one time had cancel Google Colabs and really I couldn't figure out have to yell at them in support ticket to remove my subscription (eventually they did)
I don't get why the prices jumped so much - looks like panic and hoarding because:
- The Middle East produces roughly 30% of the world's oil
- But about 20% of total global oil consumption flows through this strait (less than 30% because of of domestic consumption and some pipelines avoiding strait)
I would understand if prices increases e.g. 50% but like more than 100% seems like a panic or manipulation
Oil demand is mostly inelastic. No matter how much or how little is produced, those who need it NEED it, so they'll compete with all others who similarly need it. The richest ones from among them get the oil first, and the poorest get nothing. The end price ends up being a function of how much oil is available versus how much the richest countries' absolutely irreducibly need for oil is versus how much wealth those countries can throw at the problem not to be left without before someone else with deeper pockets gets it.
Also, a third order effect of oil inelasticity is that if it cost too much , it decrease production and trade (might be less true now, but it was absolutely true in 2010), it lowers global demand, so the prices go down. Which is why markets can't predict oil prices, not really.
It would make sense if there was mostly inelastic demand for 80% of current production. I'm not sure if that is the case, but it might be. There is a lot of usage with no good substitute.
This is the exact reason that prices for U.S. benchmarks (e.g. WTI) are diverging from international ones (e.g. Brent). The spread between the two is the largest since 2011 and widening. There is only so much export capacity in the U.S., so there will be a glut compared to the rest of the world as a whole.
U.S. prices are up, but international prices are up even more.
they do are cheaper than SOTA but not getting dramatically cheaper but actually the opposite - GPT 5.4 mini is around ~3x more expensive than GPT 5.0 mini.
Similarly gemini 3.1 flash lite got more expensive than gemini 2.5 flash lite.
The crazy cheap models may be adequate for a task, and low cost matters with volume. I need to label millions of images to determine if they're sexually suggestive (this includes but is not limited to nudity). The Gemini 2.0 Flash Lite model is inexpensive and performs well. Gemini 2.5 Flash Lite is also good, but not noticeably better, and it costs more. When 2.0 gets retired this June my costs are going up.
agree, also they should take into account that their children will be eventually an adult and will be living in such system. Goverments should only focus on educating parents (available tools, recommendations) and maybe provide some open source tooling for parents.
> agree, also they should take into account that their children will be eventually an adult and will be living in such system
We also have to consider that "children" covers anywhere from birth to approximately 18 years old.
It is reasonable to expect a parent or their proxy (e.g. caregivers and teachers) to moderate access to the Internet in the early years. Yet older children and teenagers gradually gain more independence. For example: they are able to go places on their own, get their own phone, etc.. In the physical world, we have laws that recognize this, things like forbidding the sale of alcohol and tobacco to minors. Responsibility is placed on the vendor to check identification when selling such products and the customer's age is suspect. It would be absurd to place responsibility on parents in this case since the most a parent can do is educate their child.
Now I understand the Internet poses problems when it comes to similar transactions. For face to face transactions, appearing old enough is often sufficient (perhaps with a buffer to avoid liability) for access without presenting identification. While it isn't truly anonymous, there are cases where it can be reasonably anonymous. Unfortunately, transactions are mediated by machines on the Internet. You cannot make any assumptions about the other person. Making matters worse: it is extraordinarily difficult to do age verification without disclosing identify information, and to do so in a manner that is easily recorded. Whether that information is provided directly or through a third party is a moot point. It is still being provided.
I don't know how we go about solving this problem, but I do know two things:
- Placing all responsibility into the hands of parents is absurd, and would ultimately prove harmful to adolescents. It is creating a nanny-state where the nanny is the parent. The youth would be unable to gradually gain independence, nor develop an identity independent of their parents' whims.
- We live in a world which is eager to age-gate things that should not be. Sometimes this is for semi-legitimate reasons due to how the Internet is structured. For example: there is no good reasons why children and youth cannot participate in things like discussion forums, but those forums definitely cannot look like the "social media" we have today. Other times it is for despicable reasons, such as making value based judgements based upon ideology. (The left and right are both guilty of this.)
> Placing all responsibility into the hands of parents is absurd, and would ultimately prove harmful to adolescents. It is creating a nanny-state where the nanny is the parent. The youth would be unable to gradually gain independence, nor develop an identity independent of their parents' whims.
Nonsense. What kind of nanny state? Of course parent is a nanny because they growing and taking care of their own children. This has been for this way for centuries. Nanny state is the oppposite when state is reponsible for growing your kids.
You didn't explain how "Placing all responsibility into the hands of parents is absurd," or how "would ultimately prove harmful to adolescents".
> It is reasonable to expect a parent or their proxy (e.g. caregivers and teachers) to moderate access to the Internet in the early years. Yet older children and teenagers gradually gain more independence.
its easy to moderate or at least limit access to internet for kids.
- less than 5 years then have no own phone
- less than 10 years most likely also don't have own phone if have don't need sim but only wifi (parent control wifi and router)
- more than 15 years -> no control anyway and this age limitation trying not to limit above 15 years in most countries anyway
- between 10-15 you can just not sell simcards for those ages less than 15. Parents decide if will buy such simcard for their kids. Allow buy kid like simcard that restrict access to social media. You need unrestricted simcard that you can get only if you are >15 years old or your parent will buy for you. Sure it wont restrict everything but will limit access significantly
> Makes no sense, same with "I'm in a mood for asian food"
Thai, Japanese, Vietnamese, Indian food / cuisine even thought different is more probably closer to each other same like e.g. Polish and Spanish is closer to each other than to most other asian cuisine.
