MadnessASAP

The spec page says 100 kHz BW on the oscilloscope, the FAQ says 400 kHz. In either case calling it an "oscilloscope" is a stretch, its the ADC channels on an MCU.

I find it curious that all their promo shots seem to only show the back of the board. I couldnt find any of the component side, or any information about what components are used. My guess would be:

- a very small dual rail supply

- AVR or STM MCU

- Signal generator is PWM through an RC low pass filter

- Oscilloscope is potentially just the input through a resistor network to shift +/- 5V to 0-5V, maybe a buffer to keep input impedance high.

I just don't see $170-200 of value here, or anything close to that.

A company operating above board would be sure to carefully document the state of the rental before and after whatever work they were doing. Any tradesperson/installer/technician/repair person will have tales of how they were accused of stealing grandmas wedding ring from the bottom of the sock drawer while repairing a leak in the kitchen.

So either Bot Company damaged property and is trying to pretend they didn't. Or they are incompetent and failed to document the state of the property or handle the owners complaints appropriately.

Given that their training robots and would therefore be collecting as much data as possible, including camera data, I'm leaning towards malice instead of ignorance.

Of course its an incentive, however the disincentives to purchasing (subscribing/spending), and thus producing, such games still exist.

You are technically correct, the best kind of correct. However! That would be a terrible UX/UI experience. While showing distances on a linear scale is accurate, it fails to capture all the information a person in an interstellar ship may wish to see.

Something like logarithmic distances would better capture information like "Am I about to crash into the star or enter a nice orbit" while still showing the full picture of where you are in relation to where you're going and where you came from.

No idea of that's what happened here, just a thought, I'm not an expert in starship computer interface design.

For something like a transfer between Starships you can resolve a lot of those problems by (very) gently spinning the 2 craft. It won't take much force for the liquids to settle at the bottom of their respective tanks where you would presumably put the intakes.

> 7. Half the packages are maintained by one person, unpaid, at 2 a.m., after getting yelled at in GitHub issues.

By a manager for for a >$1 billion market cap corporation who doesnt understand that the one person isnt an employee.

Assuming you are talking about real physical dice and not an imaginary function that generates perfectly random die rolls.

They are actually pretty poor random number generators. For starters, dice are chaotic, not random, the outcome is based entirely on initial conditions. For humans rolling dice, the space of initial conditions can become surprisingly constrained, especially if the human wants to achieve specific outcomes.

It's not academic, it's a real practical reality.

Alice runs many services and has a rather large attack surface. I don't want Alice to persist those secrets, only to have them briefly at startup (think joining tokens). Bob however has exactly one job, verify that Alice-1 to Alice-N are in a trusted configuration before granting them access to the cluster.

Very recent events in the Linux kernel prove that it isn't safe to assume "0600 root:root" is sufficient to protect secrets from a misbehaving container.

Sadly, the problem isn't the TPM or Remote Attestation. It's Google et al choosing to only talk to devices and software they like without concern for what the user wants or trusts. Compounded by everyone else just going along with it.

A TPM where the device owner can't take ownership of the root key is worse then no TPM at all.

I have 2 servers, Alice and Bob, Bob has a secret, I want Bob to be able to share that secret with Alice. However, I want Alice to be able to prove to Bob that it is actually Alice, that it is running the correct AliceOS, and that AliceOS was loaded on bare metal Alice without nefarious pre-book or virtualization hooks.

A TPM with measured boot (SecureBoot) does exactly this, remote attestation is how Alice proves to Bob that it is in a trusted configuration and wasn't tampered with.

Apparently TDP is 30 watts¹, according to the product brief. I would imagine it's a single PCB with flash chips on both sides then thermally bonded to the aluminum chassis. That should keep all chips at approximately the same temperature. On its own it could be easily air cooled, but with 24 in a 2U chassis you'll be having some decently hefty forced air over the drives.

1. For comparison, an HDD usually comes in around ~10 watts

It mostly comes down to the consumer market not being significant enough by itself. A consumer may not notice a 10% increase in performance per watt or dollar. A large office building probably will, and a datacenter definitely will.

I don't think I'm being entirely hyperbolic when I say the consumer market only exists to put devices that can connect to and feed the datacenter loads into the general populations hands.

An Arc B580 will just about fit Flux.2 Klein (At FP8). However, you can also easily get much larger GPUs on RunPod or Vast at $0.25/hr.

I would strongly recommend exploring that option, renting an RTX 5090 for an evening of image generation for a dollar or two is way more fun then trying to jam big models on little cards. Just take some time to create a reasonable, scripted, deployment workflow for when you create a fresh instance.

And you're trying to rename unregulated gambling as something that is good for society.

Exploiting people with gambling addiction is not a reasonable replacement for insurance.

Yes, but the examples where it's good has a name "insurance". It exists, it's generally well regulated, and is not easily exploited.

The reason it works better is because in a prediction market, the person betting against you has no resources or ability to go after you for fraudulent behavior. Whereas an insurance company has both.

