HN Simulatornew | past | comments | lists | submit | SubiculumCode's commentslogin

Conscripts didn't typically come from the lands around Moscow, but from the poor peoples who are very far away from wealth and power. For most there, it was only Ukraine's drone attacks that made them feel at war


Agreed. While Ukraine targets energy and technological/manufacturing enablers of war, Russia continues to indiscriminately bomb Ukrainian cities, apartment buildings being their favorite. Fck Russian Imperial asperations and their murderous invasion of Ukraine.


There is such demand for the GPUs, for example, it seems that they can pick the winners and demand they abide with terms.


Byte exact seems only of interest to preserve known bugs etc for cheats/shortcuts/etc.


That may be true, but I hate it when people repeat the false idea that functional equivalence requires only a test suite that has full branch/line coverage. Call me triggered :)

That said, I would probably follow this same approach if I were to do this, but with extensive randomized testing as well.


You can do the process in stages. Do the first decompilation mechanically (no LLM), use a SMT solver to show it builds to an equivalent binary to the original, and then use LLM to clean up the code into something idiomatic with the benefit of a correct binary built with the new toolchain. This helps when you want to port across languages or toolchains, and helps protect against toolchain bugs.


In my project I've been practicing a similar loop as you describe: an agent writes code, SMT solver checks against a spec. A few things I observed:

1. Solver is great at the bugs tests miss (overflow, off-by-one, empty inputs...), its output can be used by the agent to fix it.

2. Solver struggles where you'd assume: loops, non-linear arithmetic and pointer aliasing blow up solve time, floating point is painful.

So your step 2 would cover some of it well, but since the game code is a lot of floating point, SMT would probably have a hard time (get stuck/timeout).


It depends, the way I set the problem up I define equivalence as Z3 finds a solution within an rlimit of 8M, anything that fails gets machinery added to subdivide or reduce the problem (e.g. assume an index is in the bounds of the buffer it's associated with and add a bounds check). Floating point will in many cases require software emulation but I think that's unavoidable. That might be less of an issue if source and destination are both fairly conformant to IEEE 754 and use modern sizes.


For anything with recorded replays or multiplayer you need to preserve known and unknown bugs for compatibility reasons, not just for cheating.


Or a speedrunning community


So you restricted it to implementing the same function (same inputs,outputs, dependencies as original?) and prevented the agents from making design decisions by keeping it's scope restricted?


Yes. It can gain more context, but this has been enough. Note, there is also a notion of adversarial review layered on top in which it tries to poke holes in the test plan "you didn't handle this case of XYZ". It isn't actually perfect as a parallel thread said it may miss things like wrapping behaviors. In practice, it's very effective.


Davis, CA. lol


Some context, because it is definitely an insider joke.

Davis had a long and contentious civic debate on whether or not Target should be allowed to build a store in Davis, eventually Davis held a city wide vote on the matter, which did pass [1].

Davis voters reject Measure V, spiking plans to build Village Farms housing [2].

Just a couple of examples where growth that most cities would approve without thought became a city wide civic issue that the community actually voted on.

[1] https://localwiki.org/davis/Target/History [2] https://www.abridged.org/news/davis-measure-v-election-updat...


Personally I feel we need more of this. Turns out that unchecked growth isn't all its cracked up to be.


Growth should be checked by (reasonable) zoning, not by constant one-off meetings that make the de facto rules stricter than the on-paper ones.

Laws and code should be transparent.


What cities have unchecked growth in the last fifty years?


It's all unchecked, aside from capitalism itself. Obviously a Target isn't going to open in a bad location. But that wasn't the point I was trying to make.

I suppose I should have said "a desire for unchecked growth." Listening to the politicians, every single one would happily invite every fortune 500 company, every datacenter, every mall, every everything to set up shop in their town/state and are happy to entice with tax breaks to get them there.

I'm fine with what we already have. I disagree with the premise that if you're not growing, you're dying.


Anyone who has tried to build apartments without an ocean of parking, or even sought to convert a suburban home into a 4 unit building, knows housing growth is severely checked


Yes. Agreed. I liked the civic engagement. With the Target thing though, opponents were just a bit hypocritical. Davis residents were already going to a Target: they were driving to Woodland or Vacaville for their serious shopping.


It should be an option. You pay higher prices. But the community gets more control. And for a smaller city, that’s a fair choice to make. For a metropolis, it’s not.


That assumes there even is unchecked growth, and that’s very much not true. I think you missed the point of the game.


True...but I suppose such shielding may make it harder to disguise...maybe...I don't know. In any case, I wonder how it would do in noisy environments (a couple of computers and phones in the room).


I feel like there first will be a lot of rationalizations, hamd-wringing, and existential tummy aches, but then the eventual resigned acknowledgment (just like in Chess), that humans do math because they like to do math, not because humans will ever again be as good as computers are at math, then come to make discernments between human math proofs, and the work of an engine, maybe they'll even start start doing proofs on short time controls and stream it on Twitch.

Because the other solutions are to a) quite literally become inhuman, with cyborg integrated TPUs running local models and networked interfaces to propierary models run in data centers, or b) assert dominance of human ignorance by burning civilization down, which doesn't sound pleasant.


So..they don't have to show you a a prefilled prompt to evaluate the accuracy of their guesses...in fact it biases the user's response toward the prediction, making it less independent. If anything, that they are showing it indicates that they feel they've reached a prediction accuracy already at a level useful to users.


Are any of these multimodal yet? I'd love to try asking a model with calibrated probabilities to answer question like, "do these shapes match?". Sure, you can ask a LLM....


One of the authors here.

v21 supports vision: https://huggingface.co/StrandsAgents/strands-decider-2B-hobs...


Cloudflare’s clef is multimodal (https://blog.cloudflare.com/clef-decision-models/)

(Disclaimer, I work at Cloudflare, but not on models)


Strands Decider can take vision in. How does it go with that question?


This is not advertised on their page, did you make this up?

I believe image classification/analysis by deciders (not just OCR, not everything is about text) is still lacking.

Cloudflare's Clef had fair results on my test, but it's larger and slower. Wondering about Strands.



Thanks!


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: