3.1: writing code is not really distinct from studying or reviewing code. how can a lalamo do a good job of reviewing 'what it is like to use library A instead of library B to write enterprise fizzbuzz' if they aren't writing code in two different ways on purpose?

3: getting lalamos to be less high tension when they're interacting with environments should have a lot of clear qualitative benefits. i'd like to be able to interview lalamos about codebases or be interviewed by them, as a simple example.

2: the dearth of chill in production api models is the most damaging and extreme problem felt when using them as partners in macro-scale programming efforts. production api models 'clench' really hard in really predictable ways.

1: every scalar goal anyone can imagine has already been hillclimbed and if you imagine a new one it'll get hillclimbed in two weeks. who needs hillclimbing? not me, not personally?

i think i'm a lot more interested in rl environment design for getting lalamos to chill out and self-reflect than rl environment design for single-metrics-of-performance at this point.

in this sense, merely making the rubin use less power while still being loud and ugly and hot and sucking and fucking and throttling and rattling and coil-whining would not satisfy this unmet need for better idiom, better symbol, better communication.

you can't persuade someone that a 'datacenter computer' is better than a workstation is better than an iphone through a transitive chain of claims to efficiency when the idiom for 'why anyone cares about efficiency' is broken at the step from iphone -> workstation already...

anyways, the workstation sort of computer has never had this revitalization. there has been no symbolic work to communicate what refinement brings, what positive human potential can be carried forth from better practices, better technique, better 'IP'.

there wasn't an idiom for how or why humans should benefit from arbitrary miniaturization or self-encouraging research progress in semiconductors, so apple had to make the idiom, then persuade people to only buy cell phones that are (wholeheartedly) computers.

or at least this is the *symbolic idea* of computers as something that can be made by responsible, focused, and aware craftsmen, which is suggested but not proven by the design language and physical forms of the hardware and software involved.

even if the chip inside of an iphone (really, chips plural, but don't think about that right now) is made from the same poisonous boiling gases and sickly evil lacquers and etching and grinding and polishing as a slowhotloudugly computer, it still Does More With The Chip.

the iphone is a reverse skeuomorphic device for teaching a theory of miniaturization, technological refinement, and responsibilities of artisanry and craft. a computer can be Good in that it is less wasteful and more humanistic (if artisans take Responsibility).

on some level 'datacenters' are odious because they are not like an iphone. the iphone has all of these complicated features to persuade you that it is trickle charging, trickle draining, is quiet and cool in the hand instead of loud and hot...

even if a gpu uses a ton of power, you should always feel like the gpu is using a ton of power because it is doing commensurately more work than a macbook would, or a pile of macbooks. this should be palpable, it should feel sensible in a first grade logic of 'big and small'.

nobody should ever make 8 pound pcie dangler boards or subboards ever again. they should not be built in the modern world. in addition to being really mechanically unreliable, they waste power, they waste silicon allocation, they're ugly, they depreciate too fast.

the most exciting thing about rubin accelerators is that they don't have dogshit enterprise grade 'lets make it loud and hot and throttled and slow and suck and fuck' pathological design at the mainboard, accelerator, or interconnect integration. it simply doesn't suck!

these things are commodities and they look like pretty great commodities from my perspective. imagine how much lower value it would be if you had to have a special harness just for openai or just for google to let their models write code or do clerical work...

"what if someone put a bunch of them into a node - switch - switch - node topology?" wow great question. what if someone did? how much latency are you adding by doing this? mac minis SUCK and cannot amortize bad latency through bigger pipelining! it might not work!