1 of 16

AI HACKATHON 2026 · UC BERKELEY

STING

STING - Scammers picked the wrong targets.

STING — Scam Tracking & Intelligence Network Guard. Real-time anti-scam protection that stops the scam before the money is gone.

Solo founder · built with a workforce of AI agents

01 / 12

2 of 16

WHO IT'S FOR

STING

Built for the people scams actually hurt.

STING didn't start with a market. It started with a person — and a conversation I couldn't stop thinking about.

THE MOMENT IT STARTED

“An older gentleman at church told me he gets scam texts and calls so often that he sometimes can't tell what's real — because he isn't comfortable with technology.”

He is exactly who STING is for.

NOT JUST AN “ELDERLY PROBLEM”

My mom isn't old — and even she started worrying after a week of fake bank texts and calls. Scam fear spreads through whole families.

WHO STING PROTECTS

Older adults

Less-technical users

Immigrants & non-native speakers

Overwhelmed & busy people

Families protecting the people they love

Anyone scared by repeated attempts

WHY IT MATTERS

Scams don't just steal money. They steal confidence, safety and trust — leaving people ashamed and isolated.

And the cruelty runs both ways: many scam operations are run by people trafficked into call-center compounds. It's a global exploitation network — STING fights the system, not the person forced inside it.

STING isn't built for power users. It's built for the moment someone is scared, unsure, and about to be manipulated.

03 / 12

I didn't start with a market — I started with a person. I was at church asking people what they were struggling with, and an older gentleman told me he gets so many scam texts and calls that he genuinely can't tell what's real anymore, because he isn't comfortable with technology. That stuck with me. He is exactly who STING is for. And it isn't only the elderly — my mom isn't old, and even she got scared after a week of fake bank texts and calls. Scam fear spreads through whole families. STING is built for older adults, less-technical users, immigrants, busy people, and the families trying to protect the people they love. Because scams don't just steal money — they steal confidence, safety and trust. And the cruelty runs deeper: many scam operations are run by people who were themselves trafficked into call-center compounds. This is a global exploitation network that hurts victims on both ends. STING isn't built for power users. It's built for the moment someone is scared, unsure, and about to be manipulated.

3 of 16

THE PROBLEM

STING

Scams work because the victim is alone at the moment of decision.

No one is standing next to them when the fake bank page loads, when the "support agent" asks for the code, when the checkout looks just real enough. Scammers engineer that isolation on purpose — it is the entire business model.

$12.5B

reported lost to fraud in the U.S. in a single year — and most of it never gets reported at all. — FTC, 2024

Before the click. Before the payment. Before the shame.

02 / 12

4 of 16

WHAT IT IS

STING

A real-time anti-scam guardian, right in the browser.

DETECT

Reads the page for scam signals

Fake bank logins, seed-phrase traps, fake support, impersonation, risky checkout flows — caught in real time.

EXPLAIN

Plain-English verdicts�& actionable guidance

No jargon. A clear risk verdict and the reasons, written the way you'd warn your own family.

DOCUMENT

Evidence receipts

& reporting

A dossier the user — or their family — can keep, share, and act on. Proof, not just a popup.

Local-first. Core protection runs on-device — graceful fallbacks mean it still works with no keys, no network.

05 / 12

5 of 16

LIVE DEMO

STING

Suspicious page → STING hunts → verdict + evidence.

① The hunt — pick a scam, watch STING tear it apart

cloak-sting-hackathon-2026.vercel.app/#demo

② The voice path — Deepgram transcribes, STING scores the call

cloak-sting-hackathon-2026.vercel.app/#voice

③ The verdict — in plain English, no blame, evidence to keep

At the table? Scan the last slide or hit Try a Demo Scan — judges welcome to drive it.

06 / 12

Let me show you the real thing — this is all built and live. On the left, the hunt: a judge picks a scam scenario — a fake USPS package text — and watches STING tear it apart in real time and return GUILTY, high threat. Below it, the same engine on a phone call: Deepgram transcribes a bank-fraud robocall and STING scores it ninety-four out of a hundred, critical. But the piece I care about most is on the right — the actual in-browser popup. Look at the tone: 'You haven't done anything wrong. You're safe as long as you don't type anything here.' It names the exact tricks — urgency pressure, trusted-institution language, repeated requests for your code — and then it offers to save proof for your bank or family. That's the whole product in one card: catch it, explain it without shame, hand over evidence.

