<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Bobby</title>
    <description>The latest articles on DEV Community by Bobby (@bobbyhasfallen).</description>
    <link>https://dev.to/bobbyhasfallen</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F2308701%2F8d076893-477c-4557-ba94-557ded486cd3.jpg</url>
      <title>DEV Community: Bobby</title>
      <link>https://dev.to/bobbyhasfallen</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/bobbyhasfallen"/>
    <language>en</language>
    <item>
      <title>I copy-pasted text into Gemini 47 times in one day. Then I built something</title>
      <dc:creator>Bobby</dc:creator>
      <pubDate>Fri, 15 May 2026 07:15:00 +0000</pubDate>
      <link>https://dev.to/bobbyhasfallen/i-copy-pasted-text-into-gemini-47-times-in-one-day-then-i-built-something-5c5b</link>
      <guid>https://dev.to/bobbyhasfallen/i-copy-pasted-text-into-gemini-47-times-in-one-day-then-i-built-something-5c5b</guid>
      <description>&lt;p&gt;I counted. Not obsessively — I only started counting after I noticed I was doing it constantly and wondered how bad it actually was.&lt;/p&gt;

&lt;p&gt;47 times in a single session, I highlighted a sentence from a Gemini response, hit Ctrl+C, clicked the input box, typed a quote mark, Ctrl+V, closed the quote, then wrote my actual question.&lt;/p&gt;

&lt;p&gt;Every. Single. Time.&lt;/p&gt;

&lt;p&gt;Because if I didn't, Gemini would answer something adjacent to what I asked — but not &lt;em&gt;exactly&lt;/em&gt; what I asked. It would drift. Respond to the vibe of my question rather than the specific fragment I was looking at. And I'd get a perfectly confident answer to the wrong thing.&lt;/p&gt;

&lt;p&gt;The copy-paste wasn't optional. It was the only way to keep Gemini focused.&lt;/p&gt;




&lt;h2&gt;
  
  
  Why Gemini loses context (and it's not a bug)
&lt;/h2&gt;

&lt;p&gt;Here's what I eventually understood: Gemini isn't reading your follow-up question in isolation. It's reading your &lt;em&gt;entire conversation history&lt;/em&gt; and making a judgment call about what you probably mean.&lt;/p&gt;

&lt;p&gt;Most of the time that's fine. But when a response is long — multiple paragraphs, several distinct ideas — and you ask a follow-up, Gemini has to guess which part you're reacting to. It doesn't know if you're asking about paragraph two or paragraph five. So it hedges. It answers the general topic instead of the specific thing.&lt;/p&gt;

&lt;p&gt;The fix is almost embarrassingly simple: &lt;strong&gt;quote the exact text you're asking about.&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;"The light-dependent reactions occur in the thylakoid membranes"

Can you explain this like I'm 16?
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;That's it. That's the whole trick. You're not doing prompt engineering. You're just removing the ambiguity. Gemini stops guessing and starts answering the thing you actually pointed at.&lt;/p&gt;

&lt;p&gt;The quality difference is immediate and obvious. I started doing it on every follow-up question without exception.&lt;/p&gt;

&lt;p&gt;Which brought me back to the 47 copy-pastes.&lt;/p&gt;




&lt;h2&gt;
  
  
  The workflow was right. The friction was the problem.
&lt;/h2&gt;

&lt;p&gt;I knew quoting worked. I just hated doing it manually. The steps were:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Read Gemini's response&lt;/li&gt;
&lt;li&gt;Find the sentence I wanted to dig into&lt;/li&gt;
&lt;li&gt;Highlight it&lt;/li&gt;
&lt;li&gt;Copy it&lt;/li&gt;
&lt;li&gt;Click the input&lt;/li&gt;
&lt;li&gt;Type an opening quote mark&lt;/li&gt;
&lt;li&gt;Paste&lt;/li&gt;
&lt;li&gt;Type a closing quote mark&lt;/li&gt;
&lt;li&gt;Write my actual question&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Nine steps to do one thing. And I was doing it dozens of times a day.&lt;/p&gt;

&lt;p&gt;So I built a Chrome extension to collapse all nine steps into two: &lt;strong&gt;highlight, click.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;You select text in any Gemini response. A floating button appears above your selection. You click it. Your highlighted text is automatically quoted and locked into the reply bar as a context chip — visible, persistent, impossible for Gemini to ignore. You type your question and send.&lt;/p&gt;

&lt;p&gt;That's it. → &lt;strong&gt;&lt;a href="https://chromewebstore.google.com/detail/jhkodgigeemnmdmdikdkpcbmgbbopgni" rel="noopener noreferrer"&gt;QuoteReply for Gemini — free on the Chrome Web Store&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  What I learned building it
&lt;/h2&gt;

&lt;p&gt;A few things surprised me along the way.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Users highlight way more than they reply.&lt;/strong&gt; After launching &lt;a href="https://chromewebstore.google.com/detail/jhkodgigeemnmdmdikdkpcbmgbbopgni" rel="noopener noreferrer"&gt;QuoteReply&lt;/a&gt;, I watched the analytics and saw over 3,300 text selections in the first 30 days — but only ~150 replies sent. People are selecting text constantly. Most of them aren't doing anything with it. Which tells me the friction I felt wasn't just my problem.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The context chip changes how you think about the conversation.&lt;/strong&gt; When your highlighted text sits visibly in the reply bar, you write better questions. You're not vaguely gesturing at something — you're literally pointing at it. It makes you more precise without trying.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Small workflow fixes compound.&lt;/strong&gt; This isn't a revolutionary product. It's a 10-second interaction that happens dozens of times per session. But those 10 seconds add up, and the quality of the conversation improves because you stop losing the thread.&lt;/p&gt;




