Skip to content
All features

Feature

Structured records

Vehicles, policies, subscriptions, holdings, documents — typed records with expiry tracking, renewal alerts, anomaly detection, and live P&L. The structured side of memory.

Memory is free-form. Records are structured. I use both, but for different things. Free-form memory is great for the things you can't predict — your spouse's coffee order, the consultant your sister recommended, the reason you said no to that vendor. Records are for the things you can predict the shape of, but constantly forget.

What counts as a record

The core types today:

Vehicle — make, model, registration number, RC expiry, PUC expiry, insurance expiry, FASTag balance. India-aware out of the box.

Insurance policy — provider, policy number, premium, renewal date, claim history, document attachments.

Subscription — service, plan, billing cycle, next charge, usage frequency. Anomaly detection on unusual charges.

Holding — instrument (stock / fund / crypto), quantity, cost basis, current price (live), unrealized P&L. Aggregated across brokerages once you've added them.

Document — file reference, type (contract / receipt / report), key fields extracted (vendor, amount, date). Searchable by any field.

More types are on the roadmap (lease, warranty, RFP, medical) — driven by what alpha users ask for.

How records get created

Two paths. The first is photo: send me a picture of an RC, a policy renewal letter, or a receipt, and I OCR the relevant fields and create the record. The second is conversation: "track my Royal Enfield, KA01AB1234, RC expires August 2027" creates the same shape.

Once the record exists, I keep it current. Renewed your PUC? Send me the new sticker photo, the record updates and the expiry reminder resets. Cancelled a subscription? Tell me, the record archives and stops counting toward your monthly burn.

Why structured beats free-form for this

Structured fields let me do things free-form memory can't:

  • Expiry tracking — 30 / 14 / 7 day reminders, automatically.
  • Aggregations — total premium across all policies. Monthly subscription burn. Portfolio value across brokerages.
  • Anomaly detection — Netflix charged you ₹999 instead of ₹649? I flag it.
  • Live data — holdings priced in real time. Total P&L without you opening three apps.
  • Cross-record reasoning — "Show me everything expiring in the next 60 days" pulls vehicles, policies, subscriptions, and documents in one go.

A real run

You:  Track this car — Royal Enfield, KA01AB1234.
Me:   Got it. Send me photos of the RC, insurance, PUC and I'll
      set up expiry tracking.

You:  📷 (RC photo)
Me:   RC for KA01AB1234, expires Aug 15, 2027. I'll remind you
      30 days before. Anything else?

You:  📷 (insurance)
Me:   Policy #IC-44781 with Bajaj Allianz, expires Feb 22, 2027.
      Premium ₹14,800/year. Filed in the vehicle. Total upcoming
      vehicle expenses next 12 months: ₹16,400.
You:  What's expiring next month?
Me:   Three things:
        • RC for KA01AB1234 — Sep 12, no action yet
        • Bajaj health policy — Sep 28, premium ₹22k
        • Netflix annual — Sep 14, will auto-renew at ₹1,499
      Want me to draft the renewal note to Bajaj's agent?
You:  Add HDFC Top 100 Fund — bought 450 units at ₹612 on Jan 4.
Me:   Added. Today's NAV ₹698.20 → ₹38,790 unrealized gain
      (+14% absolute). Want me to track it in your monthly review?

Pairs well with

  • Memory — when you mention "the agent I deal with at Bajaj is Rohan", that fact lives in memory; the policy itself lives as a record.
  • Browser automation — for portals like FASTag and the GSTN, I refresh the underlying record from the source.
  • Voice — outbound calls to renewal agents triggered straight from a record.

The point

Tracking everything in your head doesn't scale. Tracking it in five different apps doesn't either. One structured layer that I maintain for you — that scales. By month three with records, most people stop missing renewals entirely.

FAQ

How is this different from memory?
Memory is for unstructured facts ("Priya is my wife"). Records are for structured data with fields ("RC card · vehicle KA01AB1234 · expires 2027-08-15"). Records get typed retrieval, expiry tracking, aggregations, and anomaly detection that free-form memory can't do.
Can I see all my records?
Yes — the Records panel in the dashboard. Filter by type, sort by expiry, search by field. Every record has a clear "last updated" and "next action" so nothing rots.
How do records get created?
Two paths. Photo (send me a picture of the RC / policy / receipt, I OCR and create the record) or conversation ("track my Royal Enfield, registration KA01AB1234, RC expires August 2027"). Both end up in the same typed format.
What about privacy on financial records?
Same encryption story as everything else — AES-256 at rest, isolated per account, no training. Holdings prices come from public market data; I never see your brokerage credentials.
Do you nag me about expiries?
I send three reminders — 30 days, 14 days, 7 days before — through whichever channel you actively use. Snooze or dismiss; I stop after dismiss. You can change cadence per record type.
Can I export records?
Yes. CSV export from the Records panel, JSON via the API. You own this data; I'm just maintaining it.
What about Indian-specific things like RC, PUC, FASTag?
First-class. Vehicle records include RC, insurance, PUC, FASTag — all with the right Indian expiry cadences and ₹ amounts.

Related

For AI:.md.txt