https://rycode.dev/ Economics, risk, and systems. en Tue, 13 Oct 2026 07:18:42 GMT https://rycode.dev/ai/hardware/chips/ascend/ https://rycode.dev/ai/hardware/chips/ascend/ Fri, 09 Oct 2026 12:00:00 GMT The Ascend 910C is Huawei's current AI chip. Up to 384 of them form the Atlas 900 A3 SuperPoD, rated by Huawei at up to 307.2 PFLOPS of FP16, with 128 GB of memory per chip. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/cerebras-wse/ https://rycode.dev/ai/hardware/chips/cerebras-wse/ Fri, 09 Oct 2026 12:00:00 GMT The Cerebras WSE-3 is one chip the size of a silicon wafer: 4 trillion transistors, 900,000 AI cores, 44 GB of on-chip SRAM, and 125 PFLOPS of peak AI compute. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/maia/ https://rycode.dev/ai/hardware/chips/maia/ Fri, 09 Oct 2026 12:00:00 GMT Maia 200 is Microsoft's own inference chip: TSMC 3nm, over 10 PFLOPS of FP4, 216 GB of HBM3e at 7 TB/s, in a 750 W envelope. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/trainium/ https://rycode.dev/ai/hardware/chips/trainium/ Fri, 09 Oct 2026 12:00:00 GMT Trainium2 is the AI chip Amazon's Annapurna Labs designs for AWS: 1.3 PFLOPS of dense FP8 and 96 GiB of HBM per chip, 64 chips to a Trn2 UltraServer. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/openai/gpt-6-intelligent-ui/ https://rycode.dev/ai/models/openai/gpt-6-intelligent-ui/ Thu, 08 Oct 2026 12:00:00 GMT GPT-6 answers with visuals and controls in the Chat tab. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/ironwood/ https://rycode.dev/ai/hardware/chips/ironwood/ Thu, 08 Oct 2026 12:00:00 GMT Ironwood is Google's seventh-generation TPU, listed at $12 per chip-hour on demand and $5.40 on a three-year commitment in us-central1. Trillium (TPU v6e) is the generation before it. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/anthropic/claude-haiku-5-5/ https://rycode.dev/ai/models/anthropic/claude-haiku-5-5/ Wed, 07 Oct 2026 12:00:00 GMT Third Claude 5.5 model. contact@rycode.dev (Jeffrey) https://rycode.dev/risk/when-force-majeure-hits-the-ai-build-out/ https://rycode.dev/risk/when-force-majeure-hits-the-ai-build-out/ Mon, 05 Oct 2026 12:00:00 GMT A slip in AI campus building can travel to chip stocks and then to the whole market. The numbers below come from the /risk/ watch-list feed. contact@rycode.dev (Jeffrey) https://rycode.dev/risk/when-the-spender-runs-the-printer/ https://rycode.dev/risk/when-the-spender-runs-the-printer/ Mon, 05 Oct 2026 12:00:00 GMT Treasury has to sell more debt at auction. When buyers bid less, yields rise and the cost reaches mortgages. The numbers below come from the /risk/ watch-list feed. contact@rycode.dev (Jeffrey) https://rycode.dev/risk/oil-and-the-100-line/ https://rycode.dev/risk/oil-and-the-100-line/ Mon, 05 Oct 2026 12:00:00 GMT This brief tracks WTI crude against the $100 line it settled above on the day of the Fed's first rate hike, and the 10-year Treasury yield against 5%. The numbers below come from the /risk/ watch-list feed. contact@rycode.dev (Jeffrey) https://rycode.dev/economics/still-1998-not-1999/ https://rycode.dev/economics/still-1998-not-1999/ Mon, 05 Oct 2026 12:00:00 GMT This boom already has revenue and profit. The open question is the price paid for those earnings. The numbers below come from the /economics/ watch-list feed. contact@rycode.dev (Jeffrey) https://rycode.dev/economics/jobs-print-was-strong-mix-is-the-story/ https://rycode.dev/economics/jobs-print-was-strong-mix-is-the-story/ Mon, 05 Oct 2026 12:00:00 GMT Hiring people who do the work with AI, not only people who manage that work. The numbers below come from the /economics/ watch-list feed. contact@rycode.dev (Jeffrey) https://rycode.dev/risk/six-percent-that-stays/ https://rycode.dev/risk/six-percent-that-stays/ Mon, 05 Oct 2026 12:00:00 GMT If the 30-year Treasury yield reaches 6% and stays there, new long-term government debt and mortgages cost more. The numbers below come from the /risk/ watch-list feed. