<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
  <channel>
    <title>Renest — field reports</title>
    <link>https://renest.ai/blog</link>
    <description>Real measured runs on real hardware, failures included.</description>
    <language>en</language>
    <item>
      <title>Reference image, depth and a text mask in one workflow: handing it to someone else</title>
      <link>https://renest.ai/blog/ip-adapter-depth-clipseg-handed-off</link>
      <guid isPermaLink="true">https://renest.ai/blog/ip-adapter-depth-clipseg-handed-off</guid>
      <pubDate>Thu, 15 Oct 2026 00:00:00 GMT</pubDate>
      <category>Field reports</category>
      <description>A workflow that uses three node packs, models in four places and one model your licence won't let you pass on. We handed it to a second account and watched what they got.</description>
    </item>
    <item>
      <title>ReActor face swap on a rented GPU: keeping the models it downloads for itself</title>
      <link>https://renest.ai/blog/reactor-face-swap-on-a-rented-gpu</link>
      <guid isPermaLink="true">https://renest.ai/blog/reactor-face-swap-on-a-rented-gpu</guid>
      <pubDate>Wed, 14 Oct 2026 00:00:00 GMT</pubDate>
      <category>Field reports</category>
      <description>ReActor fetches a dozen models the first time it runs and needs a GPU build of onnxruntime. We moved a working ReActor setup to a fresh machine and watched what came with it, what the machine didn't have to download, and what to know about the licences.</description>
    </item>
    <item>
      <title>Wan 2.2 5B text-to-video on rented GPUs: where the time goes when you move it</title>
      <link>https://renest.ai/blog/wan22-video-on-rented-gpus</link>
      <guid isPermaLink="true">https://renest.ai/blog/wan22-video-on-rented-gpus</guid>
      <pubDate>Tue, 13 Oct 2026 00:00:00 GMT</pubDate>
      <category>Field reports</category>
      <description>An 18 GB Wan 2.2 text-to-video setup, built once and restored on a 3090, a 4090, a 5090 and an A5000. Download, dependency install and render times side by side, and what held up the slowest run.</description>
    </item>
    <item>
      <title>FLUX.2 Klein 4B on a 3090, 4090, 5090 and A5000: the same setup, four rented cards</title>
      <link>https://renest.ai/blog/flux2-klein-on-four-gpus</link>
      <guid isPermaLink="true">https://renest.ai/blog/flux2-klein-on-four-gpus</guid>
      <pubDate>Mon, 12 Oct 2026 00:00:00 GMT</pubDate>
      <category>Field reports</category>
      <description>We built a FLUX.2 Klein 4B text-to-image setup once, then restored it on four different cards. How long each took, what the pictures looked like next to the original, and what to take from it if you rent by the hour.</description>
    </item>
    <item>
      <title>What's new in renest 0.1.15</title>
      <link>https://renest.ai/blog/renest-0-1-15</link>
      <guid isPermaLink="true">https://renest.ai/blog/renest-0-1-15</guid>
      <pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate>
      <category>Releases</category>
      <description>After a restore, renest now tells you how to start the app, adds a renest start command, checks the GPU before you pay for the download, and names plugins that came back broken.</description>
    </item>
    <item>
      <title>Moving a ComfyUI setup off a network volume</title>
      <link>https://renest.ai/blog/move-from-a-network-volume</link>
      <guid isPermaLink="true">https://renest.ai/blog/move-from-a-network-volume</guid>
      <pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate>
      <category>Guides</category>
      <description>A network volume keeps your files in one region of one provider. How to pack the setup that works on it into a nest and bring it back on a GPU somewhere else, step by step.</description>
    </item>
    <item>
      <title>Every file verified, environment still dead: the host-side failures</title>
      <link>https://renest.ai/blog/verified-files-not-a-working-environment</link>
      <guid isPermaLink="true">https://renest.ai/blog/verified-files-not-a-working-environment</guid>
      <pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
      <category>Mechanisms</category>
      <description>Every file came back checked and the app still would not run: a GPU newer than the packed kernels, a CPU missing avx2, a driver too old. What we check now.</description>
    </item>
    <item>
      <title>The pack we verified that no clean machine could restore</title>
      <link>https://renest.ai/blog/same-machine-verification-trap</link>
      <guid isPermaLink="true">https://renest.ai/blog/same-machine-verification-trap</guid>
      <pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
      <category>Self-reports</category>
      <description>A pack that verified on the machine that made it failed to install on a clean one: half its lock came from the OS image. What same-machine checks prove.</description>
    </item>
    <item>
      <title>Right vendor table, wrong install path: how our driver check turned away good machines</title>
      <link>https://renest.ai/blog/driver-floor-wrong-install-path</link>
      <guid isPermaLink="true">https://renest.ai/blog/driver-floor-wrong-install-path</guid>
      <pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
      <category>Self-reports</category>
      <description>Our pre-check rejected machines that worked: the driver floor came from the right vendor table for the wrong install path. Three corrections, one measured fix.</description>
    </item>
    <item>
      <title>Same GPU, two images: the CPU vendor decided which one came back</title>
      <link>https://renest.ai/blog/cpu-vendor-decides-identical-images</link>
      <guid isPermaLink="true">https://renest.ai/blog/cpu-vendor-decides-identical-images</guid>
      <pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
      <category>Field reports</category>
      <description>One archived ComfyUI workflow, same-model GPUs: regenerated images split into two byte-identical groups, and the host's CPU vendor decided which. Aug 2026.</description>
    </item>
    <item>
      <title>A 23GB video nest, and the exact boundary of 'identical'</title>
      <link>https://renest.ai/blog/field-report-vol-3</link>
      <guid isPermaLink="true">https://renest.ai/blog/field-report-vol-3</guid>
      <pubDate>Wed, 15 Jul 2026 00:00:00 GMT</pubDate>
      <category>Field reports</category>
      <description>Same GPU model: the regenerated video is literally the same file. Different model — even same architecture: a uniform ~0.94 plateau. And the 0.9814 mystery is solved: it was the CPU vendor.</description>
    </item>
    <item>
      <title>A real workflow survives its first exam — and two clean conclusions get caveats</title>
      <link>https://renest.ai/blog/field-report-vol-2</link>
      <guid isPermaLink="true">https://renest.ai/blog/field-report-vol-2</guid>
      <pubDate>Wed, 15 Jul 2026 00:00:00 GMT</pubDate>
      <category>Field reports</category>
      <description>A workflow with 3 custom nodes and a LoRA is captured and revived intact — including a node that was already broken — while new data forces honest revisions to Vol. 1.</description>
    </item>
    <item>
      <title>Field report vol. 1 (July 2026): 26 restores across 2 clouds, judged by an earlier standard</title>
      <link>https://renest.ai/blog/field-report-vol-1</link>
      <guid isPermaLink="true">https://renest.ai/blog/field-report-vol-1</guid>
      <pubDate>Tue, 14 Jul 2026 00:00:00 GMT</pubDate>
      <category>Field reports</category>
      <description>July 2026: one ComfyUI + SDXL environment packed, the machine destroyed, and the nest restored 26 times across 2 clouds, 7 datacenters and 6 GPU types, every file verified. The pass line used here has since been replaced by a stricter one; the current record is on the Proof page.</description>
    </item>
  </channel>
</rss>
