The problem: 73% of applications never reach humans, 250+ people apply to each job, but 85% of positions are filled through networking.
Our solution: AI agents that find hiring managers for any LinkedIn job, analyze your profile fit, and generate personalized outreach messages. Instead of competing with hundreds in the application pile, you reach decision-makers directly.
Technical bits: LangChain + OpenAI for job parsing, Next.js + Supabase, custom contact discovery algorithms that work across different company structures.
Results: Users getting 70% more response rates vs regular applications.
In a world where AI is automating jobs away, we wanted to build something that helps people actually get jobs.
almost•13h ago
So you've put a effort in to build a product just to make the world slightly worse on net. Not hugely worse, but still it doesn't seem like the best way you could have spent your time.
fhd2•12h ago
1. One from my network, just announced it, someone I had worked with in the past reached out, quick chat, hired, great.
2. One with the usual approach of posting job ads and all that. We got an _insane_ amount of noise, even as an obscure, small company. I've hired hundreds of people, but most of those three years ago or earlier. Never seen such noise, most candidates barely meeting any of the requirements, weird auto generated cover letters and CVs. It was a bit exhausting, but I went through everyone manually to make sure I'm not accidentally filtering out a solid candidate. We found two in the end, one quickly backed out because they got another offer. But there was one good candidate left, and they accepted the offer. I don't remember this being so hard.
A few years ago I'd call people trying to automate screening or mainly hiring from their network lazy. In this time, I see the appeal.
The last thing I need is more bots spamming me on my LinkedIn account on top of all this madness.
notpushkin•12h ago
tough•11h ago
it doesnt really matter where the message or how was created, only that you're 100% ok with its contents reaching its destinatatary.
that still puts some limits to how much you can -spam- and should make you not want to -spam- but only send real messages, but you can certainly use AI to help you research/generate those leads initial messages
i have not tried to do this, but i wouldn't see how its problematic?
dloku•9h ago
tough•4h ago
dloku•9h ago
neilv•12h ago
Hopefully we'll get a big backlash against disingenuous passing off AI-slop as a communication from a human.
Maybe sending someone an AI-slop message will become universally recognized as trashy behavior -- employed only by the corporate communications of companies that really don't care, but not by anyone respectable.
dloku•9h ago