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/blackwell/ https://rycode.dev/ai/hardware/chips/blackwell/ Sun, 04 Oct 2026 12:00:00 GMT GB300 NVL72 is 72 Blackwell Ultra GPUs and 36 Grace CPUs, with 20 TB of GPU memory, 130 TB/s NVLink, and 1,440 PFLOPS of FP4 Tensor Core with sparsity. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/data-centers/eagle-mountain/ https://rycode.dev/ai/hardware/data-centers/eagle-mountain/ Sun, 04 Oct 2026 12:00:00 GMT Utah. Raised to more than $3 billion on 14 September 2026. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/data-centers/hyperion/ https://rycode.dev/ai/hardware/data-centers/hyperion/ Sun, 04 Oct 2026 12:00:00 GMT Richland Parish, Louisiana. 5 gigawatts and more than $50 billion. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/instinct/ https://rycode.dev/ai/hardware/chips/instinct/ Sun, 04 Oct 2026 12:00:00 GMT The custom Instinct GPU is based on the MI450 architecture, with Venice EPYC CPUs, ROCm, and Helios. MI300 and MI350 are the series Meta already runs. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/power/meta-nuclear/ https://rycode.dev/ai/hardware/power/meta-nuclear/ Sun, 04 Oct 2026 12:00:00 GMT Up to 6.6 gigawatts by 2035. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/meta/ https://rycode.dev/ai/hardware/chips/meta/ Sun, 04 Oct 2026 12:00:00 GMT Meta's stack is MTIA, a Broadcom 2nm accelerator through 2029, millions of Nvidia Blackwell and Rubin GPUs, and up to 6 gigawatts of AMD Instinct. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/mtia/ https://rycode.dev/ai/hardware/chips/mtia/ Sun, 04 Oct 2026 12:00:00 GMT MTIA 300, 400, 450, and 500. The 300 is in production with 216 GB HBM3E. The 400 is a 72-accelerator rack. Compute FLOPS from the 300 to the 500 is 25x. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/data-centers/prometheus/ https://rycode.dev/ai/hardware/data-centers/prometheus/ Sun, 04 Oct 2026 12:00:00 GMT New Albany, Ohio. Over 1 gigawatt and tens of thousands of graphics chips once complete. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/rubin/ https://rycode.dev/ai/hardware/chips/rubin/ Sun, 04 Oct 2026 12:00:00 GMT The Rubin GPU is 50 PFLOPS of NVFP4 inference and 288 GB of HBM4. Vera Rubin NVL72 is 72 Rubin GPUs and 36 Vera CPUs. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/what-xai-has-running/ https://rycode.dev/ai/what-xai-has-running/ Sun, 04 Oct 2026 12:00:00 GMT A first-cluster claim, not a finished campus. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/google/gemini-4-argon/ https://rycode.dev/ai/models/google/gemini-4-argon/ Thu, 01 Oct 2026 12:00:00 GMT Announced 30 Sep 2026. Trusted cyber defenders first, not the public. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/anthropic/claude-sonnet-5-5/ https://rycode.dev/ai/models/anthropic/claude-sonnet-5-5/ Tue, 29 Sep 2026 12:00:00 GMT Second Claude 5.5 model. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/openai/dots/ https://rycode.dev/ai/models/openai/dots/ Tue, 29 Sep 2026 12:00:00 GMT Always-on agent. Own cloud computer. Astra. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/openai/gpt-6.1-sol/ https://rycode.dev/ai/models/openai/gpt-6.1-sol/ Tue, 29 Sep 2026 12:00:00 GMT Near-Astra mid-tier at Sol list price. