<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>TokenTriage — Writing</title><description>TokenTriage is an honesty-first LLM gateway. Point your app at one base URL and see where every dollar goes — then cut your LLM bill only after the cheaper path is proven just as good, with evidence you can audit.</description><link>https://tokentriage.com/</link><language>en-us</language><item><title>Why we’d rather refuse than fabricate a savings number</title><link>https://tokentriage.com/blog/refuse-before-fabricate/</link><guid isPermaLink="true">https://tokentriage.com/blog/refuse-before-fabricate/</guid><description>An LLM router that always shows a win is telling you something — that it isn’t measuring. Here is the line TokenTriage draws, and why a refusal is the feature.</description><pubDate>Mon, 20 Jul 2026 00:00:00 GMT</pubDate><category>honesty</category><category>routing</category><category>evaluation</category></item><item><title>UNPRICED is not zero</title><link>https://tokentriage.com/blog/unpriced-is-not-zero/</link><guid isPermaLink="true">https://tokentriage.com/blog/unpriced-is-not-zero/</guid><description>Rendering a cost you don’t know as $0.00 is the most common lie in LLM cost tooling. Here’s how TokenTriage handles the models it can’t price.</description><pubDate>Tue, 14 Jul 2026 00:00:00 GMT</pubDate><category>honesty</category><category>cost-analytics</category><category>provenance</category></item></channel></rss>