Why I am building TrySignalHire

Most hiring funnels still start with a resume-shaped filter and ask candidates to keep re-entering the same story. TrySignalHire changes the order: a candidate builds one evidence profile from resume context, projects, writing, recommendations, preferences, and proof.

Companies complete their side too: the company context, open jobs, and role-specific evidence signals they actually want to review. The matching layer compares the two sides instead of asking every candidate to rebuild a new application from scratch.

AI helps structure the material into sourced claims and match signals. App logic scores overlap in a way a person can inspect. The goal is not for AI to make the hiring decision. The goal is to make real evidence visible before someone disappears into the funnel.

AI makes it possible to build almost anything. That is useful, but it also makes it easy to jump from idea to idea without giving one problem enough attention.

I use a simple filter before I commit serious focus to a product idea. I wrote more about that filter in how I choose projects to work on. TrySignalHire had to clear all four questions before I gave it this much attention.

I applied to roughly 40 roles. About half turned into rejections, and the rest mostly turned into silence. Across job portals and repeated application forms, I kept translating the same work into slightly different boxes.

The visible portals included LinkedIn, Wellfound, YC Work at a Startup, and Welcome to the Jungle.

I used AI to help answer the questions because that is how modern candidates work now. At one point a platform told me I was a top 20% candidate. The outcome did not match the label: no offers, no meaningful outreach, and no clear signal that the system understood why I might fit.

I reached out directly and got the kind of generic answer candidates know too well: sorry, hang in there, the hiring companies decide who they respond to, and the platform cannot intervene. I understand the constraint. I do not accept it as the best possible system.

The hardest version came after a role I cared about, with conversations that felt strong from my side. I understand that good conversations and real signal are not the same thing. Still, when you spend time with people in an interview process, you usually walk away with some gut feeling for where you stand. This one ended with a polite rejection from someone I had never spoken with, saying the conversations were positive but there was still not enough signal to move forward.

I do not know exactly why that happened. Maybe the evidence never made it cleanly through the process. Maybe the screens, recorded answers, notes, or internal summaries told a different story than the conversations did. The uncertainty is the point: even a careful human process can still leave the candidate and company without enough shared proof.

TrySignalHire did not start from that rejection. I had already started building it because the problem was obvious. But that moment sharpened it: you can prepare hard, spend real time with people, feel the fit, get shut down, and still have no clear answer for what evidence failed to make it through.

That is the emotional cost hidden behind hiring metrics. The market tells candidates to apply everywhere, so the process starts to feel like throwing darts at a board you cannot see. If a handful of companies respond, that can count as progress. If one gets close and still ends with a thin signal gap, you are left trying to reverse-engineer a decision from almost nothing.

I have also been on the hiring side through Hack Upstate and Careers in Code, including searches for program managers and other roles where the resume was never the whole story.

When a real decision had to be made, the question underneath everything was trust. Do we trust this person to do the work, communicate clearly, learn, handle ambiguity, and be good for the people around them?

I keep coming back to a short clip where Elon Musk describes asking candidates to tell the story of their career and the hard problems they worked through. People who really did the work usually remember the details, the tradeoffs, and the decisions. A truthful story backed by proof creates signal, and signal is what lets trust form before a hiring decision.

The current funnel does not capture that well. It turns candidates into fragments and asks companies to infer the person from thin evidence.

That is also why so many company sites lean on images of happy people smiling. The image is rarely the product. It is a trust signal. I think the next web gets quieter, but the need underneath stays the same: help people believe there is a real person and real evidence behind the page.

I am not trying to reinvent recruiting from the outside. I am trying to step into the matching layer between candidates and companies, where trust and evidence can be established before someone disappears into a funnel.

The first version of Signal OS is meant to sit behind that workflow. It helps a candidate build a reusable evidence-backed profile and helps a company describe the job signals that matter before the match queue starts.

The goal is not to manufacture a persona or let AI hide the source material. The goal is to make real work easier to understand, easier to match to a real job, and harder to lose inside a keyword screen.

For companies, the starting point is a clearer review surface: why this candidate matched, which sources support the match, what still needs human review, and how long they have to respond before the candidate is left waiting.

I am not claiming TrySignalHire should make hiring decisions or replace human judgment. The MVP is still early, and the first job is to make candidate evidence clearer.

The current system is not aligned. Applicants use AI to write and tailor resumes. Companies use AI to screen and rank them. Recent research on AI self-preferencing in algorithmic hiring reports 67% to 82% LLM-vs-human self-preference bias across major models after controlling for content quality.

The paper I am watching is AI Self-preferencing in Algorithmic Hiring, which also reports that candidates using the same LLM as the evaluator were 23% to 60% more likely to be shortlisted than equally qualified candidates submitting human-written resumes in simulated hiring pipelines.

I am also exploring whether some kind of personality or working-style assessment belongs in the product. That needs more thought. If it becomes part of the system, it has to support trust and context without pretending people can be reduced to a simple score.

The practical bet is simpler: hiring needs less noise, more evidence, and a better way to understand the person behind the packet.