The latest articles on DEV Community by zephyr whimsy (@zephyr_whimsy_e62111ac412). https://dev.to/zephyr_whimsy_e62111ac412 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4036322%2Fe1359b8d-f753-45ef-8fd8-f52bbe8979ef.jpg https://dev.to/zephyr_whimsy_e62111ac412 en zephyr whimsy Sun, 19 Jul 2026 09:33:29 +0000 https://dev.to/zephyr_whimsy_e62111ac412/stop-pasting-webpages-into-chatgpt-you-are-wasting-65-of-your-tokens-402e https://dev.to/zephyr_whimsy_e62111ac412/stop-pasting-webpages-into-chatgpt-you-are-wasting-65-of-your-tokens-402e <p>If you paste webpages into ChatGPT or Claude regularly, you are probably wasting a third of your context window on nav menus, cookie banners, and "related articles" widgets.</p> <p>I measured it. A typical article page is ~24,000 characters as rendered HTML-to-text; the actual article is ~8,000. Everything else is website furniture — and the model has to spend attention (and you have to spend tokens) figuring out which is which.</p> <h2> The usual fixes, and where they break </h2> <p><strong>Copy-paste.</strong> Fast, but you drag in everything: sidebars, footers, comment CTAs. Tables turn to mush, code blocks lose their fences.</p> <p><strong><code>r.jina.ai</code> URL prefix.</strong> Genuinely clever — prefix any public URL and get Markdown back, no install. But it's a server-side fetch, so it breaks exactly where you need it most: Reddit threads (blocked), X posts (login wall), paywalled articles, anything behind your session.</p> <p><strong>Readability-based scripts.</strong> Work fine on clean blog posts, fall over on structured sites (docs with tabs, forums, video pages).</p> <h2> What actually solved it for me: convert in the browser </h2> <p>The insight is boring but important: <strong>your browser already has the fully rendered, logged-in page</strong>. Server-side readers have to fight for access; an extension just reads what you can see.</p> <p>My daily workflow now:</p> <ol> <li>Open any page — Reddit thread, X post, YouTube video, docs.</li> <li>One click → clean Markdown (headings, lists, tables, code fences preserved).</li> <li>Send straight to ChatGPT/Claude, or save as <code>.md</code> for Obsidian.</li> </ol> <p>On token count: clean Markdown of the same article runs <strong>~65% fewer tokens</strong> than raw copy-paste, because none of the furniture comes along. Over a month of heavy research use, that's real money on API workflows.</p> <p>Site-specific extractors matter more than I expected. Generic readability guesses; a Reddit extractor <em>knows</em> what a comment thread is (score, nesting, OP), a YouTube extractor grabs the transcript. Web2MD ships 16 of these.</p> <h2> When each tool wins </h2> <div class="table-wrapper-paragraph"><table> <thead> <tr> <th>Scenario</th> <th>Best tool</th> </tr> </thead> <tbody> <tr> <td>Public article, one-off, no install</td> <td> <code>r.jina.ai</code> prefix</td> </tr> <tr> <td>Automation/pipelines on public URLs</td> <td>Jina Reader API / Firecrawl</td> </tr> <tr> <td>Reddit, X, paywalls, logged-in pages</td> <td>Browser extension</td> </tr> <tr> <td>Batch-convert 50 URLs for an agent</td> <td>Extension + MCP</td> </tr> </tbody> </table></div> <p>I build <a href="https://web2md.org" rel="noopener noreferrer">Web2MD</a> (free tier, no signup for the web tools), so the browser-extension enthusiasm is earned bias — but the comparison table holds regardless of whose extension you use. If your AI workflow chokes on exactly the pages that matter, move the conversion into the browser.</p> <p>Happy to answer questions about the extraction internals — the Reddit one was the most annoying to get right.</p>