InterSnap
The real work of an interview happens live. The write-up happens after, once the details have faded, and the signal that mattered leaks out in that gap. InterSnap closes it: I capture my read in the moment, and it returns structured feedback in my own voice, sharper every round. I designed and specified the system end to end. Claude Code wrote the code.
The problem it solves
The pattern never changes. While it is happening, I am fully present, reading fit and capability across several dimensions and following up in real time on whatever surfaces. Then I sit down to write it up, and the precise moments are already blurred: the exact phrase the candidate used, the concern that flickered and passed, the strength I meant to weigh. What survives is a generic summary, and the signal I actually collected leaks out in the gap between the interview and the write-up.
I ran 82 interviews before I built this in April. What used to take days now takes under five minutes, and comes back as concise feedback with a follow-up action plan.
What it does
InterSnap moves the capture into the interview itself, so nothing waits for after. Before it starts, I set up the role, the competencies, and the profile of an ideal candidate, then drop in the CV and the HR screening notes, so the tool walks in with the same context I have. During the interview, I score my read on each question as it lands, flag a concern the moment it shows, and add a follow-up without breaking eye contact, in Hebrew or English, whichever is faster in the room. It tracks time so I always know how much room is left.
When the interview ends, it produces the write-up: overall assessment, key strengths, key concerns, a recommendation on a clear scale, and a follow-up action plan for the next round. Every line is built from what I captured live, not from what I could still remember an hour later.
How it learns my voice
A write-up that does not sound like me is useless. InterSnap learns my voice from evidence: real write-ups I have written, one for a candidate I moved forward and one I passed on. It models how I actually write: how direct, how long, what I lead with, what I leave out. The draft comes back in my register, ready to edit.
How it sharpens every round
The second time is closer. That is the whole design. When a draft misses, I fix it in one tap: too formal, too long, wrong tone, missed a strength, missing a concern. Each correction feeds the next round. The corrections are fixed signals, because "make it better" teaches a tool nothing, while "you buried the concern" moves it.
The rest of the choices came from the same place. Notes work in Hebrew or English. The goal was to remove every friction sitting between the interview and the write-up.
How vocabulary shapes thinking.