An AI-powered job application coach for real people
TL;DR
- Problem. The job market is broken. Candidates face automated screeners, generic AI resume tools that hallucinate experience, and a process that rewards volume over truth. Good people get filtered out before a human ever reads their application.
- My role. Solo founder, product designer, and engineer. I designed and shipped the full product end-to-end.
- What changed. I built an evidence-first application coach. It ingests resumes, job descriptions, portfolios, GitHub, and LinkedIn recommendations, then generates tailored resumes and cover letters where every claim traces back to a real source, labeled by strength.
- Result. A working, free production app on Next.js, Postgres, and OpenAI structured outputs, with a Prisma-backed job queue, encrypted credential storage, and an adaptive interview flow. Shipped and in use.
- Why it mattered. It shows what an AI-native product looks like when grounding and evidence are first-class citizens, not afterthoughts, and it is the opposite of the generate-and-pray tools flooding the category.
I built Human, Actually because I got tired of pretending the job market is normal.
It isn't.
I've been feeling increasingly uncertain in my own day job. Unappreciated too. It's not great. But that's for another day. Maybe I'm catastrophizing. Who knows. For the moment at least I'm still employed, so there's that, right?
Whether I'll be joining the millions of jobless folks on LinkedIn soon or not, the bigger point is: uncertainty is everywhere right now, and pretending otherwise is ignorant.
It does not matter whether you work at a FAANG company, a giant retailer, a hot startup, or a tiny local business. In this climate, anyone can find themselves suddenly on the outside looking in. Especially in the United States, where stability can vanish the moment somebody updates a spreadsheet, a headcount target, writes you a nasty performance review or tinkers with an org chart.
So what do we do?
We prepare.
Because the corporate-industrial complex will not take care of us. We'll have to fend for ourselves.
Where it started
That was the starting point for this project. I wanted to apply everything I've learned about product design, AI, systems, interfaces, writing, and workflow design to build something that could actually help me navigate this reality. Then I realized there was no reason to stop with me.
So I built it for everyone.
One could absolutely argue that I "rage-developed" this product.
That would not be entirely inaccurate.
But the deeper reason is this: the hiring process itself has changed, and most people are still approaching it like it's 2018.
It isn't just that AI is reshaping jobs. It's that AI is now sitting directly in the path between you and your next paycheck.
Where a recruiter might once have read your resume, a machine now parses it, scores it, ranks it, filters it, and often rejects it before a human being ever laid eyes on it. Applicant Tracking Systems are no longer some minor back-office utility. They are a front gate. And for a huge number of job seekers, that gate is locked.
That means a lot of rejection now happens automatically. Quietly. Instantly. Inhumanly.
And no, this is not just because recruiters got lazy. Entire recruiting organizations have been cut to the bone. The people still doing that work are buried, overextended, and increasingly forced to rely on automation to survive the volume. So candidates are left trying to impress machines before they are ever allowed to meet a person.
That is the world we live in now.
You hear the same story over and over: people apply to hundreds of jobs, get a few calls, maybe a few interviews, and still end up nowhere. It's exhausting. It's demoralizing. It makes good people question their own value.
So I decided to build something that fights back.
What Human, Actually is
Human, Actually is an AI-powered job search and application copilot built to help real people navigate an increasingly machine-mediated hiring system. It helps you prepare better, present yourself better, understand where you actually stand, and generate stronger materials grounded in your real experience instead of generic AI sludge.
The idea is simple: if machines are going to stand between you and opportunity, then we may as well spin up some machines on our side too.
How it works
At its core, Human, Actually helps you build a much richer, more complete representation of who you are as a candidate than a single static resume ever could.
You start by creating a case around a real opportunity. You add the job description, your resume, and any supporting evidence that helps tell the truth about what you can do: portfolio links, website content, GitHub, notes, writing samples, and other useful context. The system ingests all of that, processes it in the background, and turns it into structured evidence instead of just a pile of text.
From there, Human, Actually analyzes the role against your evidence, matches requirements, identifies strengths and gaps, surfaces missing must-haves, and shows you where your story is strong, weak, partial, or unsupported. It is not just spraying out content and calling it intelligence. It is trying to build a grounded picture of fit.
Then it helps you act on that picture.
That is the whole point.
What it can do
It analyzes job descriptions against your actual evidence.
Human, Actually compares the requirements of a role against the evidence you provide and gives you a structured fit assessment. It highlights where you clearly match, where the match is partial, where something is missing, and where more evidence is needed. It also prioritizes weak areas first so you know what actually deserves attention. It will also be brutally honest with you when a job is NOT a good match for you. After all there is no sense in applying for something someone clearly doesn't have the qualifications for.
It helps you build a stronger evidence base.
A lot of people are more qualified than their resume makes them look. Human, Actually is designed around that problem. It lets you pull in not just your resume, but your website, GitHub, notes, writing samples, interview answers, and even facts extracted from freeform chat. That matters because most careers cannot be accurately represented in one brittle document written for a machine parser.
It guides you through targeted interview-style questions.
When the system detects weak or partial areas, it can ask focused follow-up questions to help uncover the missing context. Instead of leaving you alone with a vague sense that your application is weak, it actively helps you strengthen it. Those answers feed back into the analysis and improve later outputs.
