josephh

I'm always confused by LLM proxies that claim to support tool calling. Even for Bifrost that claims to be doing it, at least when I was checking it out, I found out that while it injects the list of MCP tools that's available on the proxy-side, it doesn't actually make the call on client's behalf, and clients get confused by it (response returns MCP call request whose tool doesn't exist on the client-side).

If a multinational utility operates power grids in both the EU and the US, and a European company chooses to import power from the US grid, I don't know why EU would penalize the utility when the US grid has an outage. They really should be penalizing their European entities for failing to architect a local redundancy.

If EU's going to use this as an excuse to designate them as gatekeepers, maybe they should just block EU accounts from provisioning resources outside of Europe. These providers are taking all the blame for architectural decisions made entirely by their customers.

Rumor has it that it was an inside job and they intentionally shared the photo online as an alibi.

Much like there is 64-bit "code", there is also 32-bit "code" that can only be executed in the 32-bit (protected) mode, namely all the BCD, segment-related, push/pop-all instructions that will trigger an invalid opcode exception (#UD) when executed under long mode. In that strictest sense, "64-bit and 32-bit code can’t be mixed".

They still give you 500kbps of speed, which is enough for checking emails, voice calls, navigation, music streaming, etc.

The warranty isn't going to cover underside damages caused by going over a shallow bump

You can't just spin up an LLC and call it a separate company. OVHCloud is still OVHCloud US' subsidiary company.

From the FAQ page I linked:

> In accordance with our Privacy Policy, OVHcloud will comply with lawful requests from public authorities. Under the CLOUD Act, that could include data stored outside of the United States. OVHcloud will consider the availability of legal mechanisms to quash or modify requests as permitted by the CLOUD Act.

But then who can? No global cloud providers, including Hetzner and OVH, are free from CLOUD act because they have US presence[1].

1. https://us.ovhcloud.com/legal/faqs/cloud-act/

Stick to O-1 visa if you truly want the smartest people.

Wow, the actions alleged in the article sound like a corporate espionage that rival what state sponsored actors would do.

For the simple reason of not using Gecko

AFAIK, McDonald’s does this with their mobile app (they weren’t letting me log in with my password) But the problem with their implementation was that the magic link that they send you is wrapped in a click tracker whose domain is blocked by pihole (and the likes), and I could not reach the actual auth URL to complete the login process.

A more sensible thing that should've been done is to tokenize the case ID so that you can't just guess it with a numerical range. Also important that you don't leak your key business metrics (# of support cases over time).

> It’s almost as though the SAT score is not the only thing considered in admissions.

That's exactly the problem these Asian countries want to avoid/address: as soon as you deviate from the standardized test scores, you lay yourself open to potential biases and discrimination.

> Remind me of where the best universities in the world are again?

Quality of Bachelors programs (which is what's being discussed here) has pretty much zero bearing on how universities are ranked.

Except that the 130M figure absolutely isn't going to be correct if adjusted for literacy skill of their primary language.

When you make this kind of claim, please make sure to 1) disclose whether you're a lawyer, and 2) your jurisdiction.

Thanks! It never occurred to me that I can just tweak the system prompt to make sure the LL model never outputs symbols and initialisms as is.

Does anyone know of a good text normalization (?) library that converts symbols and initialisms into plain English before feeding them into a TTS model? All the models that I've used so far do a horrible job at synthesizing speech for them and I'm wondering whether this is the missing piece in the pipeline.

Which countries have laws that require their citizens to "support, assist, and cooperate with national intelligence efforts"?

Does this also impact server manufacturer customized images?

Could you provide a list of names who can/are willing to do it for less? I think our executive recruiting folks (from a lage-cap company) would be more than happy to get some referrals :)

/s

Of course, we need to continue the effort to put out fire with gasoline.

This could be a double negative...

ACC with LKAS is not the same as "you don't need to keep your hands on the wheel". AFAIK, It's only GM (Super Cruise), Ford (Blue Cruise) and Mercedes Benz (Drive Pilot) who have ACC systems that are designed to be completely handsfree on select motor ways.

Does anyone know the latest on having WG support dynamically assigned IP for peers? There have been some work done[1], but it seems to have stalled for quite some time.

1. https://github.com/WireGuard/wg-dynamic

In case you're referring to their servers (from https://dl.dell.com/manuals/common/poweredge_pcie_cooling.pd... ):

> The automatic system cooling response for third-party cards provisions airflow based on common industry PCIe requirements, regulating the inlet air to the card to a maximum of 55°C. The algorithm also approximates airflow in linear foot per minute (LFM) for the card based on card power delivery expectations for the slot (not actual card power consumption) and sets fan speeds to meet that LFM expectation. Since the airflow delivery is based on limited information from the third-party card, it is possible that this estimated airflow delivery may result in overcooling or undercooling of the card. Therefore, Dell EMC provides airflow customization for third-party PCIe adapters installed in PowerEdge platforms.

You need to use their RACADM interface to update the minimum LFM for your card.

FYI, in conjunction with Perspective API[1], this is the list of words that authors used to determine whether a tweet is considered hateful: https://github.com/dan-hickey1/musk-hate-lexicon/blob/main/h...

1. https://perspectiveapi.com/

South Koreans have realized that their alternative choice is just as corrupt, if not more.

Some of the things that have happened since:

- President Moon's right hand was convicted for opinion rigging leading up to the election by overseeing a bot farm.

- The country's chief prosecutor resigned after the administration's desperate attempt to cover up their own corruption, joined the opposition party, won the primary to become their presidential candidate, then actually winning (!) the election (as absurd as it sounds, this is roughly equivalent to Merrick Garland joining the GOP, getting nominated as their candidate, and winning the election).

> I like to see more tests moving to open-book model where test takers are allowed to use all resources, as in work.

Open-book model does not mean you're entitled to all resources, including paying someone to solve all the questions for you, or mooching answers off your peers. Cheating isn't limited to closed book exams, and to that extent, it's still an unsolved problem.