diff --git a/.github/workflows/ci.yaml b/.github/workflows/ci.yaml
index 9f47da1dfb5..89cfc506ffa 100644
--- a/.github/workflows/ci.yaml
+++ b/.github/workflows/ci.yaml
@@ -7,11 +7,11 @@ on:
       - "v*"
 
 env:
-  ORIGIN_REPO: smira/talos
-  ORIGIN_TAG: backports/v1.8.2
-  KERNEL_TAG: v1.8.1-6-g7553341
-  SBC_TAG: v1.8.1
-  RK3588_TAG: v1.8.1-6-gc9d4a5c
+  ORIGIN_REPO: siderolabs/talos
+  ORIGIN_TAG: v1.8.2
+  KERNEL_TAG: v1.8.2
+  SBC_TAG: v1.8.2
+  RK3588_TAG: v1.8.2
 
 jobs:
   talos-build:
diff --git a/website/content/v1.8/talos-guides/install/single-board-computers/turing_rk1.md b/website/content/v1.8/talos-guides/install/single-board-computers/turing_rk1.md
index 8d4a60e9790..b5b894681a1 100644
--- a/website/content/v1.8/talos-guides/install/single-board-computers/turing_rk1.md
+++ b/website/content/v1.8/talos-guides/install/single-board-computers/turing_rk1.md
@@ -72,7 +72,7 @@ docker save ghcr.io/nberlee/sbc-turingrk1:v1.8.<currentversion> | tar x --strip-
 tar xvf _out/sha256/$(ls _out/sha256/ -1Sr | head -1) --strip-components=4 artifacts/arm64/u-boot/turingrk1/u-boot-rockchip-spi.bin
 ```
 
-As the spi-image may be too small for the BMC you may have to flash the raw image first to the eMMC:
+As the spi-image does not contain the full GPT tables need have to flash the raw image first to the eMMC, *even if you used Talos on eMMC before*:
 ```
 tpi flash -n <NODENUMBER> -i metal-turing_rk1-arm64.raw
 ```
@@ -103,7 +103,12 @@ Following the instructions in the uart output to connect to the interactive inst
 talosctl apply-config --insecure --mode=interactive --nodes <node IP or DNS name>
 ```
 
-Once the interactive installation is applied, the cluster will form and you can then use `kubectl`.
+or use `talosctl gen config` and apply it
+```bash
+talosctl apply-config --insecure --nodes <node IP or DNS name> -f <worker/controlplane>.yaml
+```
+
+Once the (interactive) installation is applied, the cluster will form and you can then use `kubectl`.
 
 ## Retrieve the `kubeconfig`