Some recent awesome optimizations merged in MLX: Built-in sliding window support in fused SDPA Use mma/nax for QMV with small rows and large N/K, making batched decoding 2~3x faster Dequantize fp4 into float16 directly, which makes prefilling 1.4x faster for old macs Add a 2-stage nax QMM kernel that makes small prompt prefill almost 2x faster for some models 1-bit affine quantization Thin GEMM kernels making small-M/large-N and large-M/small-N 1.5x faster 1.4x faster fused SDPA by prefetching K/V And there are actually a lot more awesome changes that I'm too exhausted to list them all, the commit history of MLX is an interesting read for anyone who wants to make open source contributions.

so gamblers bet heavily on brazil elections after seeing a clear trend and thinking ez money but then got hit by an out of distribution data point. and now they are crying? you love to see it!

Seeing all these AI startups in the US raising millions is honestly bittersweet. My own startup received €60K in funding, while on a recent trip to the US, I learned that a startup at a very similar stage had raised $1.5 MILLION?!

All these companies complaining about higher memory prices - just get your finance dept to invest in the same memory shares you are buying from. Math is simple use cash to buy the shares (always buy shares using cash) and use credit to buy the actual product! GGEZ

Markets are down because of CPI? We moved from Pluggables to CPO to NPO to CPI in a year? HOLY SHIT. We're witnessing a generational acceleration in technological development. Keep buying! Your favourite photonics stock will soon be making Co-Packaged Internet too! nw.