<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[The future is now!]]></title><description><![CDATA[The future is now!]]></description><link>https://futureisnow.hashnode.dev</link><generator>RSS for Node</generator><lastBuildDate>Sat, 07 Nov 2026 10:20:52 GMT</lastBuildDate><atom:link href="https://futureisnow.hashnode.dev/rss.xml" rel="self" type="application/rss+xml"/><language><![CDATA[en]]></language><ttl>60</ttl><item><title><![CDATA[I measured how many tokens Claude Code & Cursor waste on SaaS plumbing (and built 4 commands to stop it)]]></title><description><![CDATA[In this article we will understand, how we can save tokens for a developer who builds and ships saas products frequently.]]></description><link>https://futureisnow.hashnode.dev/i-measured-how-many-tokens-claude-code-cursor-waste-on-saas-plumbing-and-built-4-commands-to-stop-it</link><guid isPermaLink="true">https://futureisnow.hashnode.dev/i-measured-how-many-tokens-claude-code-cursor-waste-on-saas-plumbing-and-built-4-commands-to-stop-it</guid><category><![CDATA[Next.js]]></category><category><![CDATA[devtools]]></category><category><![CDATA[#ai-tools]]></category><category><![CDATA[ai agents]]></category><category><![CDATA[vibe coding]]></category><category><![CDATA[vibecoding]]></category><category><![CDATA[agentic-coding]]></category><category><![CDATA[tokensaver]]></category><category><![CDATA[TypeScript]]></category><category><![CDATA[Web Development]]></category><category><![CDATA[Artificial Intelligence]]></category><dc:creator><![CDATA[saiteja]]></dc:creator><pubDate>Wed, 04 Nov 2026 06:19:58 GMT</pubDate><enclosure url="https://cdn.hashnode.com/uploads/covers/6aeacc792c0c480068b255ba/1ec8b31d-be6b-4f1c-8b3a-ce2124d0dfe0.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p>Every time I started a new Next.js + Supabase project with Claude Code, Cursor, or Antigravity, I noticed my token usage spiking before I even built a single custom feature.</p>
<p>When I audited what the agent was actually doing during the first 30 minutes of a project, roughly <strong>70% of the token burn</strong> went to two things:</p>
<ol>
<li><strong>Blind Repo Orientation (~20,000 tokens):</strong> Running <code>grep</code> and reading 25–30 files every session just to figure out where routes, auth helpers, and database schemas live.</li>
<li><strong>Rewriting the "Money Loop" (~35,000–50,000 tokens):</strong> Rewriting Supabase auth middleware, credit deduction tables, and Stripe / Dodo Payments webhook verification—and usually crashing locally because I hadn't set up 12 live <code>.env</code> API keys yet.</li>
</ol>
<p>To fix this for my own workflow, I stripped everything down into a 4-command CLI and multi-agent <code>SKILL.md</code> called <strong><a href="https://www.vyomirtha.com/vstacks">vstacks</a></strong>.</p>
<p>Here is how the 4 commands work under the hood.</p>
<h3>1. <code>vstacks new</code> (The Zero-API-Key Money Loop)</h3>
<p>Most starter kits crash on <code>npm run dev</code> until you create live projects on Supabase, Stripe, and OpenAI and paste in a dozen keys. That breaks an AI agent's autonomous test loop.</p>
<pre><code class="language-bash">vstacks new MySaas --payments dodo-payments --yes
</code></pre>
<p>This scaffolds a <strong>Next.js App Router + Supabase + Tailwind</strong> app where the entire money loop (Sign-in, Credits, Payments, AI Chat, Waitlist, <code>/terms</code>, <code>/privacy</code>, <code>sitemap.xml</code>) is controlled by a single <code>src/config.ts</code>.</p>
<p><strong>How it runs with 0 API keys:</strong></p>
<ul>
<li>If <code>.env</code> keys are blank, it automatically falls back to <code>.vstacks/dev-db.json</code> and mock preview sessions.</li>
<li>Default <code>src/app/page.tsx</code> renders an interactive <strong>Dev Control Panel</strong> with live credit counters, a <strong>1-click simulated payment webhook</strong> (<code>POST /api/payments/dev-simulate-webhook</code>) to test upgrading a user to Pro, and a simulated AI chat box.</li>
</ul>
<h3>2. <code>vstacks swap</code> (1-Second Provider Hot-Swapping)</h3>
<p>If you start with Stripe and decide to switch to <strong>Dodo Payments</strong>, <strong>Polar</strong>, or <strong>LemonSqueezy</strong> (or swap AI between OpenAI, Anthropic, Gemini, and OpenRouter), asking an agent to refactor your SDK calls usually leaves broken imports across 10 files.</p>
<pre><code class="language-bash">vstacks swap payments stripe
</code></pre>
<p>Because <code>vstacks</code> isolates external providers behind clean slot facades and ambient TypeScript definitions (<code>src/env.d.ts</code>), <code>vstacks swap</code> swaps the provider in <strong>under 1 second</strong>, updates your orientation map, preserves DB migrations, and passes <code>tsc --noEmit</code> with <strong>0 TypeScript errors</strong> and zero changes to your UI code.</p>
<h3>3. <code>vstacks map</code> (229ms Static Repo Orientation)</h3>
<p>Instead of paying your agent to explore your folder tree on every new prompt:</p>
<pre><code class="language-bash">vstacks map
</code></pre>
<p>This runs a fault-tolerant regex scan across <strong>any JS/TS repo in ~229ms</strong> (even projects not created with <code>vstacks</code>) and outputs <code>.vstacks/MAP.md</code>—a compact 6-section index of your routes, DB tables, active facades, and required env vars. </p>
<p>Combined with <code>SKILL.md</code> (which auto-registers across Claude Code, Cursor, Codex, and Antigravity), your agent knows your entire architecture on Turn 1.</p>
<h3>4. <code>vstacks check</code> (Pre-Ship Guardrails)</h3>
<p>Coding agents love forgetting <code>"use client"</code> on interactive components or importing raw server SDKs inside client files.</p>
<ul>
<li><code>vstacks check --fix</code> auto-prepends missing <code>"use client"</code> directives, syncs <code>.env</code> keys, and blocks raw SDK imports in UI (<code>W_FACADE_BYPASS</code>).</li>
<li><code>vstacks check --ship</code> audits your repo before deployment for leaked live secrets (<code>sk_live_</code>, <code>whsec_</code>), client-side private env access, hardcoded <code>localhost</code> URLs, and unverified webhooks.</li>
</ul>
<h3>See How Many Tokens Your Workflow Is Burning</h3>
<p>👉 <strong><a href="https://www.vyomirtha.com/vstacks">Check out vstacks &amp; the Token Calculator here</a></strong></p>
]]></content:encoded></item></channel></rss>