This ThunderCore hardfork went pretty smoothly as expected. In fact, ThunderCore has a better track record of maintaining liveness and consistency than a lot of younger L1 and L2 chains :)

No. 1 on SushiSwap! ThunderCore Official ⚡️ is taking the lead on #DeFi Sushi.com V3 with a top-tier #TVL! 🏆 Top 6 chains: #Thundercore >> #Basechain #Arbitrum #Ethereum #Optimism #Polygon Exciting times ahead for the next bull run! 🐂

The blockchain state storage designs in Erigon and Geth and their evolution are quite interesting for the technically minded.

So, I'm rejoining ThunderCore and again taking on management responsibilities :)

Finally received my copy of "Make in Taiwan" (clarissawei.com/madeintaiwan) by Clarissa Wei!

I'm thrilled to share that shecc is designed to target 32-bit Arm and RISC-V architectures, it stands out as a self-compiling tool for a specific subset of C language. Simple yet effective, it employs fundamental optimization techniques. github.com/sysprog21/shecc

Lita has interesting zkVM tech. People interested in high level zero knowledge programming should check them out.

1/ As promised, here's Pt. 2: zkVM Design & Trade-offs. Great design is all about navigating constraints & making right choices. The same goes for zkVMs. Full Article: lita.foundation/blog/a-zero-kn…

AMD's new gpu software integration build system (TheRock) is pretty reasonable and I do like my 128GB Asus laptop with the Strix Halo chip but I still want more memory bandwidth for local LLMs.

PyTorch + ROCm on Windows: community powered. Thanks to the great work by Aaryaman "Jam" Vasishta and Scott Tsai on the port and TheRock team building a community first build and CI system for ROCm

Here are self-contained Pytorch wheels that support the Strix Halo GPU (gfx1151) for Windows and Linux. It's been awesome working with Aaryaman "Jam" Vasishta and Anush Elangovan's team on this for the past month. Expect the speed of improvement to increase. github.com/scottt/rocm-Th…

MicroBPF-Kernel is a RISC-V edge-device unikernel built from first principles to combine seL4's formal verification, Unikraft-style efficiency and in-kernel eBPF programmability. Each application is compiled into a minimal unikernel image (<2 MB) that boots in under 10 ms, safe!

Pytorch, torch vision, and audio, supporting the gfx1101, gfx1102, gfx1103, gfx1151, and gfx1201 delivered, with ComfyUI running 😃