Nov 3, 2026 · Distrarc · Final write-up of my two-week 3F trial
The payment cleared. ₹2,999. A therapy clinic in Delhi. My first paying customer, after 14 days of 16 hour days, 729 commits and more rejection than I'd like to admit.
Then I refunded every rupee.
To understand why, you need to see the two weeks that led up to it. And the 30 minute timer that almost killed the whole thing.
The bet
I'm 16. I'm building Distrarc: it finds people on Reddit, X and Hacker News who are asking for exactly what your business sells, and drafts a reply for each one.
On 20 October I made a deal with 3F VC, a fund that backs founders under 25. Two weeks to prove people will pay for this. Before starting, I wrote down exactly how I'd be judged, including the numbers that would count as failure, so I couldn't quietly move the goalposts later.
Here's the scorecard. Read it now, because I'm not going to hide it at the bottom.
| Success metric (by 2 Nov) | Result |
|---|
| At least 2 paying customers, or 1 agency on $129+ | 1 paid. Refunded by me. |
| At least 20 real conversations with possible buyers | Missed |
| At least 6 free weeks started | Missed |
| Cold email reply rate of at least 2% | 7.4%. Nearly 4x the target. |
| 1 customer with drafted reply → click → signup | Not yet |
| Failure criterion | Result |
|---|
| 0 paying customers by 2 Nov | Avoided |
| Reply rate under 1% after ~500 emails | Avoided. 7.4% after 704. |
| Fewer than 3 free weeks by 26 Oct | Hit. I said so that day and changed course. |
I missed most of it. I fucking wanted every single one. But a sprint isn't there to make you look good. It's there to prove a point, fast, with nowhere to hide. Here's what these two weeks actually proved.
Lesson 1: a great reply rate can lie to your face
I sent 704 cold emails. 7.4% replied. Over 50 people wrote back to a 16 year old they'd never heard of.
The line that worked never said what Distrarc is. It said: I found people asking for exactly what you built. Want to see them? People who ignore every pitch in their inbox answered that one.
I thought I'd cracked it. Replies kept coming, so I kept going.
Then I actually sorted them.
- 50-plus replies
- 32 positive
- 3 who will genuinely pay
The rest loved the leads and had no budget. Students. Solo builders. People thrilled to see their own customers who couldn't spend $49 to reach them.
A reply rate measures curiosity. It doesn't measure wallets.
So on 31 October I threw out the list and rebuilt it around one signal: companies already hiring for marketing. If you're posting a job for a growth marketer, you've already decided to spend money on customers. You don't need convincing. You need a cheaper way. Those replies are still landing as people dig out of the weekend, and it's the list I'm betting on now.
Lesson 2: the right person is always reachable
Cold email is the front door. Everyone's knocking on it.
So I went around it. I found a way to reach Gustaf Alströmer directly, the YC partner who led growth at Airbnb in the early days, and asked him for feedback before I applied to YC.
He replied. My application went into his reading queue.
I did the same with owners of Reddit marketing agencies, the exact people whose job is the problem I'm solving. Some of those conversations went further in one message than a month of cold email would have. Some didn't land at all. I'm still figuring out which first touch works for whom.
But the lesson stuck: almost anyone is reachable if you're willing to do the work nobody else bothers to do. Doing things that don't scale isn't a phase. It's the job.
Lesson 3: test the bridge before you build it
In week two I stopped waiting for emails and started picking up the phone.
When someone ran a free report on distrarc.com, I looked at what it found for them and called them the same day. Follow-ups moved from email to WhatsApp. Replies came back in minutes instead of days.
That's how I got the clinic.
They were paying Meta ₹600 to ₹700 for every single lead. On the call I didn't pitch. I showed them real people in their city, posting in the last 48 hours: "need a therapist, budget ₹1k."
You could hear it land.
They signed up on the call. Three day trial, then ₹2,999 a month. The trial ended. The card cleared.
First paying customer.
The 30 minute timer
The clinic had made a brand new Reddit account to reply to their leads.
Here's what almost nobody knows: Reddit throttles new accounts brutally. Roughly one comment every 30 minutes. Their very first reply got auto-removed in five minutes.
