miniblog.

Text to speech systems seem to have largely avoided the uncanny valley effect. I've encountered robotic sounding voices but it's way less unsettling than bad CGI. I'm not sure why this is. Maybe looking at faces is just way higher bandwidth so more things can go wrong?

RISC-V seems like a great example of "commoditise your complements". If you're getting value higher in the hardware or software stack, a royalty-free ISA is valuable. Creating an ISA is a ton of work. Once it's basically viable, it creates interesting opportunities.

Valve exploring a custom shader compiler on Linux that leads to significantly higher frame rates than native code on Windows: https://www.forbes.com/sites/jasonevangelho/2019/07/11/valves-latest-linux-gaming-work-is-boosting-amd-vulkan-performance-by-up-to-44-percent/ (Shader compilers run during the game, so compilation time can make a big difference.)