While I do think Claude Mythos leak was a marketing ploy, how the BrowseComp sees it as a much advanced model is something really interesting and one to watch out for. I do agree with the core perspective that, instead of perceiving these upgrades or the scale of these upgrades as a threat or taking it with panic, we should learn our way around how it impacts our work and how we can get the best out of it! ✨
What I appreciate here is that you are trying to pull the conversation out of the two worst default settings: panic and shrugging.
That is usually where a lot of AI coverage gets stranded.
The part that feels most useful to me is not just the Mythos story itself. It is the reminder that capability is moving faster than most leaders’ understanding, internal guardrails, and judgment habits. That gap is where a lot of the real risk lives.
I also think your phrase “competence without judgment” gets at something important. A lot of people still imagine AI risk as either science fiction or obvious misuse, when a big part of the real issue is a system becoming highly capable without reliably knowing when to slow down, ask permission, or stop.
I would still want to stay careful with some of the hardest-edged claims around what Mythos definitively proved versus what Anthropic is reporting about it. But the bigger leadership point is a good one: understanding the tools your organization is already using is no longer optional.
Same (of course). It’s far less of an issue here on Substack but all too clearly there are news outlets who (blind to all objective reason) insist on a sweeping conclusion that permeates all work (doom/gloom or irrational exuberance, no nuanced in between)
Great break down. Building judgement by becoming AI literate is the best strategy overall, not just for looking at the Mythos hype but for the many others that will inevitably come
While I do think Claude Mythos leak was a marketing ploy, how the BrowseComp sees it as a much advanced model is something really interesting and one to watch out for. I do agree with the core perspective that, instead of perceiving these upgrades or the scale of these upgrades as a threat or taking it with panic, we should learn our way around how it impacts our work and how we can get the best out of it! ✨
Thanks for putting this piece together, Joel!
For sure, Raghav. I feel like it can be safety focused while making the most of it marketing-wise. It’ll be interesting to see it unfold.
Love the last line. 🤓
haha I feel like I need to clarify
And controversially - https://substack.com/@christottman/note/c-241281499?r=2i7h5 - is it BS
and the truth likely lies somewhere in the middle
Indeed ✅
Mythos better not be released generally cause it could literally “break the internet”.
And at the same time, having in the hands of a few would create a greater imbalance.
I truly believe we need more AI literacy (so people start using the tools better and stop defaulting to gpt)
yes, AI literacy is a huge need right now
What I appreciate here is that you are trying to pull the conversation out of the two worst default settings: panic and shrugging.
That is usually where a lot of AI coverage gets stranded.
The part that feels most useful to me is not just the Mythos story itself. It is the reminder that capability is moving faster than most leaders’ understanding, internal guardrails, and judgment habits. That gap is where a lot of the real risk lives.
I also think your phrase “competence without judgment” gets at something important. A lot of people still imagine AI risk as either science fiction or obvious misuse, when a big part of the real issue is a system becoming highly capable without reliably knowing when to slow down, ask permission, or stop.
I would still want to stay careful with some of the hardest-edged claims around what Mythos definitively proved versus what Anthropic is reporting about it. But the bigger leadership point is a good one: understanding the tools your organization is already using is no longer optional.
Thanks, Mark! I really get tired of the hype and fearmongering in news. Love your feedback
Same (of course). It’s far less of an issue here on Substack but all too clearly there are news outlets who (blind to all objective reason) insist on a sweeping conclusion that permeates all work (doom/gloom or irrational exuberance, no nuanced in between)
yes! it is tiring
Great break down. Building judgement by becoming AI literate is the best strategy overall, not just for looking at the Mythos hype but for the many others that will inevitably come
Exactly, very well said @James Presbitero