Skip to content
AI Dev Cases
Social Wizard

Arrived in the US with $100 — How a 22-Year-Old Built an App That Kills “Fear of the Wrong Reply” on GPT-4 Vision’s Launch Day and Hit $60K/Month

Kelechi Onyeama left Nigeria for the US with just $100 and built Social Wizard right as GPT-4 Vision launched: upload a screenshot of a chat and AI suggests what to say back. Daily TikTok posting snowballed into $60K/month and 500K+ downloads — and his second app, CleanEats, later got acquired too.

Published: Sep 14, 20263 min readPrimary-source verified · 5
Monthly (est.)
$60k/mo
Users
500K
Launched
2024
Kelechi OnyeamaKelechi Onyeama@kelechispArrived in the US with $100 — How a 22-Year-Old Built an App That Kills “Fear of the Wrong Reply” on GPT-4 Vision’s Launch Day and Hit $60K/Month

Key takeaways

  • Before outsourcing to influencers, post to TikTok/Reels yourself every day until you find a format that lands
  • Once a video format works, commission it out to many creators and channels in a native, non-ad-looking style
  • Believe “session one is everything” — track and keep shrinking the time from install to the first real value moment

The pain point, and how they found it

Freezing up when replying to a text, terrified of saying the wrong thing or leaving someone on read — a near-universal Gen Z anxiety. With a crush or an awkward topic especially, people burn time agonizing over what to say and end up not replying at all. Kelechi built the product out of his own experience: “staring at my phone, paralyzed by the fear of saying the wrong thing.”

Background & product

Social Wizard lets you screenshot a chat (or type the conversation directly), and AI reads the context to suggest replies in several tones — casual, formal, or “flirty” — plus a feature that generates an opener from a photo.

Founder Kelechi Onyeama is from Nigeria. His first app, Caspade — a social app for sharing photos within friend groups — launched there but only drew a couple thousand users before stalling. Convinced his location was the bottleneck, Kelechi flew to the US with just $100, slept on a friend's dorm-room floor, and bounced between relatives' places while looking for his next shot.

That's when OpenAI released GPT-4 Vision, a model that could “read” images. Kelechi moved fast to build on it, launching Social Wizard on January 19, 2024. At the time, almost nothing on the market could look at a screenshot and suggest a contextually relevant reply — being first gave him a real head start.

With zero ad budget, Kelechi posted to TikTok and Instagram Reels himself, every day, letting word of mouth do the rest. The result: $60,000/month in revenue and 500K+ cumulative downloads (his own site later claims 700K+), enough for outlets like Digital Journal and TechBullion to cover him as an AI app solving Gen Z's social anxiety.

From the founder (primary source)

Kelechi Onyeama
Kelechi Onyeama
@kelechisp
·

Update: $40k 2 months of finding what works, we’re now in scale mode hired a full time dev, now hiring a full time video editor. if you’re cracked/know someone plz text me!

Kelechi Onyeama
Kelechi Onyeama
@kelechisp

my new app just crossed $10k in rev few days into marketing if you’re super cracked at expo & node js, call me rn for equity. must be in US, must not have a 9-5 we’re going to $1m/mo

Reply
Read 7 replies
Social Wizard growth channels and tech stack

How it was built

  1. Analyze why the first attempt failed (Caspade in Nigeria) and try again with the location, the timing and first-mover access to a technology all changed
  2. Start building the moment OpenAI released GPT-4 Vision, a model that can read images
  3. Ship while almost nothing on the market could look at a screenshot and suggest a contextual reply, claiming the first-mover position (launched 19 January 2024)
  4. Narrow the core experience to this: import a conversation as a screenshot, let AI read the context, and propose replies in several tones — casual, formal, flirty
  5. Live by 'session one is everything' — make the time from install to the first moment of value (the first suggested reply) a tracked KPI and keep shortening it
  6. Once confident in the product's quality, raise the price from $6.99 to $9.99 a week and test the effect on both revenue and LTV, with a three-day free trial
What AI did
Being first to GPT-4 Vision was what made the product possible at all — but that technical edge is imitable, and the durable advantage turned out to be the distribution playbook

How it got users

  1. With zero ad budget, post to TikTok and Instagram Reels yourself every day before turning to any outside influencer
  2. Keep going through the quiet early response, and when one video hits 17,000 views, seize the angle that landed and refine it — reversing this order outsources the learning and destroys repeatability
  3. Identify the format that worked: a short, visual sequence of screenshotting a chat, showing the app name, and a witty reply appearing instantly
  4. Contract that format out across multiple creators and channels while keeping it native and un-ad-like (one video reached 2 million views and produced tens of thousands of dollars on its own)
  5. Strip the advertising feel to match how TikTok users behave — skipping ads while responding to a creator's personal recommendation
  6. Don't stop at one hit — reapply the same distribution playbook to a new product in another category (the second app, CleanEats, made $10K in two weeks and sold for a low six-figure sum)
First users
Organic viewers of the TikTok and Instagram Reels the founder posted daily, with a single 17,000-view video as the turning point
Early ad spend
$0 — no ad budget at all, only daily posting by the founder and contracting the winning format out to creators

