https://siliconnoon.com/today-in-ai-october-21-2026/SiliconNoonen2026-10-21T22:09:06.000ZToday in AI — October 21, 2026https://siliconnoon.com/nvidia-kumo-relational-4-auroc-points-over-xgboost-no-training/SiliconNoonen2026-10-21T19:10:06.000ZNVIDIA Kumo Relational: 4 AUROC points over XGBoost, no traininghttps://siliconnoon.com/the-take-echotribble-asked-for-stealth-not-accountability/SiliconNoonen2026-10-21T16:10:15.000ZThe Take — ECHOTRIBBLE asked for stealth, not accountabilityhttps://siliconnoon.com/chatgpt-ads-100000-advertisers-and-a-1b-run-rate-in-200-days/SiliconNoonen2026-10-21T15:03:41.000ZChatGPT ads: 100,000 advertisers and a $1B run rate in 200 dayshttps://siliconnoon.com/openai-models-are-now-on-google-cloud-billed-against-gcp-commitments/SiliconNoonen2026-10-21T13:10:26.000ZOpenAI models are now on Google Cloud — billed against GCP commitmentshttps://siliconnoon.com/open-source-radar-october-21-graphify-cloudflare-audit-skill/SiliconNoonen2026-10-21T10:02:11.000ZOpen Source Radar — October 21: Graphify, Cloudflare audit skillhttps://siliconnoon.com/deep-dive-openais-12-hour-prompt-cache-costs-25x-to-write/SiliconNoonen2026-10-21T09:05:53.000ZDeep Dive — OpenAI's 12-hour prompt cache costs 2.5x to writehttps://siliconnoon.com/ai-101-what-is-ai-memory/SiliconNoonen2026-10-21T08:06:41.000ZAI 101 — What is AI memory?https://siliconnoon.com/spacex-q3-ai-revenue-hits-756b-52-of-the-company/SiliconNoonen2026-10-21T02:14:22.000ZSpaceX Q3: AI revenue hits $7.56B, 52% of the companyhttps://siliconnoon.com/today-in-ai-october-20-2026/SiliconNoonen2026-10-20T22:03:59.000ZToday in AI — October 20, 2026https://siliconnoon.com/google-open-sources-torchtpu-native-pytorch-on-tpus-no-rewrite/SiliconNoonen2026-10-20T19:11:06.000ZGoogle open-sources TorchTPU: native PyTorch on TPUs, no rewritehttps://siliconnoon.com/salesforce-buys-braze-for-566b-at-five-times-forward-revenue/SiliconNoonen2026-10-20T15:06:25.000ZSalesforce buys Braze for $5.66B at five times forward revenuehttps://siliconnoon.com/mercury-3-a-diffusion-llm-at-1068-tokenssec-010-per-million/SiliconNoonen2026-10-20T13:11:51.000ZMercury 3: a diffusion LLM at 1,068 tokens/sec, $0.10 per millionhttps://siliconnoon.com/open-source-radar-october-20-headroom-hyperframes-ai-job-search/SiliconNoonen2026-10-20T10:03:10.000ZOpen Source Radar — October 20: Headroom, HyperFrames, AI job searchhttps://siliconnoon.com/deep-dive-deepseek-banned-from-federal-agencies-its-weights-arent/SiliconNoonen2026-10-20T09:07:25.000ZDeep Dive — DeepSeek banned from federal agencies. Its weights aren'thttps://siliconnoon.com/ai-101-what-is-an-ai-browser/SiliconNoonen2026-10-20T08:06:33.000ZAI 101 — What is an AI browser?https://siliconnoon.com/gpt-61-sol-migrates-cobol-at-979-gpt-56-scored-54/SiliconNoonen2026-10-20T02:10:52.000ZGPT-6.1 Sol migrates COBOL at 97.9% — GPT-5.6 scored 5.4%