The talent graph for the AI era.
AI broke the résumé. LatchLock replaces it with months of real conversation and a real project shipped — and hands a company the one person who can actually do the job. Below: the thesis, the moat, the numbers, the team, and the honest answer to every objection.
AI broke hiring on both sides.
- 20 applications today
- 500 by the end of the cycle
- 0 replies, and counting
500 applications to land one offer. Every form re-types the résumé. The kid running AI cheating tools through the interview got three callbacks. Honesty is punished. The best kid doesn't get seen.
- 1,000 résumés this week
- ~80% obviously AI-written
- 1 hire you're not sure about
Ten hours of sorting, no way to tell who's real. Fall back on pedigree and referrals — both miss the kid who shipped real work without the Stanford line. The best company and the best kid never find each other.
Hiring is a cheating contest now. Everyone loses.
Text us once. Never apply again.
LinkedIn, Handshake, Indeed, every career portal — they all still put a form between a student and a job. Better filters, cleaner UI, same 500 postings to chase. We take the form off the table.
One text in, and every company that comes to us is already considering the student. Companies don't post jobs — they tell us what they need, and we hand back the person whose work we've actually watched. The self-written résumé gets replaced by a multi-week conversation with an agent who knows the student, sees their work, and steers what they build toward what real companies are hiring for right now.
Two pieces of evidence. Both hard to fake.
The résumé, the 30-minute interview, the referral — all collapsed under AI, or always biased to begin with. We replace them with two pieces of evidence that compound on each other.
A conversation no résumé ever asks for.
The bot talks to a student the way a good mentor would — curious, warm, multi-turn. Weeks of conversation, not a 20-minute interview. No corporate voice. No "tell me about yourself."
Lying is self-defeating — fake your interests and we match you to jobs you'll hate. Honesty pays the other direction: students who are straight with us from day one build a track record over years, and when companies come asking, those are the names the bot surfaces first.
A real thing they actually shipped.
A week-long spec'd build tied to stated interest — and scoped from what latch has watched real companies actually ask for. It isn't a toy assignment; it's the kind of work someone is hiring for this quarter. AI assist is welcome — shipping with AI is how real work gets done now. We care that they shaped it, debugged it, shipped it.
A student who doesn't want the job will not spend a week on a real project. The cost is the filter.
Do they want it. Can they do it.
No traditional process captures both at any depth — let alone for every student we've ever talked to.
100× the reach. Work we've already watched.
The recruiter saw 5% of the funnel. Signal per candidate: one AI-polished page they didn't write.
The company texts latch what they need, and it returns the best-fit candidate automatically — with a shipped project to look at, and a student we already know ACTUALLY WANTS THE JOB.
The company didn't interview a stranger. They interviewed someone whose actual work they had already seen.
Hiring is chapter one.
Under every product we ship is a live map of which college students are good at what — who's driven right now, who's ready this semester, what they've actually built. We call it the talent graph. Every direction below only becomes possible once that asset exists.
A company like Google texts us: "We need five new-grad ML engineers who've actually trained models, not just taken the class." We come back with five names whose work we've watched ship for a year.
No application portal. No filtering ten thousand AI-polished résumés. No bright kid missed because nobody told them the deadline was Friday.
A freshman texts latch in September. By senior year they've shipped a dozen projects we watched them build, taken three summer roles sourced through us, and never applied to a job the traditional way.
latch is simply how work works for them. A steady conversation across four years — not a panic in March of senior year.
A student texts with an idea. latch pulls an engineer from one school and a designer from another — both vetted through us, both free that semester. A team forms inside a week.
That team starts a company. "We met through latch" becomes a line in the origin story, the way "we met at YC" is today.
2,000 students onboarded in tech. 50 startups texting us for hires. The pool gets real.
100,000+ students across tech, design, data, finance. First top-tier companies sourcing from us. First startup born from a latch match.
LatchLock is where ambitious kids go to find work, and where serious companies go to find them. The graph is the asset; everything else gets built on top of it.
Both sides pay — only when it works.
Free. Until they land.
- · Onboarding, conversations, projects — all free.
- · Success fee only if we place them. Scales with the role — a few hundred for a modest internship, up to a couple thousand for a career-changing offer.
- · Deducted from the signing bonus, so nothing out of pocket.
- · No subscription. No gated features. Nothing if we don't place them.
Pay for access. Not listings.
- · Free during early access. Revenue ramps in year two.
- · Subscription for search access, or ~15–20% of first-year salary when we make the hire happen.
- · The shortcut today: a company texts us a role, we hand back one person. No job post required.
We make money on every placement — from the very first one.
Both sides. From day one.
No student-only warm-up phase. No company-only warm-up phase. We onboard students and talk to companies in parallel — and the two cofounder hires we're closing next are the reason we can.
Onboarding starts now.
- · Rice first, then every school the growth cofounder can reach.
- · Bot is live. Every conversation tells us something no résumé could.
- · Projects shipping from week one. The pool gets real fast.
Conversations start now too.
