[Question] Why didn’t Agoric Computing become popular?

I re­mem­ber be­ing quite ex­cited when I first read about Agoric Com­put­ing. From the au­thors’ web­site:

Like all sys­tems in­volv­ing goals, re­sources, and ac­tions, com­pu­ta­tion can be viewed in eco­nomic terms. This pa­per ex­am­ines mar­kets as a model for com­pu­ta­tion and pro­poses a frame­work—agoric sys­tems—for ap­ply­ing the power of mar­ket mechanisms to the soft­ware do­main. It then ex­plores the con­se­quences of this model and out­lines ini­tial mar­ket strate­gies.

Un­til to­day when Robin Han­son’s blog post re­minded me, I had for­got­ten that one of the au­thors of Agoric Com­put­ing is Eric Drexler, who also au­thored Com­pre­hen­sive AI Ser­vices as Gen­eral In­tel­li­gence, which has stirred a lot of re­cent dis­cus­sions in the AI safety com­mu­nity. (One rea­son for my ex­cite­ment was that I was go­ing through a mar­ket-max­i­mal­ist phase, due to in­fluences from Ver­nor Vinge’s an­ar­cho-cap­tal­ism, Tim May’s crypto-an­ar­chy, as well as a teacher who was a liber­tar­ian and a big fan of the Aus­trian school of eco­nomics.)

Here’s a con­crete way that Agoric Com­put­ing might work:

For con­crete­ness, let us briefly con­sider one pos­si­ble form of mar­ket-based sys­tem. In this sys­tem, ma­chine re­sources-stor­age space, pro­ces­sor time, and so forth-have own­ers, and the own­ers charge other ob­jects for use of these re­sources. Ob­jects, in turn, pass these costs on to the ob­jects they serve, or to an ob­ject rep­re­sent­ing the ex­ter­nal user; they may add roy­alty charges, and thus earn a profit. The ul­ti­mate user thus pays for all the costs di­rectly or in­di­rectly in­curred. If the ul­ti­mate user also owns the ma­chine re­sources (and any ob­jects charg­ing roy­alties), then cur­rency sim­ply cir­cu­lates in­side the sys­tem, in­cur­ring com­pu­ta­tional over­head and (one hopes) pro­vid­ing in­for­ma­tion that helps co­or­di­nate com­pu­ta­tional ac­tivi­ties.

When later it ap­peared as if Agoric Com­put­ing wasn’t go­ing to take over the world, I tried to figure out why, and even­tu­ally set­tled upon the an­swer that mar­kets of­ten don’t al­ign in­cen­tives cor­rectly for max­i­mum com­put­ing effi­ciency. For ex­am­ple, con­sider an ob­ject whose pur­pose is to hold onto some valuable data in the form of a lookup table and perform lookup ser­vices. For effi­ciency you might have only one copy of this ob­ject in a sys­tem, but that makes it a mo­nop­o­list, so if the ob­ject is profit max­i­miz­ing (e.g., run­ning some al­gorithm that au­to­mat­i­cally ad­justs prices so as to max­i­mize prof­its) then it would end up charg­ing an in­effi­ciently high price. Ob­jects that might use its ser­vices are in­cen­tivized to try to do with­out the data, or to main­tain an in­ter­nal cache of past data re­trieved, even if that’s bad for effi­ciency.

Sup­pose this sys­tem some­how came into ex­is­tence any­way. A pro­gram­mer would likely no­tice that it would be bet­ter if the lookup table and its callers were merged into one eco­nomic agent which would elimi­nate the in­effi­cien­cies de­scribed above, but then that agent would it­self still be a mo­nop­o­list (un­less you in­effi­ciently main­tained mul­ti­ple copies of it) so then they’d want to merge that agent with its callers, and so on.

My cu­ri­os­ity stopped at that point and I went on to other in­ter­ests, but now I won­der if that is ac­tu­ally a cor­rect un­der­stand­ing of why Agoric Com­put­ing didn’t be­come pop­u­lar. Does any­one have any in­sights to offer on this topic?