Logo/Thumbnail
Today Dashboard
Real-Time Scam Call Detection
AI Messaging & Approval
Trusted Circle
Accessibility & Multilingual Support

Inspiration

Scammers go after older adults on the internet constantly, paying for ads and sending emails pretending to be Medicare or their bank. It just happened to my grandma last month: she spent four hours on the phone with a scammer and lost over $90,000.

The typical suggestion is letting family handle everything. My grandma refused. That's when we realized we could build a unified interface where older adults can do things themselves, in a controlled and safe environment, with someone they trust one tap away, so they never spend hours speaking to the wrong person.

We're not the only ones who see this gap. Y Combinator's Fall 2026 Request for Startups names AI for the Aging Population as an area it wants startups to build in (ycombinator.com/rfs). It points out that by 2030, one in five Americans will be over 65, that there aren't nearly enough people to care for everyone, and that 53 million family members are already doing this work unpaid. Almost no technology, even voice assistants, is built with older adults in mind. LifeLink is aimed at exactly that: a voice-first interface that keeps older adults safe and independent, with family close by.

What it is

LifeLink provides a voice-first UX for older adults.

Speak or type → review the agentic action → approve → benefit

Stop scams

Every new email is labeled Likely a scam, Warning signs, or No warning signs, with the exact reasons, so you learn the patterns too.
You can also check a text or a photo, or put a call on speakerphone for live captions and a spoken warning if a scam is detected.
When a scam is detected, one tap prepares a message to someone in your Trusted Circle, ready to send once you approve it. Family you trust can also see and monitor scam alerts proactively, without ever seeing the message itself.
Help me understand breaks down any confusing message or email: what it's asking, the dates, what's missing, and a few replies you could send.

Stay connected

Message family over text (iMessage), email, Discord, or Telegram, so you can reach people anywhere in the world from one simple interface, with all of our anti-scam features built in. Or send a single message to several people across several platforms at once.
Ask "Can Sam drive me to my appointment Tuesday?" LifeLink writes the message, and when Sam answers, it tells you whether that's a yes, a no, or a different time. It shows the weather for that day and adds the appointment to your calendar once you approve.
View photos and documents that friends and family send you.
Stay up to date on stock and commodity prices to keep an eye on your retirement investments and 401(k).

Your day

Read my day reads you the weather, who wrote to you, and whether any recent email looks like a scam, so nothing catches you off guard.
Weather shows up next to each appointment, so you know when rain or ice might change plans.
Choose uplifting news from Good News Network and Smithsonian, or world news from NPR. Tap a headline for a quick summary right in the app, so you never have to visit an outside site. No ads and no endless feed, just the latest news.
How I like help lets you pick short or detailed answers, one step at a time, your preferred language (10 languages, including English, Spanish, Chinese, Korean, Vietnamese, and more), theme, font, text size, ElevenLabs voice, and more.

Groceries

Paste a list from a text or email and it adds everything at once.
Items are sorted by aisle in the order you walk the store, with cold and frozen items last so they stay cold.
Set a budget, see the estimated total, save a usual list, reuse past trips, and print or copy the list for whoever's shopping.

Fun

Solitaire and Mahjong: free, no ads, and no sign-up.

All of it works with read-aloud, large text, high contrast, and ten languages.

How we built it

Next.js, Convex, and Better Auth, in a Bun + Turborepo monorepo. The web app runs on Vercel, and the Discord bot runs as a systemd daemon on a Vultr Debian VPS, deployed with one command.

Mercury 3 summarizes news articles, so users get safe news without visiting ad-infested external sites. It's a diffusion model, so summaries load almost instantly.
Gemini 4 Argon powers the chatbot.
Gemini 3.8 Live handles live call transcription and all speech-to-text (talking to our agents, dictating messages, and more).
GPT-6 Luna classifies scams. It's fast and cheap enough to run on every email, call, and text.
ElevenLabs v4 Turbo provides the text-to-speech voices for read-aloud, in each user's preferred voice and language.

The AI can't act on its own. Every approval is tied to a fingerprint of the exact action:

$$ \text{execute} \iff H(d_{\text{approved}}) = H(d_{\text{current}}) \ \land\ \text{permissions still hold} $$

If anything changes after you approve, you have to approve again.

Challenges we ran into

Our Discord bot (for LifeLink ↔ Discord messages) couldn't easily run on Vercel like the rest of the project, so we spun up a Debian VPS on Vultr and hosted it as a systemd daemon.
Instacart closed new developer applications and had no MCP with the tools we needed, so we cut it.
Facebook and Instagram MCPs turned out to support only business accounts, not personal ones.
Long translations kept getting cut off at large text sizes.

Accomplishments that we're proud of

Real-time scam screening on live calls. My grandma was on the phone with a scammer for four hours. We're helping prevent that for your grandmas and grandpas, and maybe even your parents.
One accessible interface to learn instead of a dozen. Weather, stocks, email, texts, calendar, documents, photos, news, groceries, and games, all in one place with the same scam protection.
Built for how older adults actually use tech. Big text, big buttons, read-aloud in their own language, and a Practice mode to try everything with fake data first.
Family can help without taking control. You pick exactly what each person can see, they can never act for you, and you can take access away instantly.