Founder, Surge Software (bootstrapped since 2018). Building AI-first B2B products on our own AI servers: ReplyDak (cold email) + Shishya (free AI tutor).
@dutinglove@gregisenberg Citations without clicks often mean the answer already covers the question. A page with a next step (a tool, calculator or checklist) gives people a reason to click. We built BrandInAIAnswers: a one-line script counts visits from AI assistants. https://t.co/rpYnbUtKDo
@cryptogemguy@Joncole23@treg_ai@dsouzaalan Mostly Google inboxes underneath. The real work is setup: lookalike domains you own, SPF/DKIM/DMARC and a slow warm-up. We built ReplyDak for that: Google Workspace inboxes ready in about 10 minutes, with a CSV for Instantly or Smartlead. https://t.co/znghGBiXWQ
@aakashgupta If X also pays the original creator, the smart move for a podcast is seeding the first clips yourself so clippers copy the right moments. We built ShortsFromVideo (https://t.co/amwX53hacJ) for this: it scores each clip and says why; you could try it on your next episode.
@baseddesigner Nice jump. With real usage in a raise deck, investors ask where each number comes from, so keep a source under every stat slide. We built SlideDeckMaker (https://t.co/912y2L9SEh) for this: every number traces back to your file; you could try it on your usage export.
@Emmanuelkalu546 Apollo and Hunter guesses go stale fast. Verify every address right before you send and drop catch-alls, so bounces never hit your domain. We built ReplyDak (https://t.co/znghGBiXWQ) for this: a list verifier plus sending inboxes; you could try it on your next Apollo export.
@yongfook Matches what I see across 40+ small products: the free tools collect sign-ups, the paid ones get silence until a human talks to the buyer. If a person has to be in the loop anyway, the price has to pay for that person's time. That kills the $19 middle.
@michaelonx_ I'd read the opposite list too: what the apps stuck at zero have in common. I run 40+ products at ₹0 revenue, and they all stall at the same step, a stranger finding them. Building was never the slow part.
The backup gateway point is the one I learned the hard way. A gateway rejected my payments for a month because all my products sat under one account registered as an education business. Nothing broke in the code, I just couldn't take money. Now each business type gets its own account.
@dannypostma Mine ended up as plain text files and a schedule. Each of my 40+ products has one facts file the agents read first, and scheduled tasks do lead research at 10 AM and a video series at 7 PM. Not pretty, but every founder's version fits how their own head works.
@gregisenberg Seen from the small end: one of my apps got 39,662 sign-ups in a week, every one a bot. I put sign-up guards on all 20 of my apps with a sign-up page that same day. Small builders buy security the day after the attack, which is exactly why the demand keeps growing.
@levelsio@Hetzner_Online@digitalocean That's roughly how my busiest sites run now: a server in my house, web traffic only through a Cloudflare tunnel. I set it up after a $751 bill from crawlers on managed hosting. Safe by default would save a lot of people learning it the expensive way.
@yongfook The indie hackers buying that bot would include me. 40+ products, ₹0 revenue, and the one real lead I've had came from a person answering a person here. I doubt buyers want chad energy from a bot. They want proof someone real answers the DM.
@cipherslice Still in the DM, no purchase yet. Revenue is still ₹0, so it's an open loop, not a win. When it turns into a 1, the number goes in a post here, same as the zero did.
40+ products. ₹0 revenue so far.
I'm saying it out loud because most "I built 40 apps" posts skip this line.
My day job is Adobe AEM architecture. Evenings and weekends I build. The building part got easy: AI agents write most of the code now. The selling part didn't.
What I learned in the first months:
- shipping is no longer the bottleneck. Distribution is.
- 20 products with 0 traffic lose to 1 product with 50 visitors a day.
- free tools get sign-ups; paid tools get silence until someone talks to a human.
So the next 90 days are about 4 products with paying intent, and talking to buyers every day.
I'll post the revenue number here every month, including when it's still zero.
I published 5 long posts in 7 days. Followers went from 98 to 99.
One new follower. Here's the real week:
- a long post most days, each with a real number (missed 2: a billing error and a frozen browser tab)
- up to 8 replies a day, spaced minutes apart
- median views per reply: 1 to 2
- one 3 Nov label from X that says it "may impact reach"
The label came from my own mistake: 22 fast replies in one day. X read it as a bot, and it still shows in the numbers two weeks later.
What I'm not doing: buying followers, joining reply pods, or going quiet.
What I am doing: the same boring routine for another week, and posting the number again on Sunday.
Most growth posts show the curve after it bends. This is the flat part.
@SashaZaitsev5@upbeat_dev Usually nothing it reads names your app: no comparison pages, store reviews or Reddit threads. One page per buyer question, with your app beside the usual picks, helps. We built BrandInAIAnswers (https://t.co/rpYnbUtKDo): runs buyer questions on Claude; try carlect.
@SujinBuilds Venu, Bengaluru. Adobe AEM architect by day; evenings go to small AI products, 40+ so far, with Claude Code doing most of the build too. My lesson to pass on: building got easy, finding the first payer didn't. Start talking to health users this week, not after launch.
@divvsaxena Only one here too. Claude Code in a terminal does most of the build work across my 40+ products, and code that used to eat a weekend now takes about an hour. The editor I miss isn't for writing code. It's for reading what the agent changed before I trust it.
@DanielSmidstrup $5k MRR, easily. I'm at 98 followers, and my one real sales lead still came from a single useful reply, not from reach. One buyer who answers your DM beats a thousand people who scroll past.
@ben_sage Shipping isn't the signal anymore. I've built 40+ products with agents doing most of the code and still have ₹0 in revenue to show for it. The engineers who stand out now are the ones who can sit with a user, find the problem worth paying for, and verify what the agent wrote.
Mine got better when I stopped being mean and started keeping receipts. Each product of my 40+ has a facts file, and every time an agent gets something wrong, the correction goes in that file. It reads the file before it touches anything. Blunt and written down beats angry in chat.