<?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: Lucas Chabrera Querol</title>
    <description>The latest articles on DEV Community by Lucas Chabrera Querol (@info_vertex).</description>
    <link>https://dev.to/info_vertex</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%2F3334376%2F18a6b0df-7c4a-4dd9-b207-d5ec416b184d.jpg</url>
      <title>DEV Community: Lucas Chabrera Querol</title>
      <link>https://dev.to/info_vertex</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/info_vertex"/>
    <language>en</language>
    <item>
      <title>When AI Gets Stuck in a Loop: A CSS Nightmare and Lessons Learned</title>
      <dc:creator>Lucas Chabrera Querol</dc:creator>
      <pubDate>Fri, 18 Jul 2025 10:19:39 +0000</pubDate>
      <link>https://dev.to/info_vertex/when-ai-gets-stuck-in-a-loop-a-css-nightmare-and-lessons-learned-3mn6</link>
      <guid>https://dev.to/info_vertex/when-ai-gets-stuck-in-a-loop-a-css-nightmare-and-lessons-learned-3mn6</guid>
      <description>&lt;p&gt;Ever had an AI model tell you "Now I see the problem!" for three days straight? Let me tell you about my recent adventure with Aura, my financial analysis app, and how a simple CSS centering issue turned into an unexpected lesson about AI limitations.&lt;br&gt;
The Initial Problem&lt;br&gt;
I needed to center content when a sidebar was expanded. Sounds simple, right?&lt;br&gt;
The problematic initial code:&lt;br&gt;
.chat-container.sidebar-expanded {&lt;br&gt;
margin-left: 280px;&lt;br&gt;
width: calc(100vw - 280px);&lt;br&gt;
}&lt;br&gt;
The AI's Descent into Complexity&lt;br&gt;
Day after day, the model (Claude-3.5-Sonnet) kept suggesting increasingly complex solutions:&lt;br&gt;
The AI's spiral into complexity:&lt;br&gt;
const calculateSidebarOffset = () =&amp;gt; {&lt;br&gt;
const viewportWidth = window.innerWidth;&lt;br&gt;
const sidebarWidth = 280px;&lt;br&gt;
const contentMaxWidth = 800;&lt;br&gt;
const availableSpace = viewportWidth - sidebarWidth;&lt;br&gt;
const optimalPadding = Math.max(20, availableSpace * 0.05);&lt;br&gt;
// This went on and on...&lt;br&gt;
}&lt;br&gt;
Every time I asked for help, the response was: "Ah, NOW I see the problem!" Spoiler alert: it didn't.&lt;br&gt;
The Three-Day Battle&lt;br&gt;
Day 1: Simple calculations&lt;br&gt;
.content {&lt;br&gt;
margin-left: calc((100vw - 280px - 800px) / 2);&lt;br&gt;
}&lt;br&gt;
Day 2: Complex JavaScript solutions&lt;br&gt;
const centerContent = () =&amp;gt; {&lt;br&gt;
const sidebar = document.querySelector('.sidebar');&lt;br&gt;
const content = document.querySelector('.content');&lt;br&gt;
// Pages of calculations followed&lt;br&gt;
}&lt;br&gt;
Day 3: Desperate measures&lt;br&gt;
useEffect(() =&amp;gt; {&lt;br&gt;
const resizeObserver = new ResizeObserver(() =&amp;gt; {&lt;br&gt;
recalculateEverything(); // Pure desperation&lt;br&gt;
});&lt;br&gt;
}, []);&lt;br&gt;
The Breaking Point&lt;br&gt;
Would you risk changing your AI model mid-project to solve something like this? The reality is that it couldn't center the content because it failed to properly calculate screen distances with an open sidebar. Even after:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Providing full context&lt;/li&gt;
&lt;li&gt;Restarting CURSOR&lt;/li&gt;
&lt;li&gt;Waiting days&lt;/li&gt;
&lt;li&gt;&lt;p&gt;Multiple approaches&lt;br&gt;
The "Solution"&lt;br&gt;
The AI finally suggested hiding the sidebar and sold it as a brilliant UX decision. Here's what actually worked:&lt;br&gt;
The actual working solution:&lt;br&gt;
.sidebar {&lt;br&gt;
position: fixed;&lt;br&gt;
width: 280px;&lt;br&gt;
transform: translateX(-100%);&lt;br&gt;
transition: transform 0.3s ease;&lt;br&gt;
}&lt;br&gt;
.content {&lt;br&gt;
width: 100%;&lt;br&gt;
display: flex;&lt;br&gt;
justify-content: center;&lt;br&gt;
align-items: flex-start;&lt;br&gt;
}&lt;br&gt;
.messages-container {&lt;br&gt;
width: 100%;&lt;br&gt;
max-width: 600px;&lt;br&gt;
margin: 0 auto;&lt;br&gt;
}&lt;br&gt;
1 Technical Improvements Made&lt;br&gt;
New Overlay Interface&lt;br&gt;
ChatGPT/Gemini-style sidebar&lt;br&gt;
Mobile-first approach&lt;br&gt;
Finally centered content (after much suffering)&lt;br&gt;
2 Backend Enhancements&lt;br&gt;
Gemini 1.5 Flash integration&lt;br&gt;
Robust anti-cache system&lt;br&gt;
Better Excel/PDF handling&lt;br&gt;
3 UX Optimizations&lt;br&gt;
Intuitive suggestions&lt;br&gt;
Smooth animations&lt;br&gt;
Enhanced visual feedback&lt;br&gt;
Lessons Learned About AI in Development&lt;br&gt;
1 AI Loops Are Real&lt;br&gt;
Models can get stuck in solution patterns&lt;br&gt;
More complex ≠ better&lt;br&gt;
Sometimes you need to step back&lt;br&gt;
2 Trust Your Instincts&lt;br&gt;
If a solution feels too complex, it probably is&lt;br&gt;
AI can be convincing even when wrong&lt;br&gt;
Simple solutions often work better&lt;br&gt;
3 Model Limitations&lt;br&gt;
Great for general problems&lt;br&gt;
Can struggle with spatial/visual logic&lt;br&gt;
May overcomplicate simple tasks&lt;br&gt;
The Path Forward&lt;br&gt;
I'm keeping the app in beta while I:&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;Gather more user feedback&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;Perfect the core features&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;Plan for pay-per-use model (no monthly subscriptions)&lt;br&gt;
Technical Stack and Tools&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;React + TypeScript&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;CSS Modules&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;Gemini 1.5 API&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;CURSOR (AI coding assistant)&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;Cloud Run + Google Cloud Storage&lt;br&gt;
Key Takeaways&lt;/p&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;ol&gt;
&lt;li&gt;AI models can be incredibly helpful but have clear limitations&lt;/li&gt;
&lt;li&gt;Don't let AI convince you that complex solutions are better&lt;/li&gt;
&lt;li&gt;Trust your developer instincts&lt;/li&gt;
&lt;li&gt;Sometimes the "wrong" solution (hiding the sidebar) leads to better UX
What's Next&lt;/li&gt;
&lt;/ol&gt;