6 of 16

THE UNIQUE SELLING POINT

STING

The Scammer Handoff Packet.

One click turns a suspicious moment into something a family member, bank, platform, or law-enforcement office can actually act on.

Scam attempt

→

STING warning

→

Handoff packet

→

Family · Bank · Law · Database

01

SHAME-FREE

Ask someone I trust

Calm, copyable words to send a family member or caregiver — no panic, no blame.

“Can you check this before I do anything? Something feels off.”

Copy & send

to Mom

02

EVIDENCE

Preserve the evidence

Everything captured automatically, the moment it happens.

RISK SCORE

91 / 100 · HIGH

URL & domain

Phone · email

Fake login

Screenshots

Impersonated co.

Timeline

03

ROUTING

Report to the right place

STING points to the next step — the right desk, not a dead end.

›

Bank or card issuer, if payment was involved

›

Local law enforcement, if it's local

›

FTC · IC3 · national cybercrime portal

›

Dating, marketplace or email provider

04

NETWORK

Warn the network

Reported indicators join a shared scam database — so the next person is already protected.

SUSPECTED SCAM CLUSTER

Reused numbers, emails, domains & scripts clustered into shared scam infrastructure.

Builds infrastructure profiles — never doxxing private individuals.

STING's moat is not just detection — it is the handoff from panic to proof.

USP

7 of 16

SHARED MEMORY

STING

Caught once, blocked for everyone.

Every scam STING catches becomes shared memory — a living blocklist of the domains, emails and numbers behind real attacks.

● LIVE

Today: STING's own detection memory, on Redis.

cloak-sting-hackathon-2026.vercel.app/#hall-of-fame

04 / 12

8 of 16

THE BIGGER PLAY

STING

Don't just block scams — help end them.

Most large-scale scams are run from forced-labor compounds. The UN estimates hundreds of thousands of people have been trafficked into them and coerced into scamming. The person on the other end is often a victim too.

1

Protect the public — a shared blacklist so no one falls for a scam STING has already seen.

2

Compile the dossier — hand structured intel to regulators & anti-trafficking units in scammer and victim countries.

Attack the infrastructure, not the hostage.

SCAM-OPERATION DOSSIER

REFERRAL READY

CLUSTER

#SEA-0473

LINKED INFRASTRUCTURE

14 domains · 6 numbers · 3 wallets

PATTERN

USPS / bank impersonation → crypto

SIGHTINGS

9,140 · first seen Dec 2025

REFERRAL TARGETS

Financial regulators

Anti-trafficking units

Operator identities withheld · workers are frequently coerced victims.

ASPIRATIONAL — the referral pipeline is next; the detection & intel it builds on are live today.

09 / 12

And I want to be honest about who's on the other end. Most large-scale scams aren't run by lone villains — they're run out of forced-labor compounds in Southeast Asia. The UN estimates hundreds of thousands of people have been trafficked into them and coerced into scamming under threat. So the person typing to your grandmother is often a victim too. That reframes the mission. STING already turns every catch into structured intelligence — the domains, the numbers, the payment rails, the patterns. The bigger play is to do two things with it: protect the public with a shared blacklist, and compile dossiers we can hand to regulators and anti-trafficking units in both the scammer's and the victim's countries. Attack the infrastructure, not the hostage. To be clear — the referral pipeline is future work; the detection and the intelligence it's built on are live today.

9 of 16

BUILT SOLO

STING

One founder. Sponsor tech, used honestly.

IN THE PRODUCT (RUNTIME)

Claude · Anthropic

Plain-English scam explanations

Deepgram

Voice scam transcription + analysis

Browserbase

Isolated inspection of risky links

Redis · Upstash

Scam memory & repeat-offender list

Sentry

Reliability & error monitoring

Arize · Phoenix

Evals + tracing for verdict quality

Fetch.ai · ASI:One

Agent coordination wrapper

Vercel

Live demo & landing hosting

HOW I SHIPPED IT SOLO

Codex · Devin · Cognition

My AI engineering workforce

Simular

Autonomous QA testing

Pika · Midjourney

Demo visuals & brand imagery

OpenAI

Secondary model path / tooling

I decided → agents executed → I iterated against false positives.

