Elektrine lite

← Feed

@tylercheung@mastodon.social

Post #2890563

2026-04-21 20:17 UTC

@dschaub@mstdn.social @atpfm@mastodon.social @marcoarment@mastodon.social I feel like this is actually one of the right calls, looking at asymco's graph of Apple vs. other FAANG capex during the AI craze. Making a LLM type model dependent on scraping is not in the DNA in the company; why force a round peg into a round hole?

Replies (1)

  • @tylercheung@mastodon.social 2026-04-21 20:17

    @dschaub@mstdn.social @atpfm@mastodon.social @marcoarment@mastodon.social But coincidentally, the company is better positioned to weather the AI revolution than most in that they have the most important part - the hardware w/ unified memory model and best perf/watt in the industry. The only thing they need (and admittedly it is a big IF) is to develop Metal/MLX into a library that beats CUDA at it's own game, and evangelize it to game and AI developers.

    Open ##2890564