&lt;ul&gt;
&lt;li&gt;More analysis models&lt;/li&gt;
&lt;li&gt;Additional file format support&lt;/li&gt;
&lt;li&gt;Premium features&lt;/li&gt;
&lt;li&gt;Developer API
Would love to hear your experiences with AI in development. Have you ever been stuck in an AI loop? How did you handle it?&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Check it out: &lt;a href="https://aura.vertex-flow.com" rel="noopener noreferrer"&gt;https://aura.vertex-flow.com&lt;/a&gt;&lt;/p&gt;

&lt;h1&gt;
  
  
  webdev #ai #programming #debugging #ux
&lt;/h1&gt;

</description>
    </item>
    <item>
      <title>Aura Financial: AI ethical tool to analyze Excel &amp; PDFs for make better data-driven decisions.</title>
      <dc:creator>Lucas Chabrera Querol</dc:creator>
      <pubDate>Tue, 08 Jul 2025 09:48:13 +0000</pubDate>
      <link>https://dev.to/info_vertex/aura-financial-ai-ethical-tool-to-analyze-excel-pdfs-for-make-better-data-driven-decisions-2ibc</link>
      <guid>https://dev.to/info_vertex/aura-financial-ai-ethical-tool-to-analyze-excel-pdfs-for-make-better-data-driven-decisions-2ibc</guid>
      <description>&lt;p&gt;&lt;em&gt;This post is my submission for &lt;a href="https://dev.to/deved/build-apps-with-google-ai-studio"&gt;DEV Education Track: Build Apps with Google AI Studio&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;This is AURA, a new (and ethical) app for small companies and freelancers who want discover new scenarios for their excels and PDF. You will discover new decisions based in the data that you have in your Excel or PDF files. Developed, thought, designed and every detail that I have made was did with these tools:&lt;/p&gt;

&lt;p&gt;AI Models Evolution:&lt;br&gt;
Started with Claude Sonnet 3.5 - For initial development and architecture decisions&lt;br&gt;
Implemented Google Gemini Pro - As the core financial analysis engine in the backend proxy&lt;br&gt;
Used Claude Sonnet 4 - For advanced debugging, UX improvements, and code optimization&lt;br&gt;
Multi-model approach - Leveraged different AI strengths: Claude for development, Gemini for financial domain expertise, ChatGPT, Google AI Studio and of course, with Cursor PRO.&lt;/p&gt;

&lt;p&gt;What it does:&lt;/p&gt;

&lt;p&gt;Upload Excel, CSV or PDF → get instant analysis.&lt;/p&gt;

&lt;p&gt;Ask questions in plain English: “Where am I overspending?”, “What if I raise prices 10%?”&lt;/p&gt;

&lt;p&gt;Generates automatic insights, projections &amp;amp; what-if scenarios.&lt;br&gt;
Why? Financial analysis often costs thousands €, out of reach for small businesses. Aura wants to democratize it.&lt;br&gt;
And for me, this is the most important part and every developer should be consider take part of it: Ethics first:&lt;/p&gt;

&lt;p&gt;No storage, local processing.&lt;br&gt;
Not available for defense companies or countries violating human rights.&lt;/p&gt;

