Home loan lending

The key to automated lending is to get a reliable measure of end point rik, make it easy to ientify borowwers who consistently operate in amber.  Then you can auto trade home loans, making sure the loand are ARM with no prepayment fee. The business model assumes sensible borrowers who will refinance when they observe gain from doing so.

There is no attempted hidden arbitrage, except borrowers cheating on their credit score. So, nail that.  Then all risk priced, you have the sandbox, group closure happens, you can have the pit autotraded. The pit boss drives bit error to zero, the graph holds loans and deposits, kind of balanced. Bidders and askers have equal observability.

Let's try the logic here.
The borrower submits a probability distribution of personal liquidity events.  The amber reading indicates he probability makes a housing payment with 99%, great.  His distribution sits mildly across zero average cash liquidity, great news, this borrower knows he is operating in a zero  transaction cost  environment.  Hence, the pit can compress the lender bid and borrower ask, distribute bit error by contract. But, loans are always compressed, and the adjustable rater appears as a partial loan recall, mostly by month. The borrower, lender, and pit boss have slightly increased risk, currency risk, as long as the loans and deposits are allowed to get out of coherence.

And the pricing ring is enforced, the economy implements ratio algebra..