diff --git a/src/app/originals/amazon-anthropic-mark-496b-backlog/page.tsx b/src/app/originals/amazon-anthropic-mark-496b-backlog/page.tsx
index e26c7489..43fd2638 100644
--- a/src/app/originals/amazon-anthropic-mark-496b-backlog/page.tsx
+++ b/src/app/originals/amazon-anthropic-mark-496b-backlog/page.tsx
@@ -86,7 +86,7 @@ export default function AmazonAnthropicMark496bBacklogPage() {
         <p>
           That resolves the first signpost from our Wednesday piece cleanly. We asked whether Amazon
           would confirm the split the market drew on Tuesday night, where backlog plus acceleration
-          gets rewarded and everything else gets sold. It did. Alphabet fell 5, Meta fell 10,
+          gets rewarded and everything else gets sold. It did. Alphabet fell 5, Meta fell 8,
           Microsoft rose 8, Amazon rose 9, and the two winners are the two companies that walked in
           holding a half-trillion-dollar order book.
         </p>
diff --git a/src/app/originals/gpt-56-luna-80-cut-sol-rewrote-inference-stack/page.tsx b/src/app/originals/gpt-56-luna-80-cut-sol-rewrote-inference-stack/page.tsx
index 5ec85116..74c6a64c 100644
--- a/src/app/originals/gpt-56-luna-80-cut-sol-rewrote-inference-stack/page.tsx
+++ b/src/app/originals/gpt-56-luna-80-cut-sol-rewrote-inference-stack/page.tsx
@@ -251,7 +251,7 @@ export default function GPT56Luna80CutSolInferenceStackPage() {
           >
             price war we wrote up in April
           </Link>
-          , with DeepSeek V3.2 setting the floor at roughly $0.28 input and $1.10 output, Gemini
+          , with DeepSeek V3.2 setting the floor at roughly $0.28 input and $0.42 output, Gemini
           3.6 Flash inside the same neighborhood, and Anthropic&apos;s Haiku tier holding a
           premium against the pack. Luna at $0.20 and $1.20 is now the floor for a closed-API
           frontier lab, undercuts DeepSeek on input, and matches or beats Gemini Flash across
diff --git a/src/app/originals/naver-nvidia-brookfield-10b-sovereign-ai-triangle/page.tsx b/src/app/originals/naver-nvidia-brookfield-10b-sovereign-ai-triangle/page.tsx
index b464f903..30369ce9 100644
--- a/src/app/originals/naver-nvidia-brookfield-10b-sovereign-ai-triangle/page.tsx
+++ b/src/app/originals/naver-nvidia-brookfield-10b-sovereign-ai-triangle/page.tsx
@@ -286,7 +286,7 @@ export default function NaverNvidiaBrookfield10BSovereignAiTrianglePage() {
           That matters because the offtake math on a sovereign AI factory is nothing like the
           offtake math on a frontier lab. NAVER cannot double revenue every six months. It runs a
           business inside Korea&apos;s advertising, commerce, and mapping economy, and its top
-          line is roughly $10B. The compute at Sejong has to service a longer, flatter demand
+          line is roughly $8 billion. The compute at Sejong has to service a longer, flatter demand
           curve than what Anthropic or OpenAI runs against, and Brookfield knows this. That is
           precisely why the term sheet is non-binding today and why the guarantee, when it
           hardens, will be a much stronger take-or-pay than any hyperscaler has ever asked a