It started because
I was applying for a PhD.

I was applying for a PhD. The postings were everywhere and nowhere, scattered across department pages, institutional repositories, semi-public mailing lists, and conference newsletters that arrive in your inbox once a year and otherwise gather dust. Some sites I visited every other day. Some I'd discover by accident, weeks late, after the deadline had already passed.

The tools I was supposed to use weren't really tools. They were keyword searches dressed up in different interfaces, all asking the same shallow question: do your words match our words? When the words matched, I got listings I'd already seen, or listings that had nothing to do with the kind of work I actually wanted to do. When they didn't, I got nothing, even though the role I'd have loved was, somewhere, sitting on a server with a slightly different vocabulary.

One evening, somewhat in despair, I copied my CV into a chat window and pasted in five postings I'd been staring at all week. I asked the model, gently, the way you'd ask a colleague, to read each one and tell me which one suited me, and why.

It read them. It actually read them. It told me which posting was best, and what the second-best one was missing, and why the one I'd thought was perfect was probably going to make me miserable. The reasoning was specific to the things on my CV, including a short, oddly-phrased line about the kind of group I'd thrived in, that no keyword filter could have surfaced. I sat there for a while.

The trouble with miracles

The next morning I tried to make it part of my routine. Every day, take the postings I'd found, paste them in, get the verdict. It worked beautifully. It also cost enough that, for a person who was living on a graduate stipend, it could not be a daily habit. I let it lapse. The miracle had a price tag.

What stayed with me, after the application deadlines passed, was the realization that this experience, sitting with a tool that had taken my background seriously and reasoned about a job alongside me, was going to be the next normal. It just wasn't, yet, a thing anybody could afford to have continuously, at scale, for free.

That, I thought, seemed worth fixing.

What it took to make it cheap

Asking a large language model to read every new posting against every user's CV would be wonderful and completely unaffordable. So we built around the expensive part instead of through it.

Suitch reads in layers. The first pass is broad and instantaneous, setting aside the postings that simply don't fit your hard requirements. The second pass is where things get interesting: instead of matching words to words, it reads for meaning, surfacing roles that fit the shape of your career even when their job titles are unfamiliar. Only the final pass is the slow, careful one, a thoughtful read that produces the single sentence you'll actually see. By the time we get there, only a handful of candidates remain, and the work that used to be a luxury becomes something you can have every morning.

The result is something that, two years ago, you would have had to build for yourself, by hand, every evening. Today, it shows up in your inbox. That's the entire point.

Where it goes from here

We are deliberately a small operation. The matching itself will always be free; a paid plan will sit alongside it for users who want more. Either way, the only person paying us is the person looking for work. No employer or recruiter can buy their way into your inbox. The math we've built makes that sustainable; the principle we've taken on makes it necessary.

If any of this sounds like the kind of search you'd like to be on the receiving end of, write to us at [email protected], or just try it.