- · First 20 startups lined up by the startup-relationships cofounder — warm, hand-sourced, no job post required.
- · Early hires are hand-matched. We'd rather place 10 people perfectly than 100 on autopilot.
- · Every placement is the next company's reason to trust us.
Beyond year one: raise seed, expand past tech into design and data, push toward 10,000 students and 200 companies. The long-term picture — a live talent graph the rest of the economy ends up building on top of — stays the same.
Just Manu. Two seats open.
Manu Prabakaran
ML engineer with research and production work across computer vision, interpretability, and AI safety.
Has LIVED the problem. Still in college, still grinding through the broken internship cycle right now. LatchLock is the tool Manu wanted — built by someone inside the pain, not watching it from the outside.
Growth · student side.
Someone who's already grown a consumer product college students actually use — through organic, content, campus chapters, and modern paid.
Owns the student side: taste, brand, referrals, the way LatchLock sounds on TikTok and in a group chat. Cofounder equity.
Startup · relationships.
Deep in the seed-to-Series-A founder network — ~100 early-stage startups they can text and get a same-day reply from. Startups are our first customer, and this is the person who brings them in.
Owns the demand side: which startups we talk to, what they'll pay for, how every placement gets sold in. Bigger enterprises come later, through hires we'll make once we've earned them. Cofounder equity.
Advisors, early engineers, and designers come after the pre-seed closes. We hire the same way we ask companies to hire — conversation, then shipped work.
The honest answers.
The eight questions a thoughtful investor asks first. Where an objection has real merit, we say so.
01 Won't students just cheat with AI on the bot? +
The incentive runs the other way. Lie to the bot and we match you to jobs you don't want. Fake "I love consulting" and you end up in consulting interviews you'll fail and hate.
And even if someone games the conversation perfectly, they still have to ship a week-long project. No amount of AI assist gets you through a full build in a field you don't actually care about.
02 LinkedIn or Handshake will just ship this. +
LinkedIn can't. Shipping LatchLock tells a billion users "your LinkedIn profile is obsolete — go text a bot instead." That's self-cannibalization of the product their entire business sits on.
Handshake would have to walk away from $200M+ of existing job-listings revenue to pivot to our model. They won't. They'll ship a half-measure — an AI assistant bolted onto their existing profile flow — and it'll be worse, because the whole point is starting from a text, not from a filled-out form.
03 What's the 10-year moat? +
The talent graph itself. Once we've watched 100,000+ college students actually ship things across four years, we have direct evidence of which young people produce real work — and a competitor doesn't.
A competitor showing up in year three has none of that history. They can copy the product surface. They can't copy years of watching the same person ship, adjust, and ship again. And no student can fake four years of consistent behavior — a track record across time is the one thing AI can't generate.
04 How do you get both sides on day one? +
Both sides, in parallel, from day one. Two cofounder hires are the unlock — one owns the student side (growth, brand, referrals), the other owns companies (warm outreach, relationships, the first 20 customers). Each side gets a full-time founder pushing on it.
We don't need a huge pool before talking to companies. We need a real one. Ten students we've watched ship genuine work is worth more to a company than ten thousand résumés — and we can surface those ten inside month one. Every early placement becomes the next company's reason to trust us, and the next student's reason to sign up.
05 Isn't there regulatory risk with student-paid fees? +
Yes, and we take it seriously. We don't collect a dollar from students until legal review closes.
The structure is simpler than an Income Share Agreement — no percent of future income, no multi-year obligation, just a one-time fee for a specific outcome we delivered. Traditional recruiters have charged candidates directly for decades with proper disclosure, so the precedent is there. If legal review surfaces real problems, we switch to company-paid only. The business works either way.
06 Two-sided marketplaces fail all the time. +
They do. Three things make this one less painful than usual.
First, the student side has standalone value — career clarity, real conversations, a bot worth texting — so students stay even while the company side is still small. Second, we only need about 100 companies to trust us for this to work, and company trust is a network problem, not a scale problem. Third, splitting the founding team across both sides means neither side ever waits for the other.
07 Why Manu, and why now? +
Manu is a Rice student and ML engineer, with research across computer vision, interpretability, and AI safety. More importantly, he has LIVED the problem — he's a college student grinding through the broken internship cycle right now. He isn't building from theory.
The bet isn't "Manu alone solves this." It's Manu plus two cofounders — one on student growth, one on startup relationships — plus the early team that follows. Who we hire next matters as much as who's already here, and we're being honest about that.
08 What's the concrete 12-month milestone? +
- 2,000 students onboarded · ≥70% still engaged after week one
- 50 companies actively asking us for hires
- 40 placements made
- $40K+ in success fees collected (or held in escrow pending legal)
- 10 unprompted company testimonials
- each student we onboard brings in 1.2–1.5 more through referrals
Hit 70%+ of those and the thesis is alive. Miss more than one by a wide margin and we owe you a hard rethink, not a spin.
Talk to Manu.
One email is enough. Manu replies same day.
Reply time is the marketing. Under three seconds, every time.