Skip to the story
MA
All work

Builder · Age 19

Hey, my name is Millad.

UC Berkeley Haas · Spieker Undergraduate Business Program

2020 → 2026. Scroll to explore.

◀◀ Rewind

2026

Six years earlier

Next: Sneaker bots

2020 · Age 13

2020. I was 13.

I bought a pair of Yeezy 350s I was super excited about. They were fake.

That got me motivated to get real ones, and to start selling them so people got authentic pairs.

This was my room. Click anything.

Receipts · Sneaker bots & reselling2020–2021
100+
Pairs flipped
$10–15K
Profit over 12–18 months, at least
+$650
First flip, in 7 days
If you don't track who's getting what, you end up paying for it.
Case study

▶▶ Fast-forward

2020

Two years later

Next: Dropshipping

◆FinderFileEditWindowWed 3:12 AM
youtube.com/watch?v=S4XpoPsy8sc

Zero to Making $100k Per Month at 18 | Biaheza Story

Biaheza

His thumbnail, from his channel. Biaheza on YouTube ↗

store #1 · recreation
Free shipping worldwide · Sale ends in 09:59

Most of my 15+ stores looked like this.

RFID-blocking wallet

Customizable water bottle

Ads Manager

Ads ManagerTraffic
ChannelResult

Click a row to see what happened.

experiments.txt

experiments.txt

  • First store: China supplier, sold as U.S. shippingkilled

    Shipping took forever. Customers got angry and filed chargebacks.

  • RFID-blocking walletskilled

    Flopped.

  • Customizable water bottleskilled

    Flopped.

  • Stores #1 and #2killed

    Paid for ads, didn't get much back. It took until store #3 to really make money.

  • 20-foot iPhone chargerscaled

    Easy to make content with. The videos went viral because the charger was so long.

Store results

Results

Best month (revenue)

$7,000

Another month

$2,000

Worst month

$0

Stores launched

15+

It started off really slow. It took until my third site to ever really make money.

store #3 · recreation
Free shipping worldwide · Sale ends in 09:59

My best-selling item, for whatever reason: a 20-foot iPhone charger.

20-foot iPhone charger

first-store-postmortem.txt

first-store-postmortem.txt

One of my very first stores.

I didn't know that for dropshipping you had to find manufacturers in the U.S., or be upfront about slower shipping if your manufacturer was in China.

I found a manufacturer in China on AliExpress and marketed it as U.S. shipping. The shipping time was incredibly long.

Customers got really pissed off and started doing chargebacks. Bad for a lot of reasons.

I bounced back, kept building stores, and learned from my mistakes.

2022 · Age 14

Then I kept seeing dropshipping all over YouTube.

One YouTuber, Biaheza, was about 18 and making way more than I ever could reselling sneakers.

So I tried something new. I launched over 15 stores. Most of them failed.

Receipts · Dropshipping stores2022
15+
Stores launched
$7,000
Best month (revenue)
#3
First store that made real money
Change one or two things on a site, and people react completely differently.
Case study

Your turn

Most of my stores failed, but the content was selling better than the ads. What do you think I did next?

▶▶ Fast-forward

2022

The next year

Next: TikTok & mentorship

2023 · Age 15

The creators

1creator

It started with one creator: a close friend from home.

Receipts · TikTok Shop & mentorship2023–2024
81
Creators in my community
15
New mentees in our best month
$1M+
GMV from my own videos, all time
$30K+
My best TikTok Shop months (sales)
$175K
GMV in 7 days, 2025
I really enjoy starting from zero with someone and trying to get them all the way to their goals.
Case study

▶▶ Fast-forward

2023

The next year

Next: Hookify

2024 · Age 16

Hookify

AI software that plugged into UGC creators' TikTok accounts, analyzed their hooks, and told them how to make better videos. Five other mentorship programs paid for it. One of them bought it when I was 16.

The sale

Soldat 16.

How it worked · rebuilt on sample data

How Hookify worked, rebuilt on sample data

  1. Customers connected their TikTok accounts. 16 videos from @smalldesk.sample, a fictional desk-setup account (metrics as of Oct 15, 2026). The original started with a TikTok connection; this rebuild loads the sample account instead.
  2. Hookify reviewed their accounts in depth using their video hooks and account/video analytics. Every opening line goes through two text rules, then each video lands on one chart: how many viewers its opening kept past 3 seconds, grouped by type of opening.
  3. It gave actionable feedback about what they could improve. Slow intros kept 42% of viewers past 3 seconds; direct openings kept 68.5% (medians, n = 3 vs 12).
  4. Customers used that feedback to adjust their videos, with the goal of making better content and getting more views. 3 experiments for the next videos, each with what to change and what to measure.
HookifyHookify reconstruction · Sample data

A fictional account in a present-day rebuild. Not the original app, and not real results.

