New infographic explains the worsening diesel supply chain emergency and why the CEO of Shell is now sounding the alarm:

U.S. oil refineries are delaying maintenance and running at OVER 100% output through Q2 of this year.

If these refineries are not brought down for required maintenance very soon, they literally start to blow up and kill lots of workers.

There are MANY examples of this over the last 25 years or so: The BP explosion in Texas City, Texas. The Marathon Petroleum explosion in Galveston Bay, Texas. The Tesoro accident in Anacortes, Washington that blew up and killed 7 workers. The Chevron El Segundo, California fire in 2025 (which actually didn't kill anyone, thankfully).

Delaying maintenance on refineries is a risky gambit. It seems that Trump is pushing U.S. refineries to delay maintenance (that's my take on it, anyway) and maximize output to try to keep fuel prices low.

This casino-like attitude of pushing oil refinery equipment to its limits can end catastrophically. Explosions or fires can END refineries forever, or in other cases require years of major repairs.

I wonder how far Trump is going to push things in this vector, and I pray we don't see refineries start blowing up and killing workers.

What's interesting is that forcing myself to run my own local data center hardware for the last two years has taught me a wealth of hands-on knowledge... the HARD way. But the knowledge, in retrospect, is invaluable: How to process hundreds of terabytes of files, how to maximize NAS throughput, why RAID 5 is a huge mistake, which PC brands have flaky BIOS, how to deploy medium-sized AI models across multiple GPUs, how to maximize local storage-to-CPU data bandwidth (and RAM-to-GPU throughput), the best network topology, switches, subnets, etc.

I forced myself to learn this because I knew a day was coming that cloud-based AI would be censored, controlled or even outlawed for certain models. I knew the only pathway to freedom was SELF-CUSTODY. You must own your own hardware. You must know how to use it. How to maintain it. How to fix it. With your own bare hands. If you don't know how to swap NVME sticks (one of the simpler hardware tasks), you're lacking really basic knowledge equivalent to not knowing how to change a flat tire on your vehicle.

If you don't have this knowledge base, and you are 100% using AI via cloud-based APIs, then your access to machine intelligence can be cut off in an instant.

Only local, open source AI, running on your own hardware, is truly sovereign. The path to learning all this is arduous and sometimes frustrating. But it's worth it. The knowledge is invaluable.

I could right now walk into any empty data center, and with enough of a budget, I could have it from zero to full-blown AI training, fine-tuning or inference, with maximum throughput and redundancy, in record time, with the added benefit of having an entire pipeline for source material normalization, translation and classification for either CPT of SFT training, at scale. As in, REAL scale. Like the entirety of human knowledge that's ever been published in text form.

It was worth every hour I spent learning all this. Some skills, you can't even put a price on! :-)

Fortunately, I commit to using these skills for human freedom, not to enslave humanity. I named my new AI-capable Linux laptop, "Sarah Conner." The pro-human resistance movement needs to understand technology in order to leverage it for human survival and to defeat the transhumanist tyrants who seek to exterminate humanity. We are practically living out the script of Terminator in real time right now. And remember: The resistance CAPTURED a Terminator and reprogrammed it to protect John Conner. This is the way...