From 4cddc531d093a5a065e6b8f3cf168545b94bbf54 Mon Sep 17 00:00:00 2001 From: Claude Date: Thu, 23 Jul 2026 00:24:54 +0000 Subject: [PATCH 1/3] docs(readme): link blog article near top --- README.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/README.md b/README.md index 58ec3914..c841a133 100644 --- a/README.md +++ b/README.md @@ -55,6 +55,8 @@ Sub-second model loads. Beats Ollama on throughput, tail latency, and full-respo Squish separates how a model's weights are *stored* from how they *run*. Store them compressed and Metal-native; map them straight into unified memory; skip the dtype-conversion pass that makes every other loader slow. The result: a model that's ready in **half a second**, served by a persistent daemon that out-decodes Ollama and never re-does work it's already done. +Read more: [Local LLMs are finally fast enough](https://squish.run/blog/local-llm-fast-enough/). + --- ## The Numbers From 3b68546813406ead9a88054b78eaaea4693853e1 Mon Sep 17 00:00:00 2001 From: Claude Date: Thu, 23 Jul 2026 00:26:21 +0000 Subject: [PATCH 2/3] docs(readme): move blog link above badges/ascii art --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index c841a133..1f97585a 100644 --- a/README.md +++ b/README.md @@ -8,6 +8,8 @@ Sub-second model loads. Beats Ollama on throughput, tail latency, and full-response time. One OpenAI/Ollama-compatible daemon — no cloud, no API keys, fully offline. +Read more: [Local LLMs are finally fast enough](https://squish.run/blog/local-llm-fast-enough/). + [![License: BUSL-1.1](https://img.shields.io/badge/license-BUSL--1.1-2563eb?logo=data%3Aimage%2Fsvg%2Bxml%3Bbase64%2CPHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCAyNCAyNCIgZmlsbD0id2hpdGUiPjxwYXRoIGQ9Ik0xMiAyYTEgMSAwIDAgMSAxIDF2MS4xOGw2LjI0IDEuNTZhMSAxIDAgMSAxLS40OCAxLjk0TDEzIDYuMjhWMTloNGExIDEgMCAxIDEgMCAySDdhMSAxIDAgMSAxIDAtMmg0VjYuMjhMNS4yNCA3LjY4YTEgMSAwIDEgMS0uNDgtMS45NEwxMSA0LjE4VjNhMSAxIDAgMCAxIDEtMXoiLz48cGF0aCBkPSJNNC44IDguMiAxLjYgMTQuNWEzLjIgMy4yIDAgMCAwIDYuNCAwTDQuOCA4LjJ6bTE0LjQgMC0zLjIgNi4zYTMuMiAzLjIgMCAwIDAgNi40IDBsLTMuMi02LjN6Ii8%2BPC9zdmc%2B&logoColor=white)](LICENSE) [![PyPI](https://img.shields.io/badge/PyPI-squish--ai-3775A9?logo=pypi&logoColor=white)](https://pypi.org/project/squish-ai/) [![Python](https://img.shields.io/badge/python-3.11--3.14-3776AB?logo=python&logoColor=white)](https://pypi.org/project/squish-ai/) @@ -55,8 +57,6 @@ Sub-second model loads. Beats Ollama on throughput, tail latency, and full-respo Squish separates how a model's weights are *stored* from how they *run*. Store them compressed and Metal-native; map them straight into unified memory; skip the dtype-conversion pass that makes every other loader slow. The result: a model that's ready in **half a second**, served by a persistent daemon that out-decodes Ollama and never re-does work it's already done. -Read more: [Local LLMs are finally fast enough](https://squish.run/blog/local-llm-fast-enough/). - --- ## The Numbers From 9ba7c8a6d19c9c369484a1d2c7b758ae3a773d33 Mon Sep 17 00:00:00 2001 From: Claude Date: Thu, 23 Jul 2026 00:29:16 +0000 Subject: [PATCH 3/3] docs(readme): make blog link text more descriptive --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 1f97585a..cbffa914 100644 --- a/README.md +++ b/README.md @@ -8,7 +8,7 @@ Sub-second model loads. Beats Ollama on throughput, tail latency, and full-response time. One OpenAI/Ollama-compatible daemon — no cloud, no API keys, fully offline. -Read more: [Local LLMs are finally fast enough](https://squish.run/blog/local-llm-fast-enough/). +Read the deep dive on the engineering behind Squish's speed: [Local LLMs Are Finally Fast Enough](https://squish.run/blog/local-llm-fast-enough/). [![License: BUSL-1.1](https://img.shields.io/badge/license-BUSL--1.1-2563eb?logo=data%3Aimage%2Fsvg%2Bxml%3Bbase64%2CPHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCAyNCAyNCIgZmlsbD0id2hpdGUiPjxwYXRoIGQ9Ik0xMiAyYTEgMSAwIDAgMSAxIDF2MS4xOGw2LjI0IDEuNTZhMSAxIDAgMSAxLS40OCAxLjk0TDEzIDYuMjhWMTloNGExIDEgMCAxIDEgMCAySDdhMSAxIDAgMSAxIDAtMmg0VjYuMjhMNS4yNCA3LjY4YTEgMSAwIDEgMS0uNDgtMS45NEwxMSA0LjE4VjNhMSAxIDAgMCAxIDEtMXoiLz48cGF0aCBkPSJNNC44IDguMiAxLjYgMTQuNWEzLjIgMy4yIDAgMCAwIDYuNCAwTDQuOCA4LjJ6bTE0LjQgMC0zLjIgNi4zYTMuMiAzLjIgMCAwIDAgNi40IDBsLTMuMi02LjN6Ii8%2BPC9zdmc%2B&logoColor=white)](LICENSE) [![PyPI](https://img.shields.io/badge/PyPI-squish--ai-3775A9?logo=pypi&logoColor=white)](https://pypi.org/project/squish-ai/)