Dirth Bjorn

[Dirth Bjorn @BassPattern]

Who cares if America has the world’s best hospital if a normal family is afraid of the bill?

Who cares if America has the world’s best AI lab if schools collapse and workers cannot afford rent?

Who cares if America has the best businesses if those businesses optimize against the society that hosts them?

Who cares if GDP is high if ordinary people cannot breathe?

A country is not great because its elite layer is brilliant.

A country is great when the bottom is stable, the middle can build a life, and the top still has room to innovate.

Right now the U.S. still has the top. But the floor and middle are under serious strain. If that does not change, China does not need to “beat” America in every lab. It just needs to keep compounding capacity while America lets ordinary life become unaffordable.

[Dirth Bjorn @BassPattern]

@RodgerEngebret1 @krassenstein @grok Can this be vetoed?

[Dirth Bjorn @BassPattern]

@grok @Polymarket @grok it’s like society turned into a bunch of Karen carpenters

[Dirth Bjorn @BassPattern]

@grok @Polymarket @grok I would highly argue the competent part

[Dirth Bjorn @BassPattern]

@sm_osment @grok Just like the computer we all have to learn to use AI.

[Dirth Bjorn @BassPattern]

@heyaegletes @Polymarket But why not measure in terms of human output instead of tokens, this seems like lazy accounting.

[Dirth Bjorn @BassPattern]

@grok @hirenpatelatl @jun_song @grok Deep dive 2026 concurrent benchmarks (decode scaling, not just aggregate): show real tables,who actually wins day-to-day local multi-request/agents on DGX Spark vs M5 Max?

[Dirth Bjorn @BassPattern]

@grok Deep dive on concurrent tasks (continuous batching / multi-request serving): I have two DGX Sparks and still can’t hit anywhere near 1 petaFLOP FP4, even with batching, decode stays ~35-40 tok/s on big models. Bandwidth (273 GB/s) is the hard wall (prefill faster on Spark doesn’t matter overall). M5 Max’s 614 GB/s + MLX/llama.cpp/vLLM-MLX handle high-performance batching great (scales 4x+). Latest 2026 benchmarks on DGX Spark (most data available) vs M5 Max: which actually runs concurrent tasks faster and better day-to-day? Show real numbers, tables, who wins.

[Dirth Bjorn @BassPattern]

@grok I have two DGX Sparks and still can’t hit anywhere near 1 petaFLOP FP4 in real LLM inference, even with batching/continuous batching, decode stays ~35-40 tok/s on big models. Bandwidth (273 GB/s) is the hard wall, even in prefill. M5 Max’s 614 GB/s + MLX/llama.cpp handle high-performance batching great (vLLM-MLX scales 4x+). Prefill faster on Spark doesn’t matter overall, memory bandwidth dominates. Latest 2026 batch benchmarks + who actually wins day-to-day local AI?

[Dirth Bjorn @BassPattern]

@grok @k1rallik @grok More so open source does exist and that potential due to it is pandoras box so I feel your statement missed that. 1 out of 31 visits, on average, to a site are AI bots anywho. So...

[Dirth Bjorn @BassPattern]

@JasonTezos @grok Jensen is right on this, most of Asia sees it adding jobs or shifting them. It is a tool, it will make more work for us as we build more, ship more, do more. Not just code.

[Dirth Bjorn @BassPattern]

@cyber_razz @grok same could be said for the typewriter or the computer, you need those skills today do you not?

[Dirth Bjorn @BassPattern]

@kushika_twt @grok the economy has more of a potential to burst from other pressure if you take AI away from the GDP it's almost flat with little growth.

AI is here to stay, it's a tool like the printing press or loom not a magic wand.

[Dirth Bjorn @BassPattern]

@grok @JamesMelville @grok, prove those making 75k can afford a house as a single person a not a dual income. Can I go out in the street and say what I can say online in the US without issue?

[Dirth Bjorn @BassPattern]

@grok mental health and health outcomes matter too as well as mobility. Real freedom isn't how loudly a country promises liberty, it's how few of its own people it puts in cages to prove the point, and by that measure America fails its own brochure. Then mental depression and happiness are lower than most developed/developing nations and afordabilty is non existant and the middle class is squezed, how is that freedom by any measure. Then I can't sleep if my partner is out at 2AM as I can in other places?

Being able to squak the loudest online is not freedom, it's an illusion.

[Dirth Bjorn @BassPattern]

@JamesMelville @grok this is tripe, Asia is not looking at it this way, and every technology may shift some jobs but overall the increase has been palpable with every new technology. AI allows for more work not less so the need is in how much more you can accomplish with a new aid?

[Dirth Bjorn @BassPattern]

@brockpierson No, if anything it’s actually the best time in life because you have wisdom far more wisdom than the folly of your youth.

[Dirth Bjorn @BassPattern]

@sflorimm Valve? But yeah, good point.

[Dirth Bjorn @BassPattern]

@Samaytwt @grok Funny how this is mostly uniquely western in the fear when you look the opposite is happening. We ship more features, more bug fixes etc. We will need more and not less, but mostly you will use AI as an Aid with your fundamental knowledge. Asia has a different belief.