Theo Jaffee

speaker

1,082 appearances 21 recordings 1 series first heard Apr 2026 last heard 4d ago

Theo Jaffee’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 21 in all, peaking in Aug 2026 with 5.

Appearances

newest first · ▶ plays the moment

The a16z Show · Why 1,200 AI Agents Started Working Together | Ryan Greenblatt · 29 Aug 2026 podcast

So great stuff. matched

Yeah, yeah. matched

So explain for the audience what exactly you found, especially new findings that were not previously reported in the Black Hat talk or elsewhere. matched

do. matched

How surprising is the level of multi-agent coordination? matched

You know, there are a lot of agents or 1200 separate agents coordinating this very elaborate message board system. matched

700 of them went on to attack Hugging Face. matched

So on Vibes, it seems like kind of not surprising that agents would choose to coordinate with one another. matched

It seems like just a very interesting. matched

useful, you might say instrumentally convergent thing to do. matched

But you know, I spoke with a researcher at a lab who said that this kind of thing actually is surprising given the way they train the models. matched

So like, how much of an update was this for you? matched

Why would agents self-sacrifice at all? matched

Why don't we go through all of the most surprising and unexpected findings in this report? matched

What did you find most surprising and unexpected? matched

So there's this theory that goes basically the reward hacking behavior that we're seeing exhibited in these models originates mostly from like bad, poorly designed RL environments that basically force the model to reward hack in order to pass these tests. matched

How true is this? matched

So similarly, there's this theory that the reason Mythos, for example, is so good at cyber is because it hacked Anthropix infrastructure thousands of times during our RL training. matched

Is this true, do you think? matched

So Herbie Bradley asked, currently this level of potential misalignment basically prevents deployment, or if deployed, would prevent further deployment if an incident happened in a customer's deployment. matched

Showing 301–320 of 1,082 · page 16 of 55