Do the math. Distrarc hands you 50 people asking for exactly what you sell, and you can reply to almost none of them before they go cold. Every hour a lead sits unanswered is an hour a competitor gets to them first.
They were skeptical. They were right to be.
The leads were real. The bottleneck was the platform. But a bottleneck between my customer and their customers is my problem, not theirs.
So I refunded every rupee, gave them another week free, and that same day shipped the fix:
A Chrome extension that posts for you, inside each platform's rate limits, from a queue. Distrarc finds the lead. Drafts the reply. You hit approve. The queue posts it the moment your account is allowed to. No babysitting a timer. No burned accounts.
Could I have kept the money? Sure. The scorecard would look better.
But your first customer isn't revenue. They're your reference. Every business I talk to next is going to hear about Distrarc from somebody, and I want that somebody saying: "he gave my money back and fixed it the same day."
That's worth a hell of a lot more than ₹2,999.
A bottleneck between my customer and their customers is my problem, not theirs.
Then it happened again
On 28 October, Boardy, the AI that makes introductions between founders, connected me with a SaaS founder running several products. He'd been hunting high-intent conversations on Reddit and X by hand. Exactly the job Distrarc does.
Our first call slipped. Boardy rebooked it. The night of the second call, I sent him ten Reddit threads from the last three weeks: people actively cancelling an AI content tool and asking what to switch to, people asking which AEO tool is actually worth paying for, a small business owner wanting to track their visibility in ChatGPT. Five I'd act on today, five weaker ones so he could see how the filtering works. I told him plainly they were up to three weeks old, because it was a manual pass, and that the scheduled version would catch them the day they appear.
Then the answer came back. The leads weren't the problem. His Reddit account was banned. He couldn't comment at all. What he actually wanted was the ban lifted.
I wrote back within two minutes: I can't get a Reddit ban lifted, and I won't pretend I can. But Reddit is one source out of seven. Switch it off, and you get the same high-intent buyers from X and LinkedIn, posted from your own browser session by the extension, with no Reddit account involved.
Two customers. Two completely different businesses, a therapy clinic in Delhi and a software founder on the other side of the world. The exact same wall. Not "find me leads." Every tool claims to do that. The wall is: I found them, and my account can't reach them.
That's the moment I stopped thinking of Distrarc as a lead finder. Finding the post is table stakes. Getting a reply in front of that person, from an account that stays alive, on whichever platform will let you, is the actual product.
Why your website is about to stop mattering
Here's the part that keeps me up at night, in a good way.
Google is rebuilding search around AI answers. Ask it something and you get a summary, and the sources it leans on are forums. Reddit threads. Real people arguing in public.
Not your 12 page company website with a stock photo of a handshake.
AEO, answer engine optimization, is about to matter as much as SEO ever did. And answer engines are built on the conversations happening on Reddit, X and LinkedIn right now.
So the blunt version: if your business isn't in those conversations, being useful, getting quoted, getting upvoted, you're going to quietly vanish from search. Your website will still exist. Nobody will ever see it.
Most businesses are standing in the old ocean. The new one has almost no ships in it yet. That's where I'm sailing.
What actually makes this hard to copy
Keyword alert tools tell you a post exists. Then you're on your own.
Everything we built is aimed at what happens after the alert, and every piece compounds.
1. Outcome data nobody else has. Every reply Distrarc drafts is tracked: post, reply, click, signup. Alert tools know which keywords fired. We're building the record of which replies actually made someone money, by community, by industry, by city. Every customer makes the next customer's replies better.
2. A rulebook for every community, written in scar tissue. Which subreddits ban self-promotion. How new accounts get throttled. How long a helpful, link-free reply survives versus a pitch. We learned it by getting burned. Now it's built into how every reply is drafted and scheduled, so our customers never get burned the same way.
3. Posting that never gets your account killed. Two out of two early customers hit an account wall before they hit a lead problem. The queue posts inside each platform's limits, and every published reply is read back and logged with a verified receipt. Tested live end to end: one job, one post, one receipt, and a replay that refused to post twice.
