Публичный реестр каждого изменения — чтобы любой мог проверить, что редакционный firewall работает.
Список — это продукт; здесь — чек о том, как он меняется.
xmr.club
{
"field": "fees_page",
"summary": "Adds new cloud models beam and kimi-k3.1-pro to the per-token price table; plan prices and included credits are unchanged.",
"url_source": "discovered_conventional",
"minor": true,
"auto_acked": true
}{
"field": "fees_page",
"summary": "Free plan appears replaced by Pay as you go with the free starter credits removed; new paid mimo models added, and mistral-large-4 rose ~3%.",
"url_source": "discovered_conventional"
}{
"checks": "site 200",
"policy": "no unacked material",
"date": "2026-10-30"
}{
"field": "fees_page",
"summary": "Pricing page adds new models (clef $0.24 input, muse-spark-open) to the pay-as-you-go table. Existing plans, prices, and privacy FAQ are unchanged; this is an optional addition.",
"url_source": "discovered_conventional",
"minor": true,
"auto_acked": true
}{
"field": "fees_page",
"summary": "Adds new optional model mistral-large-4 to the price table ($0.68 in / $0.07 cached / $2.09 out); existing plan prices and models unchanged.",
"url_source": "discovered_conventional"
}{
"field": "fees_page",
"summary": "Fee schedule materially changed: DeepSeek rates doubled ($0.15→$0.30 input, $0.60→$1.20 output) with off-peak discount replacing peak surcharge; deepseek-v4-flash, glm-5.1, qwen3.5:397b removed.",
"url_source": "discovered_conventional"
}{
"field": "fees_page",
"summary": "Fee schedule expanded: new model 'deepseek-v4.1-flash' added ($0.15/$0.003/$0.60 per 1M, peak $0.30/$0.006/$1.20), and FAQ reworded from 'add extra credits' to 'buy usage credits' (credit-purchase mod",
"url_source": "discovered_conventional"
}{
"field": "fees_page",
"actor": "xbtoshi@kyc.rip"
}{
"field": "fees_page",
"summary": "Complete pricing overhaul: new plan structure (Team now $500/mo unlimited users vs $25/seat), new per-model token pricing table, added 2x 'peak pricing' (12:00-18:00 UTC weekdays), Pro/Max/Team usage-",
"url_source": "discovered_conventional"
}{
"field": "fees_page",
"summary": "New Team plan launched at $25/seat/mo (5-seat minimum) plus Enterprise plan with custom pricing. Previously 'Team' was 'Coming soon'. New team billing/usage model with shared balance. Material change ",
"url_source": "discovered_conventional"
}{
"field": "fees_page",
"summary": "New Max plan sign-ups paused due to capacity constraints. Added 'Plans' FAQ explaining pause and new 'Accounts' FAQ stating one-account-per-person policy. Material change affecting service availabilit",
"url_source": "discovered_conventional"
}{
"field": "fees_page",
"summary": "New Team plan tier added between Max and FAQ — 'Coming soon' enterprise offering with SSO, centralized billing, model access controls, MDM installer, and dedicated Slack support. Pricing not yet publi",
"url_source": "discovered_conventional"
}{"provider":{"id":"ollama","category":"ai","subcategory":"Local Runtime","name":"Ollama","tagline":"Run open-source LLMs locally — no network, no API key, no telemetry.","review":"**Local-first LLM runtime** that pulls quantised open-source model weights and exposes them through an OpenAI-compatible API on `localhost`. Listed at **Grade A** because Ollama is the canonical \"I want OpenAI-quality output without sending my prompts anywhere\" answer for the ~80% of users who don't want to compile `llama.cpp` themselves — MIT-licensed, no account anywhere in the flow, no telemetry on the inference path, and the inference itself never leaves your machine. The strongest privacy posture available in this directory because there is no operator to trust on the data path.\n\n**What it is.** Ollama is a desktop + server application that wraps the `llama.cpp` inference engine in a clean CLI (`ollama run <model>`, `ollama serve`), a model registry (`ollama.com/library` hosts quantised GGUF weights {"provider":{"id":"ollama","category":"ai","subcategory":"Local Runtime","name":"Ollama","tagline":"Run open-source LLMs locally — no network, no API key, no telemetry.","review":"**Local-first LLM runtime** that pulls quantised open-source model weights and exposes them through an OpenAI-compatible API on `localhost`. Listed at **Grade A** because Ollama is the canonical \"I want OpenAI-quality output without sending my prompts anywhere\" answer for the ~80% of users who don't want to compile `llama.cpp` themselves — MIT-licensed, no account anywhere in the flow, no telemetry on the inference path, and the inference itself never leaves your machine. The strongest privacy posture available in this directory because there is no operator to trust on the data path.\n\n**What it is.** Ollama is a desktop + server application that wraps the `llama.cpp` inference engine in a clean CLI (`ollama run <model>`, `ollama serve`), a model registry (`ollama.com/library` hosts quantised GGUF weights {"provider":{"id":"ollama","category":"ai","subcategory":"Local Runtime","name":"Ollama","tagline":"Run open-source LLMs locally — no network, no API key, no telemetry.","review":"**Local-first LLM runtime** that pulls quantised open-source model weights and exposes them through an OpenAI-compatible API on `localhost`. Listed at **Grade A** because Ollama is the canonical \"I want OpenAI-quality output without sending my prompts anywhere\" answer for the ~80% of users who don't want to compile `llama.cpp` themselves — MIT-licensed, no account anywhere in the flow, no telemetry on the inference path, and the inference itself never leaves your machine. The strongest privacy posture available in this directory because there is no operator to trust on the data path.\n\n**What it is.** Ollama is a desktop + server application that wraps the `llama.cpp` inference engine in a clean CLI (`ollama run <model>`, `ollama serve`), a model registry (`ollama.com/library` hosts quantised GGUF weights {"provider":{"id":"ollama","category":"ai","subcategory":"Local Runtime","name":"Ollama","tagline":"Run open-source LLMs locally — no network, no API key, no telemetry.","review":"**Local-first LLM runtime** that pulls quantised open-source model weights and exposes them through an OpenAI-compatible API on `localhost`. Listed at **Grade A** because Ollama is the canonical \"I want OpenAI-quality output without sending my prompts anywhere\" answer for the ~80% of users who don't want to compile `llama.cpp` themselves — MIT-licensed, no account anywhere in the flow, no telemetry on the inference path, and the inference itself never leaves your machine. The strongest privacy posture available in this directory because there is no operator to trust on the data path.\n\n**What it is.** Ollama is a desktop + server application that wraps the `llama.cpp` inference engine in a clean CLI (`ollama run <model>`, `ollama serve`), a model registry (`ollama.com/library` hosts quantised GGUF weights {"provider":{"id":"ollama","category":"ai","subcategory":"Local Runtime","name":"Ollama","tagline":"Run open-source LLMs locally — no network, no API key, no telemetry.","review":"**Local-first LLM runtime** that pulls quantised open-source model weights and exposes them through an OpenAI-compatible API on `localhost`. Listed at **Grade A** because Ollama is the canonical \"I want OpenAI-quality output without sending my prompts anywhere\" answer for the ~80% of users who don't want to compile `llama.cpp` themselves — MIT-licensed, no account anywhere in the flow, no telemetry on the inference path, and the inference itself never leaves your machine. The strongest privacy posture available in this directory because there is no operator to trust on the data path.\n\n**What it is.** Ollama is a desktop + server application that wraps the `llama.cpp` inference engine in a clean CLI (`ollama run <model>`, `ollama serve`), a model registry (`ollama.com/library` hosts quantised GGUF weights