly good at preventing some of these other [model] companies from using our models internally." — well maybe this says something about how Opus 4.5 and Opus 4.6 have the same SWE bench score.
Does this manufacturer's practice pattern of repackaging data center components (e.g. Mellanox) imply any up and coming product creation opportunities?
By delaying allowing others to train off Mythos, they hold their SWE-Bench Pro head start longer so among other things, the USG can't but notice Anthropic's lead when they're deliberating on whether to further substantiate the "supply chain risk".
reply