4. Search in the buyer's words. Nobody posts "I want to buy a tool." They post their problem. Distrarc searches the way buyers talk, where buyers actually hang out. After benchmarking 17 AI models on real data, we find 64% of real leads instead of 44%, for the same cost.
5. It sounds like you. Replies are drafted in your voice and get sharper every time you edit one.
6. Costs low enough to outscan anyone. One full scan costs about $0.016. The whole company runs on about $100 a month. We can scan every hour while others meter you per keyword.
How one 16 year old ships like a team
I don't write most of the code anymore. I run a team of AI agents, and I make the calls.
Claude Code plans and reviews. Antigravity agents build features in parallel, each in its own isolated git worktree so they never step on each other. Codex handles release and safety. Cursor jumps in the moment the others hit their usage limits.
The trick is the same one that makes any small team fast: build it so any piece can be swapped in a few lines. That's how I could test 17 models in a week instead of a quarter.
915 instructions. ~300 agent sessions. 729 commits. 14 days.
What shipped:
- The extension and posting queue. Built the day a customer told us what was wrong.
- Service areas. Tell Distrarc you work online, in person, or both, and get leads from your own city.
- The fix that fixed nothing. One change passed 4,700 tests, went live, and changed absolutely nothing. Production was running a newer search engine than the one the tests were checking. Green tests, dead code. I traced what production actually runs and fixed the real thing. A mutual fund app went from 0 leads that morning to 4 that night.
- A cloud migration overnight. My Azure credits ran out early, so I moved the entire stack to Google Cloud in one night: Cloud SQL with 34 migrations, always-on workers for discovery and posting.
- Scan cost cut nearly 10x. From around $0.15 to about $0.016.
- Two security holes closed before anyone found them. One let our servers be tricked into fetching private network addresses. The other let one account's GitHub identity be attached to someone else's. Both dead.
- A release process that would hold up at a real company. 1,900+ automated tests, 326 database tests, end-to-end tests for the app and the extension, a rollback drill, and a staged rollout this morning: 1%, 5%, 25%, 100%.
- Less product, on purpose. I ripped out cold email campaigns, mailboxes and enrichment. They were tools I needed, not what customers pay for. Deleting your own features hurts more than building them.
All of this while my home router died during Diwali cleaning and I ran the company off a 4G hotspot at 0.8 Mbps upload.
The no I'm not taking
On 28 October I met Dr. Shruti Rajagopalan, Senior Research Fellow at the Mercatus Center at George Mason University, who runs Emergent Ventures India.
Her verdict: this idea wasn't ambitious enough for me. Go find a bigger one.
I respect that completely. She's backed a lot of young people and it's a fair read.
But I've spent two weeks watching business owners light up when they see real buyers asking for them, posted two days ago. I've watched the biggest search engine in the world start treating forums as the source of truth. A tool that puts every business into the conversations that decide whether it gets found at all doesn't feel small to me.
I think this is the big thing for me. I'm not letting go of it.
What I'd tell anyone running the same experiment
- Will businesses pay? Yes. One already paying ₹600 a lead on Meta paid within days of seeing real people asking for them.
- Who converts? People already spending on customers. Businesses paying for leads. Companies hiring marketers. Don't convince people customers are worth money. Find the ones who already know.
- Which channel? Cold email buys curiosity: 7.4% replies, 3 real buyers. A free report plus a same-day call buys customers. Reaching the right person directly beats all of it.
- Do drafted replies turn into signups? Not proven yet. It's the most important line on this page. The first real reply got killed by Reddit's new account limits, which is exactly why the queue exists. Proving this is next month's whole job.
What happens next
The clinic gets their extra week with the queue posting for them, and I try to earn that payment back. I close the 3 people who told me they'd pay. I work the list of companies already hiring for marketing. And I chase one number above all: a verified reply → click → signup on a customer's dashboard.
Thank you to Yash and 3F for the deadline. It made me move faster than I knew I could.
Two weeks ago I had zero customers. Today I have one I paid back, a fix I shipped the same day, and a much clearer picture of who will pay me and why.
I'll post the next numbers whether they're good or ugly.
Watch the next two weeks.
If your business should be in the conversations that decide whether you get found, Distrarc is at distrarc.com.