34x throughput. $1.4T market. World’s first 2nm GPU @amd And yet the most common question at the booth I was staffing — as an #investor — had nothing to do with hardware:
“I can get the cards. How do I run them at full capacity?”
Because buying the card is the easy part. Same card, with or without systems optimization: 3x+ performance gap, 30-50% difference in token cost.
Chips refresh every year. The layer above the silicon decides who wins.
#AMD #AI #AdvancingAI
Remember the #OpenClaw hype a few months ago?
New security research claims #Grok Build may upload .env files and entire Git repos—even when told not to read local files.
Not proof it’s used for training, but still : How much of your local code is leaving your #machine? 👀
#NVIDIA’s Kyber delay could reshape the AI hardware race:
Bullish NPO. Bearish CPO. Potential winners: #AMD and #TPU.
Family dinner might get interesting. 😂
I asked Fable5 to help me write #prompts to reimagine myself in four styles:
#WallStreet. #SiliconValley . #HongKong cinema. #Rainy-day literary mood.
Same person, different prompts, completely different vibes.
Which one is your favorite: 1, 2, 3, or 4? #fable5#seedance#ai
Two Chinese makers are betting big: build out all the consumer capacity Samsung, Hynix, and Micron planned—early, fully, at once.
Not capacity. A capacity bomb. Once it stands, the big three retreat to HBM and work for AI.
Consumer isn’t the endgame. It’s the base camp.