Twinly

48% never follow up after networking. Twinly fixes it: record conversation on iPhone, AI drafts personalized emails and tracking cards. 3 taps and never miss an opportunity again.

Comment

Inspiration

I've been networking for years, and I'm always that person who meets someone amazing, gets their card, and then... never follows up. Not because I don't care—but because writing a personalized email takes 30 minutes, and after meeting 10 people, that's 5 hours I don't have.

The stat that shocked me: 48% of professionals never follow up after networking events. That's not laziness—that's a broken system. Every missed follow-up is a missed job offer, a missed partnership, a missed opportunity that could change your career.

I built Twinly because I was tired of being in that 48%. And I realized: if I'm struggling with this as someone who understands the importance of networking, imagine how many opportunities are being lost every day across millions of professionals and job seekers.

What it does

Twinly automates the entire networking follow-up workflow in three iPhone taps:

Step 1: Record your networking conversation using an iPhone Shortcut
Step 2: Twinly's AI agent (built with Toolhouse) automatically:

  • Transcribes the conversation
  • Analyzes context with Claude AI to extract key details (names, companies, topics discussed, action items)
  • Generates a personalized follow-up email referencing specific points from your conversation
  • Generates a Trello tracking card with action items and follow-up reminders

Step 3: Done. 90 seconds from conversation to complete follow-up.

The result: Instead of spending 30 minutes per person (or never following up at all), you turn every networking conversation into a tracked opportunity with personalized outreach—automatically.

How I built it

Tech Stack:

  • iPhone Shortcuts - Voice recording and transcription using iOS native speech-to-text
  • Zapier MCP Server - Workflow orchestration and webhook handling
  • Toolhouse AI Agent - Central intelligence for conversation analysis and decision-making
  • Claude API - Natural language processing for context extraction and personalized email generation
  • Composio Tool Router - Intelligent routing to Gmail and Trello APIs
  • Gmail API - Automated draft creation with personalized content
  • Trello API - Follow-up tracking cards with action items

Architecture:

  1. iPhone Shortcut captures audio → transcribes → Toolhouse
  2. Toolhouse AI agent workflow
  3. Agent uses Claude to analyze conversation and extract:
    • Contact name, company, role
    • Topics discussed
    • Action items mentioned
    • Follow-up urgency and timing
  4. Agent drafts personalized email dtemplate via Zapier MCP to connect to Gmail and
  5. Agent generates personalized email template via Zapier MCP to connect to Gmail

Challenges I ran into

1. Avoiding Generic AI-Generated Emails Early iterations produced emails that sounded robotic and generic—exactly what we were trying to avoid.

Solution: Refined Claude prompts to explicitly extract and reference specific conversation details. Added context windows to maintain conversational tone. Result: Emails that actually feel personal because they reference what was ACTUALLY discussed.

2. Solo Hacker Scope Management As a solo hacker, I initially wanted to add Snap Spectacles integration, real-time processing, LinkedIn automation, calendar scheduling... and would have finished nothing.

Solution: Ruthlessly prioritized. Built the core workflow first (iPhone → Email draft) as my MVP. Only then added Trello tracking. This "ship first, enhance later" mentality was crucial for finishing in 24 hours.

3. Demo Reliability WiFi at the venue was spotty with 2500+ hackers. Live demos risked failure.

Solution: Pre-recorded backup demo video showing the complete workflow. Practiced the two-person skit 5+ times until it was muscle memory. Had multiple conversation examples ready to process.

Accomplishments that we're proud of

🏆 Built a genuinely useful product - This isn't a demo toy. I will actually use this. That 48% stat? I was in it. Not anymore.

🎯 Solved a universal problem - Everyone networks. Everyone struggles with follow-up. Twinly democratizes relationship-building for people who don't have personal assistants.

🤖 True AI agent architecture - Not just "LLM + API wrapper." Twinly uses intelligent tool routing where the agent decides actions based on conversation context.

⚡ Extreme efficiency - 5 hours of manual work → 90 seconds. That's a 200x time savings. For job seekers networking with 50 people, that's 25 hours saved.

🎭 Memorable demo execution - The two-person networking skit (playing both Yi and Mark) made judges laugh while clearly demonstrating the product. Technical excellence + showmanship.

💪 Solo execution - Went from idea to working demo in 20 hours as a solo hacker. Shipped a complete workflow automation including AI, APIs, UI, and live demo.

What we learned

Technical:

  • Use Toolhouse - Expedited agent development via prompts
  • Claude excels at structured extraction - With proper prompting, Claude consistently extracted contacts, companies, action items, and sentiment from unstructured conversation text.
  • iPhone Shortcuts are underrated - Native iOS automation is surprisingly powerful. Speech-to-text quality rivals dedicated APIs, and the UX is frictionless (just 3 taps).

Product:

  • Simplicity wins - My initial idea involved Snap Spectacles, continuous listening, automatic detection. I simplified to "3 taps on iPhone" and shipped something that works.
  • The demo IS the product - Judges don't care about your architecture diagram. They care: "Does it work? Can I use it? Will it help me?" Show, don't tell.
  • Personalization is the killer feature - Generic "nice to meet you" emails are worthless. Twinly's value is in the specific references: "Looking forward to that intro to Lisa" or "Excited about the ML projects you mentioned." That's what builds relationships.

Hackathon Strategy:

  • 48% never follow up - One powerful stat beats a wall of text. Find YOUR stat and lead with it.
  • Have a backup - Pre-recorded demo video saved me from WiFi/technical failures. Always have Plan B.
  • Playing both characters in a skit is memorable - Judges saw 50+ pitches. They'll remember "the person who played both people in the networking conversation."

What's next for Twinly

Immediate (Next 2 weeks):

  • Public Beta - Launch to 100 beta users (job seekers, startup founders, salespeople)
  • Multi-conversation batch processing - Process 5-10 conversations at once from a career fair
  • Email template customization - Let users adjust tone (formal, casual, enthusiastic) and length
  • Calendar integration - Auto-suggest meeting times based on conversation context

Short-term (Next 3 months):

  • LinkedIn automation - Auto-send connection requests with personalized notes
  • CRM integration - Connect to Salesforce, HubSpot, Notion for professional relationship management
  • Snap Spectacles integration - True hands-free recording during conversations (the original vision!)
  • Analytics dashboard - Track follow-up completion rates, response rates, relationship ROI

Long-term Vision:

  • AI Relationship Coach - "You haven't followed up with Sarah in 6 months. Here's a reason to reconnect." Proactive relationship management.
  • Company Intelligence Layer - Auto-research contacts before meetings. "Mark's company just raised Series B. Mention their expansion."
  • Multi-language support - Enable networking across language barriers with real-time translation
  • Enterprise version - Sales teams, recruiting teams, conference organizers. "Never let a lead go cold."

The North Star:
Make it impossible to miss an opportunity. Turn every conversation into a relationship.

Right now, 48% of professionals give up after one contact. With Twinly, that number goes to zero. That's hundreds of millions of missed opportunities captured. Jobs found. Partnerships formed. Careers changed.

Built With

Share this project:

Updates