&lt;h2&gt;
  
  
  The actual tip, if you want one
&lt;/h2&gt;

&lt;p&gt;You don't need the extension to apply this. Start quoting your context manually in your next Gemini session. Pick one response, find the sentence you want to follow up on, put it in quotes before your question.&lt;/p&gt;

&lt;p&gt;Notice the difference in the answer you get back.&lt;/p&gt;

&lt;p&gt;If you find yourself doing it constantly and getting tired of the copy-paste — that's what &lt;strong&gt;&lt;a href="https://chromewebstore.google.com/detail/jhkodgigeemnmdmdikdkpcbmgbbopgni" rel="noopener noreferrer"&gt;QuoteReply for Gemini&lt;/a&gt;&lt;/strong&gt; is for. Free, no account, works in about 10 seconds. Install it, highlight something, and you'll feel the difference immediately.&lt;/p&gt;

&lt;p&gt;But honestly the habit matters more than the tool. The tool just makes the habit frictionless.&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;strong&gt;&lt;a href="https://chromewebstore.google.com/detail/jhkodgigeemnmdmdikdkpcbmgbbopgni" rel="noopener noreferrer"&gt;→ Install QuoteReply for Gemini (free)&lt;/a&gt;&lt;/strong&gt;&lt;br&gt;
&lt;strong&gt;&lt;a href="https://www.producthunt.com/products/quotereply-for-gemini?launch=quotereply-for-gemini" rel="noopener noreferrer"&gt;→ See it on Product Hunt&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;
&lt;/blockquote&gt;




&lt;p&gt;&lt;em&gt;QuoteReply just launched on Product Hunt — &lt;a href="https://www.producthunt.com/products/quotereply-for-gemini?launch=quotereply-for-gemini" rel="noopener noreferrer"&gt;upvote it here&lt;/a&gt; if this was useful. Would love your feedback.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>productivity</category>
      <category>webdev</category>
      <category>vibecoding</category>
      <category>gemini</category>
    </item>
    <item>
      <title>I made a chrome extension that fixes YouTube recommendations</title>
      <dc:creator>Bobby</dc:creator>
      <pubDate>Sun, 12 Jan 2025 07:44:17 +0000</pubDate>
      <link>https://dev.to/bobbyhasfallen/i-made-a-chrome-extension-that-fixes-youtube-recommendations-2kka</link>
      <guid>https://dev.to/bobbyhasfallen/i-made-a-chrome-extension-that-fixes-youtube-recommendations-2kka</guid>
      <description>&lt;p&gt;&lt;a href="https://recfix-yt.vercel.app/" rel="noopener noreferrer"&gt;RecFix Extension&lt;/a&gt; was one of my spur-of-the-moment side projects and is still in the MVP phase. Essentially, the extension guides the user to open YouTube.com, where it directly scrapes the recommendations from the YouTube homepage. From there, users can select the videos they like, and the extension takes care of the rest to help refine their YouTube algorithm. The interface is super intuitive and straightforward, so anyone can use it without confusion.&lt;/p&gt;

&lt;p&gt;The original idea for RecFix was to be a website where users would input the URLs of videos they liked, and it would fetch their recommendations. But that approach felt way too manual, and scraping recommendations that way didn’t work well—it also hurt the overall user experience. Another challenge I ran into was integrating external user YouTube API keys. After hours of trying to figure it out, I decided to use my own API key as the default. To manage usage, users are limited to 2 API requests per day (which is honestly more than enough for most people).&lt;/p&gt;

&lt;p&gt;Despite these challenges, I’m pretty happy with how the extension turned out in its current form. It strikes a balance between functionality and simplicity, allowing users to take control of their YouTube recommendations without jumping through hoops.&lt;/p&gt;

&lt;p&gt;The scraping process works seamlessly, and the limitation of 2 API requests per day ensures that the service remains free and sustainable while still being effective. Plus, by using my own API key, I’ve removed the hassle of requiring users to generate their own, making it a plug-and-play tool.&lt;/p&gt;

&lt;p&gt;Looking forward, I’ve got big plans for RecFix. Features like enhanced algorithms for even more accurate recommendations, multi-platform support, and maybe even an AI-powered analysis tool are on the horizon. The goal is to make this not just a YouTube extension but a comprehensive solution for managing video recommendations across platforms.&lt;/p&gt;

&lt;p&gt;For now, I’d love to hear your feedback! Whether it’s bugs, feature requests, or just your thoughts, feel free to share them. Together, we can make RecFix even better. Thanks for supporting this little project of mine! Ohh and I forgot tell you guys. The project is actually open source too.&lt;/p&gt;

</description>
      <category>webdev</category>
      <category>extensions</category>
      <category>youtube</category>
      <category>javascript</category>
    </item>
  </channel>
</rss>