Asian countries developed with more overlap in basic ingredients, cooking techniques, and historical influence networks than Europe did. Historically there were 3 influence zones in Asia. There is a lot of pickling, fermenting, salting, drying. In Asia of these techniques were more or less unified. Fish sauces from different countries are Pepsi vs Coca-Cola level of difference.
> Polish and Spanish is closer to each other than to most other asian cuisine.
I'd say Polish has a lot of similarities with Asian cuisine. Sure, both have stews and sausages, but flavor profiles are very different: acidic vs sour.
I won't be able to tell difference between gyoza & wonton if they shaped the same, but surely I can tell difference between ravioli & uszka. Uszka is IMO closer to any dumpling from Asia than to anything European.
Those were brought to them most likely by China in one way or another.
> Yes, a filled pasta is a very different thing from dumpling,
You saying it like a filled pasta and a dumpling isn't the same twist on "filling encased in thin dough".
> There is nothing in South Asian cuisine similar to sashimi or to soy heavy stir fries.
Dish is ingredients and method. Stir-frying is a Chinese technique (technically multiples, but all originated in China). Ingredients get replaces all the time for various reasons. You're telling me Poriyal is not close relative to the OG stir-fry?
I'm not sure how you arrive at that opinion. Take the example of Punjabi food. It's heavily based around ghee and dairy. Does anything in Thai cuisine use butter except European style pastries?
The only major similarities I see uniting the national cuisines you listed (not regional ones) are things like curries and rice. The former arrived in Japan with European influence (where it's also common in colonial countries) and the latter isn't a feature common to all Asian cuisines (e.g. Mongolian).
FWIW this RCLI is only MIT license but their engine MetalRT is commercial. Not sure the license of their models I guess also not MIT. So IMHO this repo is misleading.
Not sure why they decided to reinvent the wheel and write yet another ML engine (MetalRT) which is proprietary. I would most likely bet on CoreML since it have support for ANE (apple NPU) or MLX.
Other popular repos for such tasks I would recommend:
Fair feedback on the README clarity, we've updated it to make the licensing distinction between RCLI (MIT) and MetalRT (proprietary) more prominent. That should have been clearer from day one.
On why we built MetalRT instead of using CoreML or MLX:
CoreML is optimized for classification and vision models, not autoregressive text generation. ANE is powerful for fixed-shape workloads but doesn't handle the dynamic shapes in LLM decode well.
MLX is much closer to what we need, and we respect what Apple has built. But MLX is a general-purpose array framework, it carries abstractions for developer ergonomics and portability that add overhead. MetalRT is purpose-built for inference only, and the numbers reflect that: 1.1-1.2x faster on LLMs (same model files) and 4.6x faster on STT.
We also needed one unified engine for LLM + STT + TTS rather than stitching three separate runtimes together. That doesn't exist in any of the alternatives listed.
The libraries you mentioned (FluidAudio, mlx-swift-audio, sherpa-onnx) are good projects. RCLI actually uses sherpa-onnx as it's fallback engine when MetalRT isn't installed. They solve different problems at different layers of the stack.
It uses hybrid retrieval (vector + BM25 with Reciprocal Rank Fusion) and runs at ~4ms over 5K+ chunks. Embeddings are computed locally with Snowflake Arctic, so nothing leaves you're machine.
Pentagon made few reports that Iran is not pursuing nuclear weapons - I bet they have better intelligence that you. Their killed religious leader made fatwa that forbid having or using weapons of mass destruction. Surprisingly now when he is gone they can pursue it after being attacked. Also worth to watch many Bibi talks since 1980s where he sais Iran will have nukes very soon and this didn't materialize in 40 years.
You realize how much work has gone into ensuring that didn't materialize in those 40 years, right? JCPOA... Stuxnet?
The Pentagon agrees that Iran is not officially pursuing nuclear weapons. However, there are CIA reports that indicated there may have been covert operations taking place that were exploring cruder nuclear weapons. I imagine that was the basis for the US bombing of Iran in 2025.
> You realize how much work has gone into ensuring that didn't materialize in those 40 years, right? JCPOA... Stuxnet?
JCPOA agreement was nuked by Trump administration. No, I don't buy your arguments, If Iran would wanted to have Nukes would have it already made those in 5 years for sure. Kim didn't have problems for making those.
Also, by his logic they succeeded in thwarting Iran's efforts for 40 years without resorting to bombing civilians. So we still need to see proof for why this is now, all of a sudden, the only way forward.
if for you to be qualified as regime is to murder tens of thousands of your own people then I think you put too high bar on it. I guess killing only few thousands or even few hundreds in your definition would rule out to someone being called totalitarian/autocratic regimes? How about not murdering own people but thousands other people? How is it called? Nazi germany AFAIK mostly murdered millions of other people.
People use this name (Regime) wrong - worth to at least read definion on wikipedia:
Look, I don't understand what you are debating here. I already agreed you can call USA regime just that should you choose to. I don't mind. You might get a scholarship to Columbia while at it.
My post was simply to clarify to the reader that PressTV is owned by the regime in Iran.
what US trauma supposed to be in this case? Only Americal Civil War and American Revolutionary War comes to my mind but have no clue how middle east mess could trigger those traumas?
The only traumas maybe related to money is ... Great Depression but it's not like somebody else was responsible for that
reply