| Hi, I have recently been applying for summer internships and got frustrated when tailoring my resumes in Word. I started learning Python last autumn, but had absolutely zero experience with web development or deploying something to the front/backend. I wanted to experiment with the new coding agents to build a resume editor that would make my application process less painful. Here it is: www.tailortojob.app How I built it:
A friend helped me set up the initial infrastructure because I struggled to connect everything with Claude alone (newer models might be more helpful; that was 4 months ago). The stack is Vercel, Render, and Supabase. Once the front- and backend were set up, I used agents iteratively over the last 4 months to build the actual application (I have probably spent on average about 10 hours per week, so a total of about 180-200 hours) My Agent Workflow:
High/Low Model Split: What worked best was using larger models (Opus 4.6 has been a game-changer) to brainstorm an implementation plan in Markdown, and then handing that off to smaller models to do the actual coding. I felt this gave me a good balance between performance and costs. I recently started exploring Claude's planning mode, which I am considering using instead of my old approach. UI Struggles: While Claude’s newer UI skills are great for quickly building a frontend, I found it often struggled with perfectly aligning hover overlays with other elements on the page (maybe my prompts were not precise enough; it often took me a couple of tries to get it right) Knowledge Cutoffs: When asking agents to implement new LLM APIs (like GPT 5.1 mini), they would often tell me they did not exist, and I should use models like 4o. This was especially annoying when I asked Claude to find errors in my code. It would often tell me that I am using models that do not exist. I learned I had to manually feed Claude the new API documentation during the session to get it working. My Biggest Lesson: The "Tutorial Trap"
Because AI makes it so easy to simply build whatever you can think of, I believe prioritization becomes extremely important. Case in point: I spent 5 hours the night before launching to my friends, building a detailed interactive tutorial. In the end, not a single person used it. Several friends even told me they could not figure out a feature, knew the tutorial existed, and still chose to just give up rather than watch it. I was really surprised. My approach definitely was not perfect, but it was impressive to see what is possible nowadays with hardly any prior coding experience. I would love for you to try out the editor and let me know how I can make it more useful. I am also very curious to hear tips from this community on how I can refine my agent workflows. Thank you for your time and help! |