  • Built it first for my own mentorship program, then broadened it and sold it to other mentorship programs
  • Learned ChatGPT could do a lot of the coding for me, before vibe coding was a popular thing, and built with it
Receipts · Hookify2023–2024
5
Other mentorship programs paid for it
100+
Creators used it
16
Sold at age
Your customers can become your buyer. The company that bought Hookify was already paying for it.
Case study

Junior year of high school · 2024–25

Profits for Purpose

An entrepreneurship program I founded at my high school in junior year, with about 50 students. The profits from the ventures they started in the program went to LLS: over $12,000 in one year.

~50

Students, approximately

$12,000+

Donated to LLS in one year

Collectively, from September to May of junior year, from the profits of the ventures students started in the program. All of it went to LLS.

One participant: Sean

  • Step by step: Sean came to me with no reselling experience. I tutored him step by step: where to find pieces, how to price them, and how to list them.
  • First sale: A vintage denim jacket he found for $6 and sold for $34 (before fees and other costs). That's more than five times what he paid.
  • Later: After that first sale, he kept sourcing and selling on his own. He still runs his vintage resale business.

Receiving an award from LLS (the Leukemia & Lymphoma Society) for the money we raised and donated.

Junior year (2024–25)

What I changed

Some students felt uncomfortable struggling publicly. So I moved to private feedback and smaller milestones. Retention improved.

The case study

▶▶ Fast-forward

2024

The next year

Next: Wealth management

2025 · Age 17

At the end of 2024, I started a wealth management business with a close friend, Andrew Elguindy.

We felt traditional firms weren't taking care of young entrepreneurs' needs, or helping them learn financial literacy.

2025 · Age 17

Elguindy Afshar Wealth Management

ActiveStarted at the end of 2024

We manage young entrepreneurs' money for them and, when needed, teach them what we're doing and why.

Assets under management

$600K+

Client money we manage, not revenue or profit.

It still runs. We're not taking new clients, so I've stepped back a bit.

Two client accounts

Snapshot
Oct 14, 2026
Date of the original screenshot file. It isn't visible in this crop.
Period
The three months to Oct 14, 2026
The app's “Past 3 months” view
Broker
Robinhood

Account value · labeled “Investing”

$78,973.81

Reported for the period · labeled “Past 3 months”

+$32,508.41 (+69.96%)

Robinhood's portfolio return for this account and period, which Robinhood says leaves out deposits and withdrawals. The source and method are under “How to read these numbers”.

Account 1. The original screenshot is dated Oct 14, 2026; that date isn't in this crop.

Accounts whose owners agreed to be shown, cropped so nothing identifies them. Not every account, and not the firm's overall results.

How to read these numbers
  • Every figure is the app's own, word for word, as of the snapshot date. The account value and the reported figure are separate numbers.
  • Account 1 (Robinhood): Robinhood's portfolio return for this account over its “Past 3 months” view. Robinhood says external money movements (ACH deposits and withdrawals, ACATS transfers and wires) are left out of its portfolio return calculations, using a method based on Modified Dietz. Robinhood: Using charts ↗
  • Account 2 (Brokerage app (not identified)): The app labels this “Year to Date Change”. The screenshot doesn't say how it's calculated, or whether deposits and withdrawals count in it.
  • There's no S&P 500 comparison. The two screenshots cover different periods from different apps, and only one app documents how its figure treats deposits and withdrawals.
  • Past results don't predict future ones. Nothing here is an offer or advice.

I learned how tentative people are with their own money.

2025 · Age 17

Jobs & internships

Jobs at two startups, and internships at a General Catalyst MD's family office and a wealth management firm with $230M+ under management.

Job

Sperm Racing

Led GTM

  • Led go-to-market (GTM) for Andros: how it reaches customers and which channels it sells through
  • Built the TikTok Shop infrastructure for Andros, their company: set up the storefront and its product listings
  • Set up the affiliate side, so creators can sell Andros products for a commission
  • Built a system for finding creators, sending them product, and tracking what their videos sell
  • Saw there wasn't much of a GEO presence, which is what led me to build Agentline
Case study

Job

Bon Health

GTM

An AI-powered health startup

  • Built out their go-to-market (GTM) systems
  • Built the creator system
  • Built everything for gaining customers through media

Internship

Marc Bhargava's family office

Intern

  • Researched companies raising pre-seed through Series B: reading their decks and sizing up the team, the market and the traction
  • Connected Marc with their founders
  • Tracked how his portfolio companies were doing after he invested
Case study

Internship

WE Alliance Wealth Advisors

Intern

A wealth management firm with $230M+ under management

  • Directly shadowed the managing director
  • Sat in on his investor meetings with families, and watched him break down his reasoning
  • Worked with the firm's employees
  • It showed us at Elguindy Afshar just how underserved our market was

Sperm Racing is also where I spotted the problem that became Agentline.

▶▶ Fast-forward

2025

I moved out.

Next: Agentline

2026 · Age 19

Agentline

Agentline helps software companies show up in AI answers. It finds whether ChatGPT, Claude and Gemini name you when buyers ask, why not, and who gets named instead. Then an agent drafts content from your verified facts and publishes it to sites like WordPress and DEV.to.

Wound down · Oct 2026

This year. I'm 19.

