I read 50 years of network science, then built a CRM that runs entirely in the browser

Last year, I stumbled on a paper that changed how I think about professional networks.

In 2022, researchers at LinkedIn, MIT, Harvard, and Stanford ran a causal experiment on 20 million LinkedIn users. They tracked 2 billion connections and 600,000 job changes. The finding: weak ties — people you barely know — are significantly more likely to help you land a job than your close contacts.

This wasn't new. Sociologist Mark Granovetter proposed the "strength of weak ties" theory in 1973. But the 2022 study (Rajkumar et al., published in Science) was the first large-scale causal test. Fifty years of theory, finally proven at scale.

I'm a software engineer. My first thought was: why doesn't any tool actually use this?

LinkedIn has the data but doesn't surface tie strength. CRMs like HubSpot, Folk, and Attio treat every contact the same. Nobody scores relationships by the research that says which ones actually matter.

So I built one.

The architecture decision that shaped everything

Early on I had to choose: build a traditional SaaS with a database, or go browser-only.

I chose browser-only. Here's why.

The people who need this most — recruiters, founders, career changers — are managing sensitive professional relationships. Candidate pipelines. Investor networks. Executive contacts. Asking them to upload that to my server felt wrong, and "we take privacy seriously" on a terms page doesn't mean much.

So the entire CRM runs in your browser. IndexedDB stores everything locally. No accounts.

No server-side database for contact data. No tracking. Your network graph never leaves your machine.

The tradeoff is real: if you clear your browser data, it's gone. That's why the paid tier includes encrypted backup and encrypted share-links. But the default is privacy by architecture, not by policy.

The stack

• 
  Next.js 14 (App Router) with TypeScript


• 
  Dexie.js wrapping IndexedDB for client-side storage


• 
  D3.js for the network graph visualization


• 
  Tailwind CSS for styling


• 
  Stripe for licensing (validates a key, no user accounts)


• 
  Vercel for deployment


• 
  Anthropic Claude for
  AI-drafted outreach messages (server-proxied for licensed users)

The original visualizer (V0) is open source under MIT github.com/navox-labs/network. V1 adds the full CRM layer on top.

What it actually does

You export your LinkedIn connections (Settings → Get a copy of your data → Connections). Upload the CSV. The app builds an interactive graph of your network.

Then it scores every contact by tie strength — based on the weak-ties research. The contacts you've been ignoring? Those are often your most valuable professional relationships.

The free tier gives you the graph, gap analysis (which industries are you missing bridges to?), company search, and an outreach queue.

The paid tier adds a full CRM: contacts table, pipeline, notes, tags, AI-drafted messages, follow-up reminders, and relationship timeline.

What I learned building it

IndexedDB is powerful but unforgiving. Schema migrations need careful handling when there's no server to coordinate. Dexie helps a lot, but you're still managing state that lives entirely on the user's device.

Privacy-first architecture limits your options — and that's the point. I can't run analytics on user data. I can't build recommendation models from aggregate usage. Every feature has to work with data I never see. It forces better design decisions.

The research is the moat. Any CRM can add a graph view. But implementing tie-strength scoring from peer-reviewed methodology — connecting Granovetter's theory to Rajkumar's empirical findings to actual product features — that's hard to replicate without doing the research work.

Recruiters are pragmatic buyers. 30+ recruiters reached out after the V0 Product Hunt launch asking for a CRM version. They don't care about the graph being pretty. They care about finding warm paths to candidates before paying for cold InMail.

Try it

The free visualizer is live at navox.tech/network —

upload your LinkedIn export, see your graph, find your weak ties. No account needed, no data leaves your browser. The CRM layer (contacts, pipeline, AI drafts) is a paid upgrade from there.

The original V0 visualizer is open source:github.com/navox-labs/network

The research, if you want to go deep:

• Granovetter, M. (1973). "The Strength of Weak Ties." American Journal of Sociology, 78(6).


• Rajkumar, K. et al. (2022). "A causal test of the strength of weak ties." Science,377(6612).

I'd genuinely love feedback from other devs building privacy-first tools. The browser-only constraint creates interesting problems — happy to discuss in the comments.