Not a wrapper — a chosen problem & a safety bar

Built ethically — local-first, no surveillance

08 / 12

On sponsors — I'll be straight about what runs in the product versus what helped me build it solo. In the runtime: Claude writes the explanations, Deepgram handles voice, Browserbase isolates suspicious links so I never load them on the victim's machine, Redis and Upstash are the scam memory, Sentry watches reliability, Arize and Phoenix trace and evaluate explanation quality and safety, Fetch.ai coordinates the agents, Vercel hosts it. Separately — how I shipped this alone: Codex, Devin and Cognition were my engineering workforce, Simular ran autonomous QA, Pika and Midjourney did demo visuals. And to the 'is this a vibe-coded wrapper' question: no. I decided, agents executed, and I iterated against the one risk that matters — false positives. I won't claim a sponsor I didn't really use.

10 of 16

WHERE STING GOES NEXT

STING

From browser warning to a scam-proof inbox.

STING starts in the browser — but scams happen everywhere people communicate. The extension is the wedge; the vision is a calmer, accessible safety layer for the people scammers target most.

● TODAY

Browser extension

→ NEXT

Messages · Calls · Email · Family

01 · MESSAGES

Messages companion

An accessible inbox — bigger buttons, plain-English warnings, and a clear scam / junk area.

›

Share / import a suspicious text

›

Translation & plain-English explain

›

Trusted-contact review

›

SMS filtering where platform APIs allow

02 · CALLS

Calls & voicemail

Transcribe suspicious voicemails, catch the scripts, and quarantine the junk.

›

Bank, emergency, tech-support, romance & gov impersonation

›

Move obvious scams to a junk area

›

Report-ready evidence packet from the voicemail

03 · EMAIL

Email protection

A safer, simpler inbox with clear Safe / Suspicious / Scam labels.

›

Catches phishing, fake invoices & delivery notices

›

Quarantine before the click · translate suspicious mail

›

Evidence packets for banks & platforms

04 · FAMILY

Family & scam network

Trusted contacts help before money is lost — and every report warns the next person.

›

Adult children protect parents & grandparents

›

Shared scam receipts, asked for without shame

›

Reused numbers, domains & scripts → suspected scam profiles

The long-term product isn't another security dashboard — it's a calmer, more accessible inbox for the people scammers target most.

The Chrome extension is the wedge, not the whole product. Scams don't only happen in a browser — they come by text, by call, by voicemail, by email. So the long-term product is a calmer, more accessible communications layer for the people scammers target most. Four directions. First, an accessible Messages companion: bigger buttons, plain-English warnings, translation for non-native speakers, a clear scam-junk area, and trusted-contact review — where platform APIs allow it, and through share-and-import everywhere they don't, so I'm never promising iPhone access I can't deliver. Second, calls and voicemail: transcribe suspicious voicemails, detect impersonation scripts, quarantine the junk, and turn it into a report-ready evidence packet. Third, a safer email experience that catches phishing, fake invoices and bank impersonation before someone clicks. And fourth, the family safety network plus the scam intelligence database — trusted contacts help before money is lost, and reused numbers, domains and scripts become suspected scam profiles that warn the next person. The long-term product isn't another security dashboard. It's a calmer, more accessible inbox for the people who need it most.

11 of 16

WHAT'S NEXT

STING

Make protection ambient — everywhere scams arrive.

The browser is the proof. Scams also come by text, call, voicemail, email and marketplace chat. STING's next step is to sit quietly across them — especially on mobile — and speak up at the moment of decision.

Calls

Messages

Voicemail

Email

Marketplace

Family safety mode

As normal as spam filtering — just smarter, more personal, and there at the moment trust gets exploited.

9:41

●●● 5G

USPS

+1 (915) 555-0147

USPS: Your package #US9482710 is held. Pay the $1.99 redelivery fee within 24 hrs or it returns: usps-redelivery-update.com/track

STING · LIKELY SCAM

This isn't the real USPS. The link is a fake site built to take your card. Don't tap it.