Everything before this taught me something I used here.

agentline · workspaceillustrative walkthrough · not a live run
Pick a scenario, then run the walkthrough.

Scripted: nothing here calls an AI model or publishes anything. The steps mirror what the real code does; real screens from the product are further down.

Where it came from

I was working at Sperm Racing and saw there wasn't much of a GEO presence: how often AI answers mention and recommend a company. I looked into the fixes out there, didn't like what I saw, and decided to build my own.

My role

  • I built the product: the audit, the diagnosis and Boost
  • I also ran the business side
  • Two co-founders: a fellow Haas student and an M.E.T. student
  • Built it AI-first: coding agents wrote a big share of our commits, with us directing and reviewing them

Stack: Next.js and TypeScript on Supabase (Postgres), the Claude, OpenAI and Gemini APIs, and background workers.

The product

Agentline's brand page for Agentline itself: what it does, and one audit's provider runs across the OpenAI, Google and Anthropic APIs, each marked Succeeded.

One audit of Agentline's own brand: each row is one question sent to the OpenAI, Gemini or Claude API, and every answer is kept as evidence.

Requested Oct 7, 2026

One question, start to finish

A real Boost plan from our audit of Agentline's own brand: what the product measured, what it inferred, and what it recommended.

  1. Buyer question

    “how to check what chatgpt says about my brand”

  2. Measured

    Agentline was named in 0 of 9 captured API responses: 0 of 3 each from Claude, Gemini and OpenAI. The answers cited sources like vertexaisearch.cloud.google.com and pingmybrand.com.

  3. Why it lost (AI inference)

    No clear winner: all three recommended other brand-monitoring tools instead of naming Agentline.

  4. Recommended action

    A third-party article on DEV.to, written only from verified facts about Agentline. Not a change to Agentline's own website.

  5. Outcome

    The plan was waiting for review when this was captured. No result was measured for it.

From a real Boost plan in our self-audit on Agentline's own brand (the Boost plan screenshot below).

Why we wound it down

We wound Agentline down.

What design partners said

Our design partners used Agentline for free and gave us feedback in return. A lot of them said there was a big need for this. One notable example was MyLife, a startup out of UC Berkeley. They were really trying to optimize for GEO presence, because that's where commerce is moving, and wanted their app to be incredibly discoverable there.

What we found in our target market

Later, we met with more startups in our actual ICP (ideal customer profile). A lot of the time, they were bringing GEO staff in-house to work on their presence, and weren't looking to outsource it. The ones that did were already working with established providers like Profound.

So the need was real, but the startups we were building for were mostly meeting it themselves, or with a provider they already had. That's why we wound Agentline down in October 2026. It was an amazing learning experience for all of us.

What worked

  • We got 5 design partners and ran the agent 100+ times.
  • We built a lot, fast, with AI doing much of the coding.

What I'd do differently

  • Do more customer interviews before building the product.
  • Ship fast, and talk to more customers than lines of code you write.
Receipts · Agentline2026
5
Design partners
100+
Agent runs
Case study

2026 · Age 19 · Now

Now, at Edgerun.

Revolutionizing the future of combat.

Edgerun company footage · “Edgerun exoskeleton, donned and powered” · Watch on X ↗

Edgerun is a YC-backed company building powered military exoskeletons, developed through field testing with Ukrainian operators. I'm helping the company work toward an $8–10 million seed round, with about $4 million already committed.

Seed round target
$8–10M
Committed to the company
~$4M

Round in progress · October 2026

October 2026 · Age 19 · An aside

My most recent fail

I tried to make a 3D rendering of myself for this site.

6 days, 29 versions of my face. The cartoon is the one that stayed.

Making of the character: 6 versions
  1. An avatar in the browserCut

    Oct 8

    An avatar in the browser

  2. A cut-out from an AI photoCut

    Oct 10

    A cut-out from an AI photo

  3. A MetaHuman of my faceCut

    Oct 10–13

    A MetaHuman of my face

  4. The facecamCut

    Oct 13

    The facecam

  5. The walkerCut

    Oct 13–14

    The walker

  6. Drawn in code

    Oct 14

    Drawn in code

Also

Everything else

Not everything fits in one story.

2026 · age 19

ActiveStudents already use it

TransferPlanner

A planner for community college students transferring to a university: what counts, what they still need, and what to confirm. Every rule comes with its official source.

Founder. Built it with AI coding agents

435
Commits
~67K
Lines of app code
50+
Paying users

From the repository, as of Oct 26, 2026.

transferplanner-psi.vercel.app/sample LiveOpen the live app ↗
TransferPlanner's sample student Jordan, Sierra College to UC Berkeley Political Science: the next step (the UC application deadline), where Jordan stands, and units completed and still needed.
Try it ↗

The real app, live. The sample students are made up and checked against real rules; nothing you enter is saved.

Case study

Experience

Side projects

Failed experiments

2026 · Age 19

At Edgerun
Working with Edgerun's founding team to put them on calls with investors and further their $8–10M seed round
At Berkeley
First-year at UC Berkeley Haas, in the Spieker Undergraduate Business Program