Should AI be perfect? & Current project introduce
Hi everyone! this is like my first post anywhere ever so be gently to me 😃
Little bit about my situation:
I'm working on a project where users can generate personalized CVs and cover letters on my website. Here's how it works in a nutshell:
Users fill out a detailed profile with their information, skills, and experience
LLM (like ChatGPT) creates completely customized CVs and cover letters based on their profile
Each application is tailored specifically for the job they're applying to, highlighting why they're the perfect fit for that particular position
My question for the community: When we use AI to generate CVs and cover letters, what should our expectations be?
Should we expect AI-generated content to be perfect?
Do people assume it's flawless just because AI created it?
Is it realistic to demand perfection from AI, or is it okay to accept that sometimes things don't go perfectly?
Im planning to make possiblity to customize generated products freely, but every feature im planning to add which feeks like had to be add ons , increase the development time and delays the deployment.
Why all this matters:
AI makes mistakes, just like humans do. But there seems to be this expectation that if AI creates something, it should be error-free. I'm curious about your thoughts on finding the right balance between leveraging AI's capabilities while managing realistic expectations.
What do you think? Should AI make everything 100% correct and im just bad at developing, or is it understandable that ai generated things are sometimes badly made.
Would love to hear your experiences and opinions (:
Replies
I tried to also make a post for p/general but it got rejected, so I'm going to describe my product here a bit more.
The main problem is that I cannot trust the AI to be error-free, as it definitely will make mistakes. So I developed the idea of different layers. In the first layer, AI's attempt to create the ready-to-go version of the cover letter or CV.
Since this version might include incorrectly written names or something that doesn't make sense, the AI will take a second look after the first attempt. It is no longer trying to change any key elements of the product, it is only looking for mistakes and trying to validate that the writing makes sense and includes the correct information.
There could also be a third layer. Since my site's idea is to create content that is suitable for the target job role and highlights the user's fit for that specific position, the third layer acts like a job recruiter. It compares the job post and the resume and decides whether it is a good fit for the role, and how well it fits that role.
Does this sound like good way to use LLM or am I trying reinvent a wheel here and there is already better solutions for this problem? :)
Since my last posts, my focus has been mostly on continuous implementation. I have been coding consistently, turning the concepts into a functional system.
The Challenge of Feature Overload and Growing Codebase
As is typical for developers, my mind constantly generates new features and ideas. However, the largest challenge now is not the lack of ideas, but managing the increasing complexity of the growing codebase. Keeping the entire application structure clean and coherent becomes significantly harder as the project expands.
To counteract this, I maintain a narrow focus on two critical objectives:
Core Functionality: The product must be good at its primary purpose: producing highly effective, job-specific, and ATS-optimized documents.
Clear User Journey: The user must be able to complete the detailed profile creation easily. This is the foundation of high-quality output, and the interface must be intuitive.
My goal is to launch the first version (V1.0) this week. To achieve this target, I have made significant cuts: For instance, I dedicated time to building a separate, visually appealing welcome page, but I decided to leave it out of the launch. I prioritize spending my time ensuring that the core functionality is robust and maintainable, rather than perfecting non-essential surrounding elements. My focus is entirely on solving the main problem for the customer.
Maintaining Progress
I am working on this project outside of my full-time job. I use these Product Hunt posts as a progress log and measurement tool. Documenting these milestones helps me visualize my advancement and ensures I maintain strong, consistent momentum towards the launch.
Seeking Advice
For those who have successfully launched a side project under a tight schedule:
What are your best practical strategies for keeping the code easy to work with and managing unexpected issues when you are under pressure to release the first version? How do you decide where to draw the line between "good enough code" and "must be refactored now"?
Any advice on maintaining launch discipline is appreciated.
Long time no see! Got some delays and whats funny, the more I delay development, the more I decide to change the product. Last months I rebranded the product, changed styling, pricing model and added new features.
I realized that if I change something, the code breaks, so I made it better to maintain and easier to change in the future, so every new thing does not break the current model.
I'm pretty close to deploy, but still trying to make it more understandable to new customers.
That's pretty much the current problem: how to make the site so simple that everyone can use it, but still add many features and be as flexible as possible, so the user can do as much as they want to modify the CV and cover letter. If I add a shiny new feature, how does the user understand the benefit it brings or know how to use it.