Explainers
Evergreen explainers on AI models, labs, chips, data centers, benchmarks, robotics and events.

Gemini 4 Argon review: strong evaluations, restricted access
Argon has strong published evaluations, but restricted access leaves limited evidence from everyday workflows.

Claude Opus 5.5 review: long-workflow gains and missed bugs
Opus catches additional bugs in CodeRabbit’s review pipeline, while missing some issues the baseline caught.

Claude Sonnet 5.5 review: strong results, heavy max-effort runs
Sonnet’s review speed and token use vary with the workload and effort setting.

GPT-6.1 Sol review: capable coding, uneven interactive speed
Sol’s published coding gains come with higher token use and mixed evidence on interactive speed.

AI data center vs traditional data center: what's different
An AI data center runs dense, liquid-cooled accelerator racks networked into one cluster, backed by large power deals.

Nvidia B200 vs H100: what Blackwell changes
B200 vs H100 from Nvidia's own spec sheets: two dies instead of one, FP4, fifth-generation NVLink and more memory.

H100 rental price: what clouds charge per GPU-hour
The H100 rental price on the providers' own pricing pages runs from $1.78 to $11.06 per GPU-hour.

Nvidia H100: specs, memory and what it is used for
Nvidia H100 specs: 80 GB HBM3 (SXM) or 94 GB (NVL), FP8 dense and sparse, 700 W per GPU, 10.2 kW per DGX H100.
