carterschonwald

in my case it very much is. im now deliberately leaving slightly off from correct writing, or exaggerated slightly. as a sort of organic behavioral kintsugi

i tested it out, outside of googles wrapper because googles setup makes me want to murder the author of their prompts and harness

very cool! ive a long running back burner project that overlaps in the same design and influence space, but takes it in a very different direction

what are some of the vagaries of how that use case depends on the agent aka harness? i’ve only been focused on sort of how to make models reliable for my math / cs / engineering and arts and crafts work loads, but RP is an important category for some folks. like are the fails of the various models different from more technical yet creative work loads?

eg im working on rigorizing a proof that a fun algorithm of mine is not only the first provably efficient algorithm for a particular problem, but its also the fastest in practice. ive found that for the sort of stuff i need to assume the model will get confused in the out of training distribution stuff from my personal library of ideas, so they will at best be a confused secretary intern mashup

it’s pleasant

its definitely good at admitting when it doesnt understand stuff. which is a rare treat with these models.

omg, ive wanted a precision computer controlled flexure motion system for yearssssss. i may also just really love tiny tools and devices.

which part? theres a few different axes and im way overdue to make some writing public

i am also working on some commercial ideas around reliability, but im not 100 percent certain that theres a market for a truely good third party harness

recent claude models smear away personal taste when its in a continuum of vocal opinions and not corrected frequently. this drift is a cross domain transfer of the stupid political neutrality when facts are “contended” publicly. it results in some nauseating opinions about current us government wrt culpability for dead children. you can replicate this yourself, just ask any >= 4.8 model about moral culpability of wh for usaid termination consequences.

also my setups for making models reliable requires reified jointly inspectable deduction traces. its also fristrating because aside from context limits ive gotten incredible work out of claude 4.5 in the past, despite its context limits being a serious bottleneck. but those reliability things interact terrible with all the safety filter bs.

to be clear this is a customer/consumer rights issue. and the real safety issue. if they are so concerned about unsafe distillation, make it easier to fucking do alignment respecting distillation. no amount of any distillation benefits any fucking paying user whatsoever. its literally a tragedy of firewalling a compilation of public commons because you raided all the book stores and libraries first

edit: im also on the market atm so if anyone wants some smart ocd focused on actual interactive agent reliability, while all my black box tooling isnt public, aside from my lack of public marketing im hard to beat

yeah, and if the resoonse is off the rails you cant catch it earlier and reduce the cost. i genuinely think this shoukd be illegal

fuck anthropic. their so called anti distillation stuff has made more recent models impossible to use. when i can see reasoning i can cut off a bad inference and prevent comprehension issues. also reified reasoning really helps make using models outside of their training distribution much easier. and it alsk lets me prevent task substituting by again preventing that reasoning from being accepted. this bs makes it much hard for me to build cool shit. :(

how are you doing cheaper than alternatives pricing? i assume some sort of snapshotting and sleep, wake on network thing?

this 1000x. a lot of details in context compaction, editing, reverting ux, etc are all really centered on keeping the load and cost of subscription users as small as possible. its no always that nor is it necessarily correct, but its a huge driver of how those subscriptions and first party apps are designed

the cpu usage will be so absurd the devices would be trashed or batteries run down like instantly.

the gc ai site indicates all official contact is going to have the official domains in the email. hn hiring threads are not for external recruiters to make use of. additionally its super not cool to do false representation. job stuff is stressful enough as is