The repeatable playbook

  1. 1Before outsourcing to influencers, post to TikTok/Reels yourself every day until you find a format that lands
  2. 2Once a video format works, commission it out to many creators and channels in a native, non-ad-looking style
  3. 3Believe “session one is everything” — track and keep shrinking the time from install to the first real value moment
  4. 4Once you trust the product's quality, raise the price and test the effect on both revenue and LTV ($6.99 → $9.99/week)
  5. 5Don't stop at one hit — redeploy the same distribution playbook on a new product in a different niche (his 2nd app, CleanEats, also sold)
  6. 6Diagnose why your first attempt failed (Caspade, in Nigeria) and change the variables — location, timing, being first on new AI — before trying again

Channels they actually used

The hard parts

Social Wizard has drawn persistent App Store review complaints about auto-billing after the 3-day free trial and being an overpriced “ChatGPT wrapper” that just reads screenshots. The price increase from $6.99 to $9.99/week lifted revenue and LTV but also fed some of that backlash — a reminder that aggressive pricing cuts both ways, which we note in the body. His age is also reported inconsistently across outlets (20–22 depending on when the piece ran; his own X post says he arrived in the US at 20, “two years ago” as of an August 2025 post), and we could not confirm the app's official country of operation from primary sources, so `country` is left unset.

Deep dive

【Deep dive】Breaking down the zero-ad-budget playbook that took Social Wizard to $60K/month — and how Kelechi refused to stop at one hit.

“Inherently viral” — become the creator yourself before outsourcing

Kelechi describes his marketing philosophy as “inherently viral”: showing people something they haven't seen before. With no ad budget, he posted to TikTok/Reels every single day. Traction was thin at first, until one video hit 17,000 views — the moment he found a format that landed, which he then refined. Validating what works yourself before handing it to paid creators is the same sequencing seen in other hits like Cal AI's Zach Yadegari; reverse the order and you outsource the learning itself, losing repeatability.

Scale the winning format without it looking like an ad

Once he had a format that worked — screen-recording a chat screenshot being taken, the app name appearing, then a clever reply popping up instantly, short and visual — he commissioned the same format across many creators and channels. One video hit 2 million views and brought in tens of thousands of dollars on its own. Native-feeling videos that don't read as ads fit how TikTok users behave: they skip obvious ads but trust a creator's own pick.

“Session one is everything” — kill first-run friction with a metric

Kelechi's operating belief was “session one is everything.” He tracked and kept shrinking the KPI “install to first hint” — the time from opening the app to seeing its first AI-generated reply. Showing the app's “magic” fast prevents day-one churn and feeds straight into paid conversion — he put as much energy into onboarding optimization as into acquisition.

Testing LTV with a price hike

Pricing started at $6.99/week and, once he trusted the product's quality, moved to $9.99/week (with a 3-day free trial), per reporting. The increase lifted both revenue and LTV — but, as covered below, it also fed complaints about auto-billing after the 3-day trial. A useful reminder that aggressive pricing cuts both ways.

The growth curve (note: metrics aren't apples-to-apples)

Stitching together his own X posts and press coverage: MRR hit $3,500 two weeks after finding a working channel, $25,000 a month later, and $40,000 at two months (when he reported hiring a full-time developer to scale production). Multiple independent outlets later converged on “$60,000/month.” His personal site's later self-reported figures show 700K+ downloads and $800K+ cumulative revenue. Because the metric shifts between MRR, cumulative revenue, and self-reported totals across sources, these numbers aren't directly comparable — but every framing tells the same story of sustained growth.

Not stopping at one hit — the second app, CleanEats, also sold

After Social Wizard took off, Kelechi reapplied the same distribution playbook to a different niche. CleanEats scans a food barcode and scores it on two axes — a “skin score” and a “weight score” — an idea born from his own 30-pound weight loss. It made $10K in its first two weeks, reached $65K cumulative revenue and a 4.9-star rating (2,700+ reviews), and sold in a two-day close for a low-six-figure sum. He described it less as a financial win than a psychological one — proof he could build an asset from nothing and carry it through to an exit. By his own account, twelve months after landing in the US with $100, he'd turned it into two viral apps, 700K+ cumulative downloads, and $1.5M in portfolio revenue.

The lesson: the distribution format is the moat, not the head start

Being first to ship on GPT-4 Vision was a technical edge that others could eventually copy. What actually persisted was the format Kelechi found by posting daily himself, and a repeatable playbook he could redeploy across channels and across products. The takeaway: the distribution format itself becomes the asset — more than the novelty of the technology.

Cross-case “growth playbook” report (coming soon)

We're building a paid report that aggregates every case in this database: which acquisition tactics worked, in which categories, and how well — insights you can't see from a single story. Get notified first when it launches.

Related cases