• 0 Posts
  • 111 Comments
Joined 3 years ago
cake
Cake day: July 3rd, 2023

help-circle








  • That is the thing though, they don’t have any actual work. In theory, they simply ingest data and spit out logical decisions based on that data. In reality, the data has been so selectively filtered and manipulated at every layer of management on the way up that it becomes functionally meaningless. Executives are so bad at making the right decisions that their subordinates actively hide what’s really happening from them.

    Further, executives serve no useful function beyond tricking employees into feeling like they matter and the company cares about them. This is why they have hour long all-hands meetings where they communicate no useful information but constantly repackage obvious business goals like “sell more stuff” as “increase sales velocity” because they think using the right keywords is the same thing as intelligence. This performative but ultimately useless behavior is the primary reason for their exorbitant salaries. They distract everyone from the reality that the company is stealing all the money the business generates.

    Sum all that up and you get a superfluously verbose moron working with bad data to make decisions that have, at best, marginal impact on the stated goal of the organization and consumes a hugely inflated amount of resources in the process. That does sound like the perfect human equivalent of an LLM to me.











  • Kind of but not as much as you might hope. Most of the time when telcos add more capacity to a fiber route they are either pulling more fiber through existing conduit or laying new ones along the same path. Being able to shoot farther is great if you’re building an entirely new route and can plan out your regen locations accordingly but not as helpful if you’re using the same huts on the side of road that were designed for shorter distances. They already have land and/or lease agreements, power, access, and all the other required infrastructure in place using shorter distance calculations. It would be expensive, and probably impossible in many cases, to set up new regen points along their existing routes.

    Then there’s the added expense of testing, certifying, stocking, and training people how to use all this new gear, which probably has longer lead times to acquire. Add all that up and a 40% reach boost isn’t quite as enticing. That’s not to say it won’t be done in some cases. I believe some subsea cables use hollow core fiber and added reach in a situation like that is awesome. I just think most of them would rather invest in improving reach on the glass they already have in the ground than going with something like this.


  • That’s pretty cool, but it isn’t going to help AI data centers get bandwidth faster any time soon. Most fiber in the ground today, especially the kind of long haul fiber being discussed here, isn’t hollow core. Laying new fiber across the country is arguably just as difficult and expensive as building these AI data centers and it’s spread out over very long distances. It costs more than $1/ft to lay relatively cheap fiber in rural areas with no permitting fees and it only gets more expensive from there. Fiber builds between major US cities cost hundreds of millions of dollars.

    Additionally, glass based fiber can do these same speeds and even higher today. Granted, this reach is about 40% longer on average at 200 km without a regen, but that doesn’t save network operators as much money as it sounds like it would. Not enough to justify overbuilding their entire network with hollow core fiber, that’s for sure.