idorosen

As a payments processor, Stripe has conversion data. They can route based on ROI with end-to-end A/B testing or online optimization. There's a tricky credit assignment / multi-touch attribution problem to solve for some use cases, too, that could benefit from transfer learning across their customers. Also, the (unstructured) data they collect from OpenRouter users can be used to improve many other dimensions of their business, including fraud detection. Lastly, both businesses look fairly similar in unit economics if you squint hard enough; for example, they both collect a ~3-5% transaction fee from most customers.

My side project this quarter has been building a system that adds steering / planning to LLMs/WAMs/etc. (T, DiT, etc.) in a market/physics/other-ground-truth-constraints aware way. I have a small model outperforming some of the largest frontier models in some games (i.e., adversarial settings) now, but that is not very impressive because "plain old LLMs" are the wrong mathematical object for gameplay, IMO. So far, it's a mix of RL + optimization (e.g., MILP) on the planning side, and not quite realtime enough to play RTSes, but that's only a few weeks away. Have been very impressed at how far the open source ecosystem for this stuff has gotten since the last time I tried something like this.

I want to buy one of these just to support projects like it.

Why invent conspiracy theories? Occam's razor likely applies here. They are leaving for companies that have filed for upcoming IPOs, so it could just as well be compensation.

This bill would likely cause game studios to release games exclusively via subscription plans as an unintended consequence.

The title does not match the article title: “We Upgraded to a Frontier Model and Our Costs Went Down”.

I would buy it. I have built a similar contraption. Let's connect.

Or to simplify the product and rebuild.

ai slop

Using the network byte ordering (big endian) of UBNT as the magic number in the protocol is a nice touch.

There is an old but still reasonable solution with mkinitcpio hooks encrypt/sd-encrypt + ssh, which is very easy to set up with EFI or grub2 onward. Tailscale is probably overkill for this use case, given that you're already exposing pre-/early- boot to the network by setting up interfaces that early. This became much more hermetic with secureboot and TPMs, too.

try firejail insread

https://github.com/libimobiledevice/ifuse might help.

This is a remarkably small number given that Gentoo Portage is load bearing infrastructure under ChromeOS.

Amen. :)

In summary, with different business requirements you would build a different technical solution.

Yup! I was referring to the original TrueTime/Spanner papers, not whatever's currently deployed. The original paper makes reference to distributed ordering guarantees at the milliseconds' scale precision, which implies many more transactions in flight in the uncertain state and coarser distributed ordering guarantees than the much tighter upper bound you can set with nanoseconds' precision and microseconds' comms latency...

Alternatively, you could guarantee the same synchronization using PPS and PTP to each host's DCD pin of their serial port or to specialized hardware such as modern PTP-enabled smart NICs/FPGAs that can accept PPS input. GPS+PPS gets you to within 20-80ns global synchronization depending on implementation (assuming you're all mostly in the same inertial frame), and allows you to make much stronger guarantees than TrueTime (due to higher precision distributed ordering guarantees, which translate to lower latency and higher throughput distributed writes).

Interesting interface. Some feedback:

  - Email delay to Gmail inboxes for verifying an SSH key used via SSH via email is longer than the timeout of the "Waiting for verification email..." stage in the SSH key registration. Wait longer or provide a non-email way to authorize a new key. You could imagine a few ways to do this: Allow users to add/delete SSH keys from the website or exe.dev shell; create a bearer token/random string that I can generate from the exe.dev shell or website to associate a new SSH key; SSH key signatures (existing key signs new key); SSH CAs (like @cert-authority); etc.
  - SSH U2F/FIDO2 authentication support has become mainstream, and offers you a way to have homogeneous auth across web and SSH interfaces. Maybe consider unifying authN this way?
  - exe.dev ssh interface does not allow me to list SSH keys, only to delete them. Consider moving all authN/authZ functionality into an "auth" subcommand/submenu (like you have for "share") and support SSH pubkey CRUD in there.
  - You make some strong assumptions about email addresses that aren't true -- what happens on email address changes, lost email access, etc. This will become more important when you start billing (and possibly costly).
  - How do I manage persistent disks? Any way to attach them to a different VM after I'm done with them on the original one? Is there always a single PD per VM or can these be managed separately? What about data or database volumes? Can PDs be attached to one or multiple VMs at a time?

At what scale do you break even on fixed costs (wages, rents, etc.)?

They claim to be the first, but a simple google search shows several other ultrasonic kitchen knives out there. (369sonic, etc.) dating back years.

Liability.

The point was that Wikipedia blocking the UK today would not have the same effect as blocking it a decade ago during previous similar censorship attempts. It would likely be less effective.

However, those aren’t not the only two options. Wikipedia could block editing from the UK, or it could simply not comply and wait for an enforcement action.

What recourse would the UK have in any case of such an enforcement action if the law or regulation Wikipedia faces does not exist in the US, where Wikipedia is ostensibly based if it removed all financial or physical presence from the UK?

To all of the commenters recommending that Wikipedia block UK visitors: This is incredibly short-sighted in the age of LLMs, where Wikipedia does not need to exist in a country in order for the benefit of its existence to be felt. Such a move would likely just drive people to obtain dubious regurgitations of Wikipedia’s (freely available) content via their favorite LLM chatbot, in my opinion.

Pure pedantry. The meaning in context was clear to a reasonable fluent reader such as yourself. Instead of focusing on pedantic form, I suggest focusing on the intent and content of TFA: Is physical media cool again and if so, why.

Vernacular shifts as society and technology change. As evidence, it doesn't seem like you or GP misunderstood the journalist's meaning. This is pedantry.

Languages change over time.

Title is incorrect, it's also accessible via their website at the second-level domain (anewsession.com).

Jacob Andreas is one of the smartest people I’ve ever met.