It lets you talk naturally, not just fill out forms.
There is a guided interview flow, but there is also a separate chat experience where you can speak freely about your work, your projects, and your accomplishments. Human, Actually can extract concrete, usable facts from those conversations and turn them into confirmed evidence. In other words, it is not just asking questions. It is listening for what matters.
It generates ATS-aware resumes that are grounded in reality.
This is not about keyword stuffing garbage into a document until it passes a robot vibe check. Human, Actually generates tailored resumes using your actual evidence, requirement coverage, and ranked bullet planning. It also runs QA on the result to catch weak bullets, repetition, unsupported claims, missing metrics, and other issues that make resumes worse instead of better.
It helps reduce obvious age-bias signals.
Whether people want to admit it or not, age bias exists. Human, Actually includes rules that avoid unnecessary age-revealing cues such as graduation dates, explicit duration framing, and stale headline conventions, while still keeping the output honest and strong. That is not gaming the system. That is survival.
It writes better cover letters without sounding like a robot wrote them.
The cover letter flow is built separately from the generic artifact system because cover letters need a different touch. Human, Actually works to avoid common AI markers, supports rewrites, and lets you steer tone toward more direct, warm, concise, or technical. It also uses writing samples and notes for light voice personalization.
It helps with application questions one at a time.
You know those custom application questions every company invents to slow you down? Human, Actually handles those too. You can generate answers per question, regenerate them, steer them, edit them, and save them so the whole thing becomes less of a repetitive soul extraction ritual.
It tells you what to do next.
One of the hardest parts of job hunting is not effort. It is decision paralysis. Human, Actually includes workflow guidance that looks at the state of your case and tells you the next best action, whether that is adding better sources, answering a key question, refreshing the analysis, or reviewing outputs. This sounds small until you are overwhelmed, at which point it becomes very big.
It shows its work.
A lot of AI tools feel like magic tricks. Human, Actually is built to make the process more inspectable. You can see what sources were captured, where evidence came from, whether it was direct or inferred, and where the system still needs help. I care a lot about that. Blind AI confidence is one of the fastest ways to make a useful tool dangerous or useless.
It supports a more complete job search strategy, not just document generation.
There is also a public presence audit flow that helps assess recruiter-facing risk across places like LinkedIn, X, Facebook, and other URLs, with findings and rewrite suggestions. Because yes, what shows up around you online can absolutely shape hiring outcomes too.
Why I made it this way
I did not want to build yet another shiny AI toy that produces polished nonsense.
I wanted to build something practical. Something that respects the fact that job seekers are usually under pressure, low on time, emotionally drained, and trying to operate inside a system that often feels impersonal and rigged against them.
That is why Human, Actually is built around evidence, iteration, analysis, and guided improvement, not just one-click content generation. The point is not to help people fake their way into jobs they cannot do. The point is to help capable people represent themselves more effectively in a hiring landscape that increasingly fails to see them clearly.
And frankly, I also built it because I think a lot of people are getting left behind by this moment, not because they are untalented, but because the rules changed underneath them and nobody bothered to explain the new game.
I also see Human, Actually as a showcase of what I can do and what I have to offer prospective employers. I envisioned it, designed it, built it, secured it, debugged it, iterated on the UX, and deployed it. It is meant to show the full breadth of my skills in one product. It even runs on the full-blown design system I originally created for my own website, ai-created.com.
Why it's free
Human, Actually is free because I do not want to make money off people who are already stressed out, burned out, scared, or suddenly out of a job.
There are plenty of businesses happy to squeeze desperate people for subscriptions, credits, premium tiers, resume rewrites, ATS scores, and other beautifully packaged forms of panic monetization. I'm not interested in doing that. I built this to help people, not to extract money from them at the exact moment they need help most.
The one thing you do have to pay for is your own AI provider. Human, Actually supports OpenAI, Claude, and Google Gemini, and you bring your own API key. I cannot realistically pay for everyone's AI tokens out of my own pocket, especially given that I, too, am living under the warm and soothing glow of modern job uncertainty. But the good news is that for an individual user, raw AI tokens straight from the hose are usually not going to break the bank.
So that's the deal. I'm giving away the product. You cover your own model usage. And if Human, Actually helps you land a job, feel free to send some AI token-money my way.
That feels like a much fairer arrangement. And frankly, a much more human one.
Why the name fits
The hiring process has become increasingly machine-mediated. More automated. More detached. More focused on filters, matching, ranking, and triage.
Somewhere in that process, the actual human being got pushed into the background.
This app is my attempt to push back on that.
Not by pretending the systems do not exist. They do.
Not by telling people to just "be authentic" and hope for the best. That is not enough anymore.
But by giving applicants a way to be more strategic, more complete, more visible, and more effective without losing the truth of who they are.
Because behind every rejected application is not just a PDF. It is a person. A person with experience, effort, strengths, gaps, history, context, and potential. A person who deserves a real shot.
And if the front door is guarded by machines now, then people deserve better tools to get through that door.
That is why I built Human, Actually.
Human, Actually was fully designed, built, and shipped by yours truly. Special thanks to Adam Cobb and Eli Yelluas for the help with penetration testing and security suggestions.