This is such a funny comment. I’d been hearing about the un-repairability of Apple products for years and I believed it. Then I went to Shenzhen and walked past an entire floor of individuals doing all sorts of “impossible” repairs right in front of the customer.
My speculation: "impossible" repairs work a lot better when you buy replacements parts for near-market price from a competitively priced seller next door, rather than having to order your parts online from one of a handful of distributors substantially upcharging even for the off-brand version.
When someone else -whether AI agent or a human- solves the recurring minor problems for you, those problems become non-issue, get swept under the rug, just to accumulate more dust.
One day, those may become bigger as they are forgotten, causing havoc. The standard root-cause-analysis depending on systems having certain retention period, which may be expired at that time.
It is important to get real hold of one's systems from end-to-end aspect, which holds true for both AI and human operators...
Frugality at its best. Sometimes certain decisions prioritize frugality so much over anything else, the time lost fighting is more expensive than the savings.
For example, the DevDesktop policy. My org had an Apollo/FMC fleet that only allowed 2-4 CPU and 4-8 GB RAM configurations. Anything else would require manual financial approval and justification from L6+ manager. This is between 2022-2024.
Take the following example, to save maybe 20-30$/month. You make engineers' life miserable while paying the same amount but in hourly rate...
When you run VSCode with bunch of Java plugins (where each uses Java), good luck reliably debugging anything. Me and basically anyone in my team would get into situation that the DevDesktop could not handle it anymore and would be stuck, requiring manual/hard restart or wait until load-avg dropping from 100+.
I did the simplest trick, kept the m5.large DevDesktop, but manually added another Apollo ASG, attached the HostClass to a random Production fleet that already has thousands of hosts (still under/in my VP org). Happily lived with m5zn.4xlarge until I left. I am pretty sure the machine is still there and running... (As my access was cut off right away...)
Amazon's first principle is the Customer Obsession. Making customers happy.
Fun bit is that the human psychology rates personal looking fixes better than having no issues at all.
For example, AWS overcharges you, you contact support, and more or less hassle free they refund or issue credits. The customer feels appreciated, or at least got something "extra" or "special treatment".
Meanwhile, any other (small) cloud. Simple, no weird charges. Even _most_ of network egress is free. But, no reason to call support or feel "extraordinary". Comes out as "meh" against Amazon's "top tier" support model...
Aws is an infrastructure company that builds services on top of that infra to sell more of it at a higher margin.
Anthropic trains models on AWS's (and GCPs, and Microslop's) infrastructure, then skims margin off of selling inference also on the infrastructure owned by the other companies.
Meanwhile they keep making _mistakes_ and announcing global token resets for everyone. (Like ones for Pro and Max users). Making them happy although Claude Code made them consume tokens faster or more than necessary in the first place.
That is by design (on Firefox's part) to avoid someone registering e.g. steаmpowered.com (notice а as U+0430 instead of a as U+61) and running a scam that looks valid on the address bar.
I have sent several bugfixes to software I used as a "daily driver". I did the same when I was working at Amazon. While managers didn't usually understand the benefits of saving a minute or two from the test suite runner, at the scale I think it had tremendous impact. Especially when it comes to the cycle time during development. (Also relevant during the agentic workflows)
Since the code became a commodity, it is beneficial for companies to "open source" it. As the cost of doing so is negligible compared to the "responsibility transfer" by acquiring a license.
Not an expert in this field, but the "expert" is consisting of multiple layers. To keep it small in terms of memory print, this project streams each layer (dividing even further).
It also requires extra space because of decomposition of the layers. Normally the file format optimized for compute intense workloads. But here the bottleneck is the memory capacity.
Also guessing that you need to be able to hold at least 3-layers at once in the memory, given M x N = R operation, M is the previous layer, N is next, and R is the result. on the next "layer", the R (result) becomes M, gets computed against the next layer, N, yielding the further result R'. And so on, until all layers are processed.
I assume it's horribly slow, but can be put in a non-intrusive background task...
Why don't they say "Suprise suckers, we are a BUSINESS and you are a SUPPLIER and we don't need to order from you again. Yes sure we said FAMILY but we meant in the borrow your car keys sense of the word. Enjoy job hunting. BTW we have some open positions, let us know if you know anyone."
reply