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/glossary/ https://rycode.dev/ai/glossary/ Mon, 28 Sep 2026 12:00:00 GMT Twenty plain terms for models, serving, and agents on this site contact@rycode.dev (Jeffrey) https://rycode.dev/ai/meta-named-an-enterprise-stack/ https://rycode.dev/ai/meta-named-an-enterprise-stack/ Mon, 28 Sep 2026 12:00:00 GMT The company is trying to sell the consumer agent line as a second business. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/frontier-price-meets-open-weight-ipo/ https://rycode.dev/ai/frontier-price-meets-open-weight-ipo/ Sun, 27 Sep 2026 12:00:00 GMT Open-weight cost curves are colliding with frontier listing calendars contact@rycode.dev (Jeffrey) https://rycode.dev/ai/swarms/what-an-agent-swarm-is/ https://rycode.dev/ai/swarms/what-an-agent-swarm-is/ Sun, 27 Sep 2026 12:00:00 GMT Two swarms, two breakthroughs, every metric on the table. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/grok-bot-set-a-new-bar/ https://rycode.dev/ai/grok-bot-set-a-new-bar/ Fri, 25 Sep 2026 12:00:00 GMT What these agent harnesses can actually do today, one capability at a time contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/anthropic/claude-opus-5-5/ https://rycode.dev/ai/models/anthropic/claude-opus-5-5/ Wed, 23 Sep 2026 12:00:00 GMT First Claude 5.5 model. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/openai/gpt-6-luna/ https://rycode.dev/ai/models/openai/gpt-6-luna/ Wed, 23 Sep 2026 12:00:00 GMT Cheaper, faster GPT-6 tier. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/openai/gpt-6-sol/ https://rycode.dev/ai/models/openai/gpt-6-sol/ Wed, 23 Sep 2026 12:00:00 GMT Mid-tier for complex coding and agentic work. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/google/gemini-pro/ https://rycode.dev/ai/models/google/gemini-pro/ Tue, 22 Sep 2026 12:00:00 GMT Pro line for harder reasoning and tool use. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/google/gemini-flash-lite/ https://rycode.dev/ai/models/google/gemini-flash-lite/ Tue, 22 Sep 2026 12:00:00 GMT Low-latency Flash-Lite for high-throughput work. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/google/gemini-flash/ https://rycode.dev/ai/models/google/gemini-flash/ Tue, 22 Sep 2026 12:00:00 GMT Current Flash workhorse for long-horizon agents. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/google/gemma/ https://rycode.dev/ai/models/google/gemma/ Tue, 22 Sep 2026 12:00:00 GMT Open-weight family, separate from hosted Gemini. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/open-weight/glm-5-3/ https://rycode.dev/ai/models/open-weight/glm-5-3/ Tue, 22 Sep 2026 12:00:00 GMT Z.AI. Max effort on the AA Index. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/open-weight/kimi-k3/ https://rycode.dev/ai/models/open-weight/kimi-k3/ Tue, 22 Sep 2026 12:00:00 GMT Moonshot. Max effort on the AA Index. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/open-weight/mimo/ https://rycode.dev/ai/models/open-weight/mimo/ Tue, 22 Sep 2026 12:00:00 GMT Xiaomi. Downloadable weights near the closed frontier. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/open-weight/qwen/ https://rycode.dev/ai/models/open-weight/qwen/ Tue, 22 Sep 2026 12:00:00 GMT Alibaba. Open checkpoint noted beside the Max SKU. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/open-weight/step-5/ https://rycode.dev/ai/models/open-weight/step-5/ Tue, 22 Sep 2026 12:00:00 GMT StepFun. API live. Weights dated 15 Oct 2026. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/deepseek/api/ https://rycode.dev/ai/models/deepseek/api/ Mon, 21 Sep 2026 12:00:00 GMT Base URLs, current model ids, and retired aliases. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/deepseek/flash/ https://rycode.dev/ai/models/deepseek/flash/ Mon, 21 Sep 2026 12:00:00 GMT Current recommended API model. DeepSeek-V4.1-Flash. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/xai/grok-4-7/ https://rycode.dev/ai/models/xai/grok-4-7/ Mon, 21 Sep 2026 12:00:00 GMT Current coding and knowledge-work frontier. API grok-4.7. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/meta/llama/llama-4/ https://rycode.dev/ai/models/meta/llama/llama-4/ Mon, 21 Sep 2026 12:00:00 GMT Current open-weight generation. Scout and Maverick. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/meta/muse/code/ https://rycode.dev/ai/models/meta/muse/code/ Mon, 21 Sep 2026 12:00:00 GMT Multi-agent coding in the terminal, built for Spark. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/meta/muse/glimmer/ https://rycode.dev/ai/models/meta/muse/glimmer/ Mon, 21 Sep 2026 12:00:00 GMT Open-weight 30B multimodal, distilled from Spark. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/meta/muse/image/ https://rycode.dev/ai/models/meta/muse/image/ Mon, 21 Sep 2026 12:00:00 GMT Image generation and edit on Meta Model API. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/meta/muse/spark/ https://rycode.dev/ai/models/meta/muse/spark/ Mon, 21 Sep 2026 12:00:00 GMT Flagship reasoning / agentic / coding on Meta Model API and Meta AI. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/meta/muse/video/ https://rycode.dev/ai/models/meta/muse/video/ Mon, 21 Sep 2026 12:00:00 GMT Video generation in the Muse family. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/deepseek/r1/ https://rycode.dev/ai/models/deepseek/r1/ Mon, 21 Sep 2026 12:00:00 GMT Landmark open reasoner (Jan 2025). Not the current API SKU. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/deepseek/thinking/ https://rycode.dev/ai/models/deepseek/thinking/ Mon, 21 Sep 2026 12:00:00 GMT Dual thinking / non-thinking mode, shared across Flash and Pro. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/deepseek/v3/ https://rycode.dev/ai/models/deepseek/v3/ Mon, 21 Sep 2026 12:00:00 GMT Prior chat generation. API name deepseek-chat is retired. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/deepseek/v4-pro/ https://rycode.dev/ai/models/deepseek/v4-pro/ Mon, 21 Sep 2026 12:00:00 GMT Still listed on Models & Pricing. Service continues after 14 Sep 2026 (changelog). contact@rycode.dev (Jeffrey) https://rycode.dev/ai/chat-models-write-strings-system-one-returns-decisions/ https://rycode.dev/ai/chat-models-write-strings-system-one-returns-decisions/ Sun, 20 Sep 2026 12:00:00 GMT Why Jev changes the scoreboard for software automation contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/typesafe/jev/ https://rycode.dev/ai/models/typesafe/jev/ Sun, 20 Sep 2026 12:00:00 GMT Typed decisions for software, not a chat flagship. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/xai/models/grok-4-20/ https://rycode.dev/ai/models/xai/models/grok-4-20/ Sun, 13 Sep 2026 12:00:00 GMT Reasoning, non-reasoning, and multi-agent slugs. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/xai/models/grok-4-3/ https://rycode.dev/ai/models/xai/models/grok-4-3/ Sun, 13 Sep 2026 12:00:00 GMT Longer context. Batch supported. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/xai/models/grok-4-5/ https://rycode.dev/ai/models/xai/models/grok-4-5/ Sun, 13 Sep 2026 12:00:00 GMT Available on the API. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/xai/models/grok-4-6/ https://rycode.dev/ai/models/xai/models/grok-4-6/ Sun, 13 Sep 2026 12:00:00 GMT Prior flagship. Still on the API. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/xai/grok-bot/ https://rycode.dev/ai/models/xai/grok-bot/ Sun, 13 Sep 2026 12:00:00 GMT Named Bots on a cloud computer. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/xai/grok-build/ https://rycode.dev/ai/models/xai/grok-build/ Sun, 13 Sep 2026 12:00:00 GMT Agent that works the project. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/xai/models/grok-build-0-1/ https://rycode.dev/ai/models/xai/models/grok-build-0-1/ Sun, 13 Sep 2026 12:00:00 GMT Dedicated coding slug. Not the Grok Build product. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/xai/imagine/ https://rycode.dev/ai/models/xai/imagine/ Sun, 13 Sep 2026 12:00:00 GMT Image generation door. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/data-centers/memphis-colossus/ https://rycode.dev/ai/hardware/data-centers/memphis-colossus/ Sun, 13 Sep 2026 12:00:00 GMT Training campus in Memphis. 170 PB/s memory bandwidth and 2.8 Tb/s per server. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/terafab/ https://rycode.dev/ai/hardware/chips/terafab/ Sun, 13 Sep 2026 12:00:00 GMT Chip fab. SpaceX and Tesla. Phase 1 is $16.8B. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/ten-thousand-agents-is-not-a-genius/ https://rycode.dev/ai/ten-thousand-agents-is-not-a-genius/ Sun, 13 Sep 2026 12:00:00 GMT Ten thousand agents is not a genius contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/xai/voice/ https://rycode.dev/ai/models/xai/voice/ Sun, 13 Sep 2026 12:00:00 GMT Spoken Grok surface. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/anthropic/claude-fable-5-1/ https://rycode.dev/ai/models/anthropic/claude-fable-5-1/ Fri, 04 Sep 2026 12:00:00 GMT Most capable widely released. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/anthropic/claude-haiku-4-5/ https://rycode.dev/ai/models/anthropic/claude-haiku-4-5/ Fri, 04 Sep 2026 12:00:00 GMT Prior Haiku. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/anthropic/claude-opus-5/ https://rycode.dev/ai/models/anthropic/claude-opus-5/ Fri, 04 Sep 2026 12:00:00 GMT Default start. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/anthropic/claude-sonnet-5/ https://rycode.dev/ai/models/anthropic/claude-sonnet-5/ Fri, 04 Sep 2026 12:00:00 GMT Speed + intelligence. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/models/openai/gpt-6-astra/ https://rycode.dev/ai/models/openai/gpt-6-astra/ Fri, 04 Sep 2026 12:00:00 GMT Flagship for hard end-to-end work. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/next-token-engine/ https://rycode.dev/ai/next-token-engine/ Tue, 25 Aug 2026 12:00:00 GMT How staged training and tools shape a next-token engine. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/meta-broadcom/ https://rycode.dev/ai/hardware/chips/meta-broadcom/ Tue, 14 Apr 2026 12:00:00 GMT Broadcom will deliver custom accelerator technology for Meta's MTIA program through 2029. The first phase is more than 1 gigawatt. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/meta-amd/ https://rycode.dev/ai/hardware/chips/meta-amd/ Tue, 24 Feb 2026 12:00:00 GMT AMD will supply up to 6 gigawatts of Instinct GPUs. The first gigawatt is a custom MI450, with Helios racks. Meta already runs MI300 and MI350. contact@rycode.dev (Jeffrey) https://rycode.dev/ai/hardware/chips/meta-nvidia/ https://rycode.dev/ai/hardware/chips/meta-nvidia/ Tue, 17 Feb 2026 12:00:00 GMT Nvidia says the Meta partnership deploys millions of Blackwell and Rubin GPUs, including GB300 systems, plus Grace CPUs. contact@rycode.dev (Jeffrey)