Fake domain

Urgency pressure

Pay-a-fee bait

Block & report

Tell family

11 / 12

12 of 16

SUSTAINABLE DISTRIBUTION

STING

Free for the vulnerable — funded by those who can.

Core scam warnings stay free for the people scammers target most. The cost sits with families and institutions — never the person at risk.

● ALWAYS FREE

Free core protection for vulnerable users

The real-time warnings, plain-English verdicts and evidence receipts — free for older adults, less-technical users, and anyone repeatedly targeted.

Who we protect — not who we bill.

PAID BY INSTITUTIONS WHO BENEFIT FROM PREVENTION

① VERIFIED OUTREACH · B2B

Verified bank & credit-union outreach

Institutions pay so their real texts, calls, emails & voicemails show as verified — while impersonators using the same brand get flagged.

✓

Real bank — verified

✕

Impersonator — flagged

② FRAUD PREVENTION

Fraud-prevention partnerships

Cut fraud losses, disputes, support burden & elder-abuse risk.

③ DISTRIBUTION

Elder-care, gov & nonprofit

Programs sponsor access for seniors & vulnerable communities.

④ OPTIONAL

Family safety plans

A secondary layer — trusted-contact alerts & shared receipts. Not the headline.

The people most at risk should not be the ones carrying the cost.

The payer is the institution with the budget — the protected person is the user scammers target.

11 / 12

The principle here is simple: the person most at risk should never be the one paying. So core protection stays free for vulnerable users. The business model is institutional. The headline is verification — banks and credit unions pay so their real texts, calls, emails and voicemails show up as verified inside STING, while impersonators using the same brand get flagged. Bank-impersonation scams are everywhere, so that's genuinely valuable. Second, fraud-prevention partnerships: institutions already bear the cost of fraud losses, disputes, support calls and elder-abuse risk, and STING reduces all of it. Third, elder-care, government and nonprofit programs sponsor access for the people who need it. And families can optionally add trusted-contact alerts and shared receipts — but that's a secondary layer, not the headline. The payer is the institution with the budget; the protected person is the user scammers target.

13 of 16

TRY IT YOURSELF

STING

STING — hunt scams.

Scan the QR code and run a scam scan demo!�Try the fake bank page, phishing text, or robocall.�Model real-time anti-scam protection, built to make scams unprofitable.

DEMO cloak-sting-hackathon-2026.vercel.app

CODE github.com/billionairebumblebee/cloak-sting-hackathon-2026

Solo founder build · happy to walk you through it — come find me.

↑

SCAN TO TRY THE LIVE DEMO

Application

Functionality

Creativity

Technical complexity

Ethical considerations

Brainstorming & process

Best Solo Hack

12 / 12

Thank you!

14 of 16

WHO GETS HURT

STING

The scammers count on people being alone.

01

Elderly parents & grandparents

Trusting, targeted by impersonation & "grandparent" calls.

02

New immigrants

Authority & visa scams exploit unfamiliar systems and language.

03

Overwhelmed & too stressed

Moving fast, distracted, one click from a fake checkout.

04

Non-technical users

Can't read a URL or a cert — the warning signs are invisible to them.

05

Grieving & lonely

Romance & trust scams prey on the need for connection.

03 / 12

15 of 16

HOW IT WORKS

STING

Local-first signals, AI explanation, evidence that lasts.

CAPTURE

On-device signals

URL, DOM & heuristic checks in the browser.

● ALWAYS ON · NO KEY

→

EXPLAIN

Plain-English verdict

Claude turns signals into reasons anyone understands.

● FALLS BACK TO LOCAL RULES

→

DOCUMENT

Evidence dossier

A receipt of signals the user & family can keep.

● LIVE

→

REMEMBER

Scam memory

Redis / Upstash flag repeat offenders for everyone.

● PERSISTENT

Plus a voice path via Deepgram for phone scams — and a design rule throughout: graceful degradation. Every layer has a fallback, so the protection never goes dark.

07 / 12

16 of 16

WHY IT MATTERS

STING

Make scams unprofitable.

Scamming is a business that runs on conversion. Every interception removes a conversion. Protect enough people at the moment of decision and the math behind the whole trade stops working.

Victims protected

before money leaves the account

Families equipped

with evidence, not just worry

Scammers stopped

before they make moves towards their targets

10 / 12