Mythos and DeFi
If you’re like me and you ingest media compulsively it feels like we’re standing at a tipping point with release of Mythos. This is either the greatest marketing strategy ever (“so powerful we can’t let you use it”) or we’ve made another leap on our acceleration to AGI (see my other article). My guess is with the release of GTP 5.5 with comparable Terminal-Bench 2.0 results at 82.7%, it’s all hype in the sense that no, this is not the model that immediately self-accelerates to destroy us all, but that yes, every 5-7% improvement in Terminal -Bench / Expert-SWE performance affirms the utility of these models.
What I’m amused by is how the first thing the media jumps to is zero days in our critical software infrastructure that is open source (the linux kernel, etc), and they’re right, that is scary. Because if all it takes is a security researcher saying “find me a zero day plz”, is 100 days really enough to patch everything? I truly doubt it, because the first thing red team will do when the model hits open access is ask:
“For each of the next 10000k non-famous, but still critical open source projects, find me zero days please”.
And if the Mythos claims are real, that WILL work with certainty. And if it doesn’t on the first pass, we’ll just keep doing it again and again and again telling the model to be more exhaustive. Because this is 100 days of wall clock time but Anthropic is obviously limiting the compute over those 100 days due to costs and resource constraints, and the red team will have infinite time after the model release.
As little as I care about using Defi myself, I’ll be watching the markets after this release. Think about it - what other markets immediately tie economic value to Mythos token. If Mythos tokens really are this magic vulnerability finding machine, then we’ll immediately see those tokens be used to extract economic value from the anonymous internet UNTIL the token values surpass the economic value to be extracted OR the TVL locked in these Defi protocols and contracts decreases. As I’m writing this the total Defi market cap is 77.21 BUSD - so if those protocols start getting breached due to the new omni-model, we’ll see the effects immediately, but I’m curious how quickly that will actually happen. As the low hanging vulnerability fruit is picked, how much token kerosene are we going to burn searching for new vulnerabilities that translate immediately to money?