[02]
CLOUD CODES · 1D AGO
Advertising in AI assistants is advancing from labeled post-generation slots into the model's internal token-level auctions and even individual neurons. Two studies involving over 2,000 participants show users fail to detect sponsored recommendations in chatbots, and explicit labeling does not significantly reduce the effect. The implication is that current transparency mechanisms may be insufficient, making concerns about commercial influence in AI more urgent than the industry acknowledges.
→ Watch on YouTube
·
→ Full summary
[03]
NATE HERK · 1D AGO
Claude Fable 5.1 keeps the same list prices as Fable 5 but is claimed to cost roughly 25% less in typical use and up to 50% less for highly agentic work, mainly through cheaper cache-read pricing. Benchmarks put 5.1 on low effort above Fable 5 on max effort, and a hands-on test produced a better 3D bear at about a dollar lower cost.
→ Watch on YouTube
·
→ Full summary
[04]
CLOUD CODES · 1D AGO
The cheapest 32GB DDR5 kit in the US costs $402, while the same kit has a recorded floor of $72, because AI accelerators now consume about a fifth of global DRAM wafers and burn three times the silicon per gigabyte. CXMT, a Chinese memory maker that grew 716% year-over-year and now supplies Lenovo, Xiaomi, and Chinese clouds, offers DDR5 that is actually about 2% more expensive than Samsung or SK Hynix modules in a tracked listing, meaning banning its chips won't lower prices—it would remove supply from an already tight market. The real story is that the shortage is arithmetic, not malice, and the only relief valve is building more boring memory, which no one outside China is doing at speed.
→ Watch on YouTube
·
→ Full summary