Skip to main content
Brian Powers

Brian Powers

Aug 19, 2026

TedHOST
14:31
And then quality is another metric that's tough to measure for the reasons that you mentioned.
14:36
Yeah, I think that for most law firm legal work, we're not yet at the point where AI is going to fundamentally change.
14:46
that relationship, I think the work is still very indeterminate, which makes it difficult to price in advance.
14:53
And that means that you're gonna have to do some variable costing.
14:59
It's also extremely difficult to evaluate quality and extremely difficult to evaluate turnaround time When that takes longer, was it good that it took longer because it really was a lot of work? Or was that because it was in the back shelf there or not a top priority? I think that a lot of times law firms aren't necessarily incentivized to disambiguate those things to clients.
15:26
Right.
15:28
And ultimately law firms, even if they cut costs internally for non-billable hours, they're going to be incentivized to charge as much as they can in the marketplace.

13 MINS LATER

TedHOST
28:24
It's still kind of TBD how, how that's going to work because, you know, yeah, there's still some things to sort out, but it does seem like there's a push now to move up the complexity ladder a little bit.
14:41
Or again, is there something just... distinct? If something is values-based, is there no amount of data in the world that will get these tools to the point that they can reliably process values? Does that make sense? How do you feel about that?
14:57
Yeah, I think it probably relates to context windows where AI is extremely good within a certain context window.
15:05
But when you think about values, this amorphous, almost feeling-based set of principles, in order to apply a value, you often have to draw on so many different parts of informational sources that you've developed throughout your memory and your history working.
15:21
And you're not even conscious of everything you're drawing on, but it's enormous.
15:25
And I think the reason why AI can break down and not feel reliable is often when the context it needs to draw from is just too limited.
15:35
And I think that's then the part that's been most surprising over the last couple of years, even though we've gotten so much smarter within a certain context window, it's still remarkably small.
15:45
I mean, people say million token context window.

5 MINS LATER

21:00
Am I understanding sort of the argument, the theory of the case for Marco correctly?

We value your privacy

We use cookies to understand how you use our platform and to improve your experience. Click “Accept All” to consent, or “Decline non-essential” to opt out of non-essential cookies. Read our Privacy Policy.