16:[[47702541,{"by":"juretriglav","descendants":24,"id":47702541,"kids":[47705460,47709431,47762139,47704380,47707143,47705771,47711536,47714337,47706446],"score":150,"time":1775736091,"title":"A WebGPU implementation of Augmented Vertex Block Descent","type":"story","url":"https://github.com/jure/webphysics"}],[47654488,{"by":"spencer9714","descendants":0,"id":47654488,"kids":[47654505],"score":5,"time":1775427240,"title":"Show HN: Runfra – Decentralized GPU cluster designed for bulk generation","type":"story","url":"https://runfra.com/playground"}],[47639207,{"by":"pykul","descendants":0,"id":47639207,"kids":[47648572,47644575,47639969],"score":7,"text":"I built this after hitting the same wall repeatedly — no good \nway to enforce token budgets in application code. Provider \ncaps are account-level and tell you what happened, not what \nis happening.

Two ways to add it:

  # Direct client wrapper\n  client = tokencap.wrap(anthropic.Anthropic(), limit=50_000)\n\n  # LangChain, CrewAI, AutoGen, etc.\n  tokencap.patch(limit=50_000)  \n

\nFour actions at configurable thresholds: WARN, DEGRADE \n(transparent model swap), BLOCK, and WEBHOOK. SQLite out of \nthe box, Redis for multi-agent setups.

One design decision worth mentioning: tokencap tracks tokens, \nnot dollars. Token counts come directly from the provider \nresponse and never drift with pricing changes.

Happy to answer any questions.","time":1775311753,"title":"Show HN: Tokencap – Token budget enforcement across your AI agents","type":"story","url":"https://github.com/pykul/tokencap"}],[47582537,{"by":"markfront","descendants":1,"id":47582537,"kids":[47585041,47644547,47582538],"score":7,"time":1774928937,"title":"Show HN: Free AI API gateway that auto-fails over Gemini, Groq, Mistral, etc.","type":"story","url":"https://github.com/msmarkgu/RelayFreeLLM"}],[47639779,{"by":"jrandolf","descendants":104,"id":47639779,"kids":[47641080,47641892,47642143,47642955,47640846,47641009,47640970,47671229,47642963,47641180,47641192,47640910,47648009,47642056,47641031,47642910,47647378,47647133,47655252,47640961,47642315,47644028,47641013,47644369,47642492,47646789,47646193,47740375,47641753,47640928,47644512,47671256,47670843,47649558,47643964,47661571,47642320,47643547,47642253,47695902,47643473,47647420,47642285,47648617],"score":188,"text":"Running DeepSeek V3 (685B) requires 8×H100 GPUs which is about $14k/month. Most developers only need 15-25 tok/s. sllm lets you join a cohort of developers sharing a dedicated node. You reserve a spot with your card, and nobody is charged until the cohort fills. Prices start at $5/mo for smaller models.

The LLMs are completely private (we don't log any traffic).

The API is OpenAI-compatible (we run vLLM), so you just swap the base URL. Currently offering a few models.","time":1775315938,"title":"Show HN: sllm – Split a GPU node with other developers, unlimited tokens","type":"story","url":"https://sllm.cloud"}],[47605792,{"by":"mijkal","descendants":2,"id":47605792,"kids":[47605806,47618688],"score":4,"time":1775073692,"title":"Show HN: Mycellm – BitTorrent for LLMs, pool GPUs into federated networks","type":"story","url":"https://mycellm.ai"}],[47612724,{"by":"AbuAssar","descendants":111,"id":47612724,"kids":[47613997,47613875,47613847,47613747,47613744,47614043,47613702,47620612,47613572,47613943,47615249,47613741,47618566,47617455,47615506,47613616,47613962,47613915,47615216,47619871,47617845,47617374,47623645,47613611,47615527,47762971,47621900,47614888,47638027,47621874,47620770,47619878,47617466,47614131,47624008,47614904,47617657,47617223,47618463,47615304],"score":572,"time":1775127891,"title":"Lemonade by AMD: a fast and open source local LLM server using GPU and NPU","type":"story","url":"https://lemonade-server.ai"}]]

1.	Show HN: Runfra – Decentralized GPU cluster designed for bulk generation(runfra.com) 5 points by spencer9714 88 days ago \| 0 comments
2.	1 points by spencer9714 91 days ago \| discuss
3.	1 points by spencer9714 91 days ago \| discuss

user:	spencer9714
created:	April 2, 2026
karma:	13