Industry's Guide to the New URAR

[J Grant]

A question I have about the new form is this, will we be able to at least partially clone prior reports? It will be helpful for specific properties and for particular subdivisions. I've completed about a dozen 1004s over the years and a couple of 1073 condos and it would be nice not to have to begin with a blank slate. I'm going to make a point of going to a live CE course around the time these become mandatory just to hear the old goats wail and screech about the change.

Ask your software provider - I assume we can but I will ask mine as well at the appropriate time ( alamode)

 

[Tom D]

Ask your software provider - I assume we can but I will ask mine as well at the appropriate time ( alamode)

Good advice. But if I'm fannie, i make it impossible to have a template. Gotta make it harder to get rid of you long time appraisers. Not going to kill you all, just annoy yous into leaving the profession.

 

[RCA]

Are you talking about R retirement. That certainly works for me, but oh so boring. Back to my fix and flip days.

R language of course. I mean I have been doing this for quite some time. I generate the output to Excel and then upload it to Alamode. I guess, if I were to continue to do GSE reports, I would find the new report form easier to work with. Not sure though, I will take up that path again. I have NOT spent much time with the new form. I am totally occupied finishing my paper on the RCA approach and writing Python and R code. --- Just about finished ... I keep saying ... then I think of something else to add. Enough math, not too much. My Stanford Ph.D. Geostatistics son-in-law asked me to send him a copy to read - or at least have some graduate student in Paris review. - but these guys spend most of their time working with Tensor level Calculus, quantum mechanics level Kriging for the petro- and energy industry --- not really applicable to GSE real estate appraisals ---- but I have to worry whether its going to come back with Tensor math recommendations .... my stuff is bad enough. Just simple-complicated simple Algebra you will nonetheless find tediously detailed.

I'll first upload to Arxiv.org, my son-in-law can get me in.

 

[Tom D]

Whatever you just said. Wait a second, i don't understand what you just said. Gotta love the lender, underwriter and the civilian owner reading your paragraph with a blank look.
Anyway, i'm super fast with plain math. Can't understand algebra.
But, fannie avm will soon solve the issue, before you solve us understanding your space math & quantum physics, R man.

 

[RCA]

Whatever you just said. Wait a second, i don't understand what you just said. Gotta love the lender, underwriter and the civilian owner reading your paragraph with a blank look.
Anyway, i'm super fast with plain math. Can't understand algebra.
But, fannie avm will soon solve the issue, before you solve us understanding your space math & quantum physics, R man.

You have more confidence in AVMs that I do. But you ought to know there is a lot of hype in marketing products like AVMs or AI. Supposedly, listening to a ton of YouTube videos on AI, "Coding is Dead". But my experience with using LLMs for coding is that they only know pretty much what they already know, hoping that the requirements they are given match the previous requirements of some task, they will suggest the previous final version of what they came up with. Only human language is not that exact and people are not perfect at saying what they mean, so it is a rare occasion when ChatGPT, Grok2, or Claude come up with the right code in the first try, or the second, or the third, ..... Yea, it is trial and error. Someone has to be there to tell LLMs they do or don't have it right, and if not, what is wrong. And the person telling the LLM that might not get it absolutely right until somewhere down the road - and around and around and around we go. And even if the LLM can get the code to produce the right answer, you need to look at the code, read the code, and make sure it makes sense, because it might fail under different conditions.

So-called "Operators" and "Agents" have the same problem. They have limited usefulness working on their own. They mostly make the programmer far more productive. Yes, they are good for that. But people are still needed for the foreseeable future.

Same with appraisals of complex properties, especially at the low and high end of the quality range - as well as for other properties that are non-conforming or in complex markets. For now anyway.

Everything is advancing and moving forward, - not just for the sake of efficiency - but to keep up with an ever changing and evolving world that continually increases in complexity. The simple good old days are gone. It's not the 60's anymore.