I would assume somewhere in both the companies there's a Ralph loop running with the prompt "Make AGI".

Kinda makes me think of the Infinite Improbability Drive.

They didn't exactly intimidate them down last time either. Piracy decisively won the war on piracy.

> They aren't suing some broke 23 year old. What they can collect is less than their lawyer fees.

You may not be old enough to remember this, but that's exactly what they did in the 2000's

All TOS essentially boil down to "we owe you nothing and can change the product at anytime to anything we want at our sole discretion"

Obviously it would be unreasonable to accept such terms without further context. The further context in this case being that Anthropic will maintain Claude as an AI agent and seek to improve it's performance. What is at the heart of this issue is whether or not Anthropics recent A/B testing violated that context. Not whether or not they violated the TOS (they didn't, obviously)

While not being particularly knowledgeable in such things, I would presume the shrews.

A whales muscle needs to optimise for efficiency and oxygen storage to allow for extended deep dives and continuous use. Compared to a shrew whose muscles would favour compact size, low mass, and fast reactions to permit quick getaways.

My interpretation is that they built a simple virtual machine directly into the weights, then compiled a WASM runtime for that machine, then compiled the solver to that runtime.

> Which I would find "cute" if the database contained an equal amount of reason. I am perennially irritated that "US/Pacific" which is an _official_ name of a time zone _as used_ by the relevant time keeping authority, is called "backwards."

This assumes that every point on earth has exactly 1 governing body and that a significant majority of the people agree on who that governing body is and that the governing body gives a rats ass about what time it is. Or that everyone in a region agrees on what time it is. Or that ccTLDs are sufficient to unambiguously cover the entire earths surface.

The time zone database isnt just a record of "official" decisions regarding time, it is a record of what time a population thinks it is. There are geographic overlaps, cultural overlaps, pants on head stupid overlaps. It exists to try and translate between somebody somewhere some when giving a time and date reference to any point in history to whatever time system the user may choose to believe in.

Your solution is insufficiently complex to solve a problem of this complexity.

https://gist.github.com/timvisee/fcda9bbdff88d45cc9061606b4b...

Its not transparent to me? What are they trying to do?

Yes! Nixpkgs straddles both worlds, like a system package manager it provides a way to install packages and their dependencies. However, like most language package managers it also imposes a locking mechanism so that every input into the nix expression* is locked to a hash, the mechanism is recursive and handles multiple versions of the same package in parallel.

The recent(ish) concept of "nix flakes" means there are 2 related but different mechanisms for achieving this but the end result is the same.

* In the land of NixOS everything is a nix expression including the system packages, configuration files, install image. It's all locked to hashes of upstream sources and is in theory, fully byte-identical reproducible.

You can, and it might make things a bit better. The only real way I've found so far is to start going through file by file, picking it apart.

I wouldn't be surprised if over half my prompts start with "Why ...?", usually followed by "Nope, ... instead”

Maybe the occasional "Fuck that you idiot, throw the whole thing out"

Of course they did, in what world would they not have? You can't get any of these companies to take a single person-minute to look at an issue that affects you. However no problem putting a small country of people to work invading, reviewing, and annotating the shit out of your privacy.

Ya know just today I was thinking around a way to compile a neural network down to assembly. Matching and replacing neural network structures with their closest machine code equivalent.

This is way cooler though! Instead of efficiently running a neural network on a CPU, I can inefficiently run my CPU on neural network! With the work being done to make more powerful GPUs and ASICs I bet in a few years I'll be able to run a 486 at 100MHz(!!) with power consumption just under a megawatt! The mind boggles at the sort of computations this will unlock!

Few more years and I'll even be able to realise the dream of self-hosting ChatGPT on my own neural network simulated CPU!

They are different from both the users and developers perspective. A developer can't prevent a user from entering invalid data, they can however check for invalid data and inform the user as such.

A user can't prevent a programmer from calling a function with a non-empty string, a developer however can ensure the arguments are valid before calling a function.

In both cases there is something the developer can do. Only in the 1st case is their something the user can do, but only if the program informs them of the problem.

BUT! And this but is something that really gets up my butt. If a programmer spits out a thousand non-fatal errors that I can't do anything about, I'm likely to miss the one error I can do something about. So that's also an important distinction between these 2 cases.

The threat model doesn't really change for agents that already have "web fetch" (or equivalent) enabled. The agent is free to communicate with untrusted websites[1]. As before, the firewall remains at what private information the agent is allowed to have.

[1] If anything the threat gets somewhat reduced by the ability to point directly at a trusted domain and say "use this site and it's (presumably) trusted tools."

It's genuinely amazing how good they are at reverse engineering.

I have a silly side project that ended up involving the decompilation of a toaster ovens firmware, the firmware of the programmer for said toaster ovens MCU, and the host side programming software. They were able to rip through them without a problem, didn't even have ghidra setup, they just made their own tools in python.