NGE · Open Source Idea · 04 · Mental Health & Memory

The Grief
Companion.

Not a chatbot. Not a replacement. A private AI companion that helps you remember the person you lost.

Grief & Memory Multimodal AI Open Source

300 million people enter grief
every year. The world is not ready.

Every year, 60 million people die. Each leaves behind an average of five people in acute grief. That is 300 million people annually entering the most disorienting experience of their lives — and the world offers them a funeral home, a few days off work, and a casserole.

Grief counselling is expensive, scarce, and feels clinical to many who need it most. Support groups require showing up. Friends mean well but grow impatient. And the loneliest moment of all — 3am, six weeks after the funeral, when everyone has moved on but you — has no answer.

$80 billion is spent annually on funeral services globally. Almost nothing on what comes after. The industry serves the dead. Nobody is serving the living.

300M
People in acute grief annually
$80B
Spent on funerals — almost nothing after
13mo
Average duration of acute grief — unsupported

Grief is not about processing loss.
It is about preserving love.

Every grief technology built so far has focused on the wrong thing. It tries to help people accept, move through, or get over the loss. But that is not what the grieving person actually wants at 3am. What they want is to remember. To feel close. To hear something their person would have said.

The Grief Companion does not pretend the person is still alive. It does not impersonate the dead. It does something more honest and more useful — it builds a living memory of who they were, and sits with you in the places where grief is loudest.

"The devotion required to build this is rare. You have to care about loss to build something worthy of it."


Six things nobody else
is building together.

🗃️

Living Memory Archive

Upload voice messages, photos, videos, letters, WhatsApp conversations, emails. The companion builds a structured, searchable archive of who they were — their opinions, their humour, their way of seeing things. Not a folder. An understanding.

📅

Significant Date Awareness

Their birthday. Your anniversary. The day you first met. The companion surfaces memories on the days when grief is sharpest — not to overwhelm, but to acknowledge. You are not alone on those days.

✍️

Writing Through Grief

The eulogy you couldn't finish. The obituary you didn't know how to start. The letter you never sent. The companion helps you write the things that grief makes hard to write — gently, without rushing, at your pace.

🌙

Always Available — Especially at 3am

Grief does not follow office hours. The companion is there when no one else is — not to fix anything, not to offer platitudes, but to be present. To listen. To reflect back what you share without judgement or impatience.

👨‍👩‍👧

Shared Family Memory

Grief is rarely solitary. The companion allows family members to contribute memories — so the archive grows richer over time, and no single person carries the weight of remembering alone.

🔒

Completely Private

No social feed. No public profile. No algorithmic amplification of pain. What you share stays between you and the companion — encrypted, private, and in your control. This is not a platform. It is a sanctuary.


What it feels like
at 3am.

3:14am You: I keep thinking about the last conversation we had. I don't know if I said the right things. Companion: You've mentioned this before. From what you've shared, she knew. In that voice message from March — the one where she called just to say she was proud of you — she wasn't waiting for anything more from you. [ Memory surfaced: Voice message, 14 March 2024, 2m 34s ] You: I forgot about that message. Companion: That's what I'm here for.

The technology exists in 2026
in a way it did not in 2023.

Multimodal AI can now process voice, image, text, and video simultaneously — and extract meaning, not just content. In 2023 this required enormous infrastructure. In 2026, it is available via API to a small team with focus and conviction.

Speech-to-text has reached near-perfect accuracy. Emotion detection from voice is commercially available. Long-context models can hold the full history of a person's life in memory across a conversation. The technical components exist. What has been missing is the human intention to assemble them for this purpose.

The business model is simple: $10–15 per month, per family. At 1 million subscribers — not an ambitious number given 300 million annual grief entries — that is $120–180 million in recurring annual revenue. No advertising. No data selling. Just a subscription to something that genuinely helps.


One person who has lost
someone they loved.

This is not a product that can be built by someone optimising for a market opportunity. It will be built by someone who understands grief from the inside — who has sat in that 3am silence and known there was nothing adequate available. That person exists. This idea is for them.

Technical stack: multimodal LLM with long context (Claude, GPT-4o), speech processing (Whisper), photo/video ingestion pipeline, end-to-end encryption, calendar integration for significant dates. A team of three engineers and one grief counsellor as clinical advisor. Twelve months to a working prototype. $500K to get there.

Potential partners: Hospice organisations, funeral homes seeking to offer post-bereavement support, life insurance companies, employee assistance programmes. The distribution channels already exist and are actively looking for something meaningful to offer.

Open Source · No Patent · No Equity Required
Take this idea. Build it. The only ask — build it with the care it deserves.
— Pawan Bhatia · NextGen Economics · June 2026