diff --git a/bin/render-streaming-markdown.ts b/bin/render-streaming-markdown.ts
index 2a25a0eb7..6649310f8 100644
--- a/bin/render-streaming-markdown.ts
+++ b/bin/render-streaming-markdown.ts
@@ -2,7 +2,9 @@
 //     cat README.md | deno run --allow-env --allow-read --allow-run bin/render-streaming-markdown.ts
 //     
 
-import $ from "jsr:@david/dax@0.41.0"
+import $ from 'jsr:@david/dax@0.41.0'
+// todo: replace dax with this:
+// import { exec } from 'https://deno.land/std/process.mod.ts';
 import { writeAllSync } from 'https://deno.land/std@v0.190.0/streams/mod.ts';
 
 let inputBuffer = ""
@@ -10,12 +12,43 @@ let inputBuffer = ""
 const decoder = new TextDecoder()
 const encoder = new TextEncoder()
 
+// This style works well for prompt..  but not chat
 for await (const chunk of Deno.stdin.readable) {
-  const decoded = decoder.decode(chunk);
-  inputBuffer += decoded
-  // console.log("$$$$$$$$$$", decoder.decode(chunk), "$$$zzz$$$")
+  // show immediately, but meanwhile…
+  writeAllSync(Deno.stdout, chunk);
+  // Collect it.
+  inputBuffer += decoder.decode(chunk);
+}
 
-  // --style auto is there to force it to output styled https://github.com/charmbracelet/glow/blob/2430b0a/main.go#L158
-  const output = await $`glow --style auto`.stdinText(decoded).text()
+// and now re-render it.
+if (inputBuffer) {
+  console.log('⬇️… and now rendered…⬇️');
+  const output = await $`glow --style auto`.stdinText(inputBuffer).text()
   writeAllSync(Deno.stdout, encoder.encode(output));
 }
+
+
+// This is a newline-buffered variant to avoid getting extra newlines in the output because we send it to glow too eagerly
+// it works but... the next problem is backtick codeblocks are broken up and... i'm sure there's more.
+//   definitely need a better solution
+ 
+// let remainingContent = '';
+// for await (const chunk of Deno.stdin.readable) {
+//   const decoded = remainingContent + decoder.decode(chunk);
+
+//   const lastNewline = decoded.lastIndexOf("\n");
+//   if (lastNewline !== -1) {
+//     // Flush everything up to it
+//     const output = await $`glow --style auto`.stdinText(decoded.substring(0, lastNewline + 1)).text()
+//     writeAllSync(Deno.stdout, encoder.encode(output));
+
+//     // Hold onto the remaining content to flush with the next chunk
+//     remainingContent = decoded.substring(lastNewline + 1);
+//   }
+// }
+
+// // Flush any remaining content
+// if (remainingContent) {
+//   const output = await $`glow --style auto`.stdinText(remainingContent).text()
+//   writeAllSync(Deno.stdout, encoder.encode(output));
+// }
diff --git a/fish/aliases.fish b/fish/aliases.fish
index 1646552d4..d56e09c3a 100644
--- a/fish/aliases.fish
+++ b/fish/aliases.fish
@@ -142,10 +142,11 @@ alias update_brew_npm_gem='brew_update; npm install npm -g; npm update -g; sudo
 abbr gemini "llm -m gemini-1.5-pro-latest"
 
 function gemi
+  # using https://github.com/simonw/llm-gemini and llm
   if test -n "$argv[1]"
-    llm prompt -m gemini-1.5-pro-latest $argv[1]   | deno run --allow-env --allow-read --allow-run bin/render-streaming-markdown.ts
+    llm prompt -m gemini-1.5-pro-latest $argv[1] | deno run --allow-env --allow-read --allow-run bin/render-streaming-markdown.ts
   else
-    llm chat --continue -m gemini-1.5-pro-latest   | deno run --allow-env --allow-read --allow-run bin/render-streaming-markdown.ts
+    llm chat --continue -m gemini-1.5-pro-latest
   end
 end