&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2F2f082wo77blsuo6ndf64.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2F2f082wo77blsuo6ndf64.png" alt="AURA APP" width="800" height="362"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  My Experience
&lt;/h2&gt;

&lt;p&gt;What I Learned:&lt;/p&gt;

&lt;p&gt;User feedback beats assumptions - I initially made UX recommendations without real data, highlighting the importance of user testing over theoretical "best practices".&lt;br&gt;
Incremental improvements matter - Small fixes like scroll behavior and animation polish significantly impact user experience.&lt;br&gt;
Cache management is critical - Proper file naming/hashing prevents deployment headaches and ensures users get updates.&lt;br&gt;
Simplicity often wins - The overly complex "glassmorphism" thinking animation was confusing; clean, simple gradients worked better&lt;br&gt;
What Was Surprising:&lt;br&gt;
How much polish details matter - Users notice subtle things like scroll appearing during file uploads or thinking animations not showing properly&lt;br&gt;
My own inconsistency - I contradicted my previous advice about UI patterns without solid justification, showing the importance of data-driven decisions.&lt;br&gt;
Deployment complexity - Managing Google Cloud Storage, cache busting, and manual uploads requires more attention than expected&lt;br&gt;
The iterative nature - We went through multiple animation versions before finding the right balance.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Technical Insights&lt;/strong&gt;:&lt;br&gt;
React useEffect dependencies matter for performance (scroll behavior)&lt;br&gt;
CSS animation complexity doesn't equal better UX&lt;br&gt;
Build pipeline optimization helps with cache invalidation&lt;br&gt;
Error handling and loading states are crucial for production readiness&lt;br&gt;
&lt;strong&gt;Process Learnings&lt;/strong&gt;:&lt;br&gt;
Beta-first approach is smart - get real user feedback before building advanced features&lt;br&gt;
Version control and backups prevent lost work during rapid iterations&lt;br&gt;
Clear communication about deployment processes saves time and confusion&lt;br&gt;
Bottom line: Building a polished product requires attention to detail, user-centered thinking, and willingness to iterate based on real feedback rather than assumptions.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;AURA BETA V2&lt;/strong&gt;:&lt;br&gt;
&lt;a href="https://storage.googleapis.com/aura-financial-app/aura-complete.html" rel="noopener noreferrer"&gt;https://storage.googleapis.com/aura-financial-app/aura-complete.html&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Feedback, feature ideas or questions are very welcome!&lt;br&gt;
✉️ &lt;a href="mailto:info@vertex-flow.com"&gt;info@vertex-flow.com&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Thanks for reading! 🚀&lt;/p&gt;

</description>
      <category>deved</category>
      <category>learngoogleaistudio</category>
      <category>ai</category>
      <category>gemini</category>
    </item>
    <item>
      <title>Aura Financial: AI tool to analyze Excel &amp; PDFs for SMEs and freelancers</title>
      <dc:creator>Lucas Chabrera Querol</dc:creator>
      <pubDate>Tue, 08 Jul 2025 08:21:35 +0000</pubDate>
      <link>https://dev.to/info_vertex/aura-financial-ai-tool-to-analyze-excel-pdfs-for-smes-and-freelancers-lf6</link>
      <guid>https://dev.to/info_vertex/aura-financial-ai-tool-to-analyze-excel-pdfs-for-smes-and-freelancers-lf6</guid>
      <description>&lt;p&gt;Hey everyone! &lt;/p&gt;

&lt;p&gt;I'm Lucas from Vertex (Spain) and I'm building Aura Financial, a web app to help small businesses, freelancers and startups make better data-driven decisions — even without finance expertise.&lt;/p&gt;

&lt;p&gt;✨ What it does:&lt;/p&gt;

&lt;p&gt;Upload Excel, CSV or PDF → get instant analysis.&lt;/p&gt;

&lt;p&gt;Ask questions in plain English: “Where am I overspending?”, “What if I raise prices 10%?”&lt;/p&gt;

&lt;p&gt;Generates automatic insights, projections &amp;amp; what-if scenarios.&lt;/p&gt;

&lt;p&gt;💡 Why? Financial analysis often costs thousands €, out of reach for small businesses. Aura wants to democratize it.&lt;/p&gt;

&lt;p&gt;Free Beta v2&lt;br&gt;
&lt;a href="https://storage.googleapis.com/aura-financial-app/aura-complete.html" rel="noopener noreferrer"&gt;https://storage.googleapis.com/aura-financial-app/aura-complete.html&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Ethics first:&lt;/p&gt;

&lt;p&gt;No storage, local processing.&lt;/p&gt;

&lt;p&gt;Not available for defense companies or countries violating human rights.&lt;/p&gt;

&lt;p&gt;Feedback, feature ideas or questions are very welcome!&lt;br&gt;
✉️ &lt;a href="mailto:info@vertex-flow.com"&gt;info@vertex-flow.com&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Thanks for reading! 🚀&lt;/p&gt;

</description>
      <category>showdev</category>
      <category>buildinpublic</category>
      <category>webdev</category>
      <category>programming</category>
    </item>
  </channel>
</rss>
