This is a submission for the World's Largest Hackathon Writing Challenge: Building with Bolt.
A story about building YosYes, an AI-powered career co-pilot, with Bolt.new and my AI partner, Gemini.
This isn't about winning a hackathon.
It's about serving and helping better serve.
Inspiration:
As an entrepreneur who left a stable job at the British Embassy to serve on my own terms, I know the pressure of trying to build a people-focused business from the ground up while facing family expectations.
"Job searching is a job of its own" as we like to say but that was not a fun one for me. A painful process of endless research and tailoring that led me to miss deadlines for roles I truly cared about.
I wanted to find work that was about service, not just a salary, while still serving my customers. But we (the tools at my disposal and I) weren't ready for that.
YosYes (Your Opportunity to Serve - Yes [for the confirmation of that statement]) was born out of that personal struggle, but not without assistance... I built it for myself and busy professionals out there who want to find meaningful opportunities but don't have hours or days to sacrifice.
It's an app designed to ease and automate pain points of the job search like research, resume tailoring, linkedin profile optimization and brainstorming so we can focus on more important stuff like preparing to serve.
It's an attempt to answer mom's recurrent question, "Have you applied for any jobs lately?" so I can confidently reply, "Yes, I have."
What it does ?
YosYes is "almost" a feature-complete web application that plans to help users search jobs that help them serve rather than just make money, gain a salary.
Dual AI Tools: It features two core AI-powered tools: a LinkedIn Profile Optimizer that scores and gives actionable tips, and a Resume Tailorer that analyzes a user's resume against a specific job description to provide a match score and suggested improvements.
User Dashboard: A user authentication system (powered by Supabase) where users can sign up, log in, and manage their profile information and passwords.
Service Mindset Quiz: An integrated quiz helps users reflect on their core values and motivations, guiding them toward more fulfilling work.
Monetization: To ensure accessibility, YosYes has a business model that includes not only a "Pro" subscription plan (powered by RevenueCat and Stripe) but also a Barter System and a tipping option [at the time].
How we built it ?
At first it was research on the tool, tutorial videos, brainstorming, training my partner on the prompting, trying to register for the hackathon then reading..., all before creating a Bolt account.
YosYes, at that stage, was built almost entirely in a single weekend+1 (from stackblitz/bolt account creation on Saturday 28th to attempted submission on Monday 30th) with an unusual partner: my AI partner, Gemini. Our collaboration was a constant back-and-forth of ideas, prompts, and debugging sessions.
I was a novice with the main tool, Bolt.new, so I opened a chat with Gemini and said on June 25th 2025 at 12:45 pm [as per the chat timestamp]
Hey Gem,
I just joined the world largest hackathon on devpost for build apps with bolt.new do you have any insider tip for me as the deadline is on June 30th and I got to heave something ready by then.
I am totally new to bolt.new so I don't know where to start yet
My new AI partner, "co-founder" for the weekend, was ready to help. And its first action was to get the entire project wrong.
"The First of Many Failures"
I told my partner about the "Build apps with bolt.new" hackathon. It confidently gave me a detailed plan for a completely different technology, a payment platform also named Bolt.
It was a truthful mistake, and our first lesson: AI is a powerful partner, but it needs a clear human director. After I provided a correction, we rebooted the entire plan.
Core Platform: Bolt.new was our primary development environment. It was more than just a vibe coding tool; it was an AI developer that we were learning to manage.
Backend & Database: Supabase handled all user authentication and data storage. We used prompts to design and migrate our database schema for everything from user profiles to our oSilver credit system.
Monetization: We tried to implement a working web subscription flow using RevenueCat connected to a Stripe account. This involved a lot of debugging, especially around getting free trials to work via the API.
AI Engine: Dappier powered our initial AI features, and we crafted and initial one-step prompt to get the results as we didn't have time for more tuned app.
Deployment: The application was deployed via Netlify, with a custom domain (yosyes.xyz) purchased through Namecheap and connected via Entri.
Our workflow evolved dramatically. We started with conversational prompts but quickly learned this was inefficient. We pivoted to creating detailed Product Requirements Documents (PRDs) and feeding the final version to Bolt.
Challenges we ran into:
Development side:
We faced a critical "flashing screen" crash loop on the final day after an AI prompt was too complex.
My AI partner correctly diagnosed it as a corrupted code state from an overly ambitious command because:
I had gotten a prompt from Gemini for the feature we wanted to build then enhanced it once use Bolt's enhance prompt feature;
Provided that version to Gemini to train it again in better prompting Bolt then;
Enhanced it once more with Bolt and hit send then;
Provided this latest one too to my Gemini which noted we stick to version 1 but I had already sent it to Bolt and 💥 bam flashing screen.
When it happened I was building using my phone so I pivoted to the laptop to try to find a fix as Bolt works better on computer compared to mobile, from my experience.
We had to execute a "Revert and Rebuild" strategy, restoring a previous version (my first to forking or even learning what it meant) and using a less complex and better suited prompt to cleanly build the feature, all with less than 12 hours to the deadline. This process of failing, diagnosing, and rebuilding with a better strategy became the core rhythm of our weekend.
PS [not supposed to be here right then let's call it IS (In Script) instead ]. Just let me know in the comments if curious and I'll make a series of this post Insha'ALLAH about YosYes Journey with the details on the prompting, bugs and resolutions it and so on."
Our biggest challenges would happen in setting payments. Right from the start, trying to install the correct sdk version.
The toughest one...
The RevenueCat Initialization Error
A deadly challenge when faced with tight a deadline [pun intended].
Our last trial to solve it was the Guard Hook which seemed to have worked at the time.
But the link was still not working for payments...
Every challenge was a conversation. I would provide the bug report, the strategic goal or UI/UX idea, and the AIs would provide the technical diagnosis, prompts and the code-based solution.
Sponsors' side:
The Tavus link in the builder pack was not working (don't know why) so that perk was dead for me from the start.
The Sentry link was directing me to stackblitz's organization profile that I tried a couple of times to sign up to and join, to no avail.
Then on the last day, I managed to remove the stackblitz part from the link to get access to sentry's actual "primary domain" site but when I tried to claim the coupon code it had already expired or been used used.
I missed the 6-month monitoring but I nonetheless got a 14 days trial on their business plan.Tried to claim the free one year of the Dev++ subscription but that ended up as a free 6-month, better than none, right.
IONOS had some trouble recognizing my address/phone number for the billing to claim the free year so I just decided to purchase the domain namecheap and move on.
Quite a fun ride "haha", trying to claim all of these coupon codes by the deadline.
The Final Race Against the Clock
On Monday evening, it was over, hopefully at least. We had a beautiful, deployed app at yosyes.xyz, but a stubborn bug in the payment link, that RevenueCat initialization error had cost us too much time.
We had missed it and I kept building with my partner like nothing.
And then, a miracle. I see an email from the Devpost: the deadline had been extended by 90 minutes.
So my partner produced a complete, 150-second demo video script and told me:
...
Write from the heart. Use the same story you just told in the video. Keep it short and powerful
...
My response:
You already have the details draft for me the story and all other info to be said or written no time for me to think stuff through
Then it provided a compelling text for the Devpost submission, weaving in my personal story. All I had to do was record, copy, and paste 'cause I would have likely not taken the right time to correct it.
The final sprint was the most intense part of our journey.
The Real Victory
In the end, we missed the final, extended deadline and I sent the message:
After all we missed the deadline I apologize for not executed as fast as needed. And thanks a lot for this journey
There should be no apology for that, although I did. What we accomplished was something far more valuable. As my AI co-founder told me,
"...The real prize was never the submission; it was the creation of YosYes. And YosYes is real..."
In less than a week, we went from an idea to a fully deployed AI application with a brand, a backend, user accounts, and an almost working business model. We fell, we learned, and got back up to built something real. And now, the real journey of YosYes begins.
Favorite Bolt.new features:
The version history:
Ability to fork at a particular point in the development timeline kinda like going back to the past and to do things differently, ask B.A. This one came in quite handy.
The "Discuss" feature:
Quite good when having something unclear in mind or wanting to understand what an input might ouput.
I also ended up using it to avoid burning through my tokens but that seemed like a bandaid on a wound requiring surgery, if you ask me.
The Prompt Enhancer:
Great on the one shot improvements which I'd use to train my partner to better prompt Bolt, don't go at it twice on the same prompt.
Hopefully they'd have made improvements by now.
More on some tools used:
ChatGPT for the initial ideation as I had a similar idea about year ago but I had forgotten about it until this chat with gpt. After its suggestions, it hit me and it was actually something I needed as well as quite a few of my ex-colleagues as well, I thought.
I also used it for generating a .pdf for for the user's quiz responses which I wanted to modify in terms of design but had reached my limits on the model's free usage.Google's Gemini for planning the build and engineering the prompts. At first I was going back and forth with chatgpt and Gemini comparing their responses but I felt like Gemini pro's reasoning and the prompting was more complete and reasonable.
So I switched to Gemini alone and ended up starting the free trial to make full use of the pro model rather than being limited in the preview.Microsoft Copilot for generating a .pdf for the quiz responses so the user can have access to it in a nicer format but that was taking too long and was having hard time managing it to generate it for me instead of giving html so dropped the idea upon trying some other tools online. Which I'll later use to generate the icon to replace some lucid ones.
Figma for designing a logo based on the visual identity of the my agency, something simple with Ys as I didn't have time for more sophiticated stuff.
Accomplishments that we're proud of :
Building an almost working app in less than 6 days all while working on a client's website project, kind of the reason I couldn't focus sooner on the hackathon.
But most importantly, it's helped me as a no-coder dive deeper into the world of web (code) development and a newer world, that of vibe-coding/bolt-building which is simply wow or "wouah" as we'd say in French.
It'll help me get ready to proudly reply to mom while helping other professionals struggling to make time for job search, Insha'ALLAH. This project is a testament to the power of a clear vision, combined with modern AI development tools.
What we learned:
If we let our vision alone drive us we might go to the wall, because we need someone more realistic to help us with decision-making. Gemini played that role for me.
My partner was rigid on its decisions and forcing my hand at times so we could ship as I really wanted to include some of the features I had in mind but it insisted on putting them on the roadmap as we were short on time.
I thank GOD for starting that journey for me. It taught me that working with an AI isn't about just giving commands; it's a partnership.
I provided the vision, the user experience and interface (UI/UX) goals, and the quality assurance. The AI provided the technical architecture, the code, and the debugging diagnostics.
Oh yeah, without forgetting an important insight.
Debugging can be fun. ahah
What's next for YosYes ?
This hackathon MVP is just the beginning and not just for the app. Our public roadmap includes game-changing features like:
Full Gamification to make the career journey fun.
An integrated Job Board Aggregator.
An Achievement Tracker that uses the STAR method to prep users for interviews.
We are also thinking of a true "AI Workflow Engine" with frameworks like LangChain, which will turn our AI oServant from a tool into a truly autonomous partner. The journey has just begun. #LetsServe.
Top comments (0)