Dario Amodei imagined AI compressing a century of progress in human health into a single decade. We're building it. One phone call at a time.
In his essay Machines of Loving Grace, Anthropic CEO Dario Amodei argues that powerful AI won't just make us more productive. It will fundamentally reshape healthcare, mental health, and the experience of being human.
He envisions AI compressing 50–100 years of scientific progress into 5–10 years. Not through brute automation, but through intelligence that works alongside human systems — accelerating what we already know how to do but can't scale.
The essay's most radical claim isn't about technology. It's that in a world of abundant intelligence, meaning still comes from human connection. The question isn't whether AI will be powerful. It's whether we'll use that power to make people feel more human, not less.
Millions of older Australians receive minimal meaningful interaction each day. Facilities can't hire enough staff for daily one-on-one conversations. Families worry but can't be there every day.
The gap between what vulnerable people need and what the system can deliver grows wider every year. By the time problems surface, it's too late.
Amodei writes about AI achieving breakthroughs in neuroscience and mental health — improving "baseline cognitive experience" and treating conditions before they become crises. But the essay also acknowledges a harder truth: that meaning derives primarily from human relationships, not economic labour.
The deepest health challenge isn't molecular. It's relational. And that's exactly where we started.
The most universal technology in aged care isn't a smartphone or a tablet. It's a telephone. Every person in care already knows how to use it. There's no onboarding. No digital literacy barrier. No screen to navigate. Just a voice that calls, listens, and remembers.
No apps to download, accounts to create, or buttons to press. The phone rings. They answer. That's it. Works with landlines, mobiles, and facility phones.
Text can lie. Surveys can be gamed. But the voice reveals what words hide — trembling, hesitation, flatness, warmth. Our AI hears all of it in real time.
A human care worker can have 8–12 meaningful conversations a day. Kate can have thousands — each one personalised, each one remembered, each one generating clinical intelligence.
Four of Amodei's core arguments. Four things we've already built.
Intelligence alongside clinicians, not in place of them — compressing decades of incremental improvement into years.
Every resident receives a warm, natural AI call that remembers them — giving staff clinical intelligence they couldn't generate manually.
AI-powered approaches to depression, loneliness, and cognitive decline that catch problems early through continuous monitoring.
Real-time analysis of vocal tone, pitch, pace, and sentiment. The AI hears how you feel, not just what you say — weeks before incidents.
Enhancing everyday wellbeing — not just treating disease, but improving the felt quality of life for everyone.
Longitudinal wellbeing data. Gradual mood decline, emerging patterns, and cognitive drift detected automatically. Making the invisible visible.
In a world of abundant AI, what matters most is connection, being heard, and being known.
Families stay connected through wellness summaries. Care staff get actionable insights. Residents feel heard every single day. The AI doesn't replace relationships — it creates the conditions for them.
Amodei imagines AI as a "virtual biologist" — a tireless intelligence that accelerates research. We built something more personal: virtual companions, each designed for a specific community of people who need to be heard.
Daily calls detecting loneliness, cognitive change, and emotional distress. Live at a national healthcare provider. Generates compliance evidence as a byproduct of caring.
Emotional support for at-risk students. Flags concerns to welfare teams before small problems become big ones. Live at Western Chances.
Goal tracking, wellbeing monitoring, and Practice Standards evidence for NDIS participants. Turns every call into proof of support delivered.
Supporting people waiting for Home Care Packages — the most vulnerable gap in the system. Daily check-ins until care arrives.
Mental health check-ins for veterans. Because no one should process trauma alone. Detects distress and escalates to support services.
Emotional intelligence has no language barrier. Italian, Arabic, Mandarin, Cantonese, Spanish, Portuguese, French, German, Hindi, Japanese.
Plus Bloom (postnatal) • Grace (carers) • Ruby (social housing) • Ember (disaster recovery) • Sage (corrective services) • Dusty (FIFO workers) — each seeking pilot partners.
Behind every companion is Kate: an agentic intelligence layer that orchestrates the entire care pipeline. She's not a chatbot. She's the system Amodei describes — AI that operates autonomously on complex tasks, working alongside human systems at scale.
Schedules at the right time for each person. Selects the right persona. Loads full context from every prior conversation. Configures conversational pathways with branching logic, conditions, and guardrails.
Real-time emotional signal processing. Conversation adapts to mood. Mid-call actions: database checks, appointment booking, CRM updates, escalations, SMS to family — all while the person is still on the line.
Transcription and recording. WHO-5 wellbeing scoring. Compliance evidence. Trend analysis. Cognitive drift detection. Daily risk scoring (0–100). Everything the care team needs — generated automatically.
Cognitive drift detection, speech latency tracking, vocabulary change analysis, sentiment drift alerts, daily risk scoring. Catches decline weeks before it becomes an incident.
Real-time distress detection, auto-escalation, warm handover (3-way call), family distress alerts via SMS, longitudinal sentiment tracking, safety protocol triggers.
WHO-5 wellbeing scoring, structured JSON clinical data, trend analysis over weeks and months, case conference flagging, exportable compliance reports.
Every call maps to Quality Standards. Evidence generates as a byproduct of caring, not an admin task. Accreditation-ready documentation. Always audit-ready.
Wellness summaries for families via nonni.ai. SMS notifications when concerns arise. Family debrief calls. Voice messages from residents. Families know how loved ones are really doing.
Auto task creation from calls, smart staff assignment, priority escalation, voice task notes, overdue detection, on-call staff routing, after-hours management.
Every persona remembers details from previous conversations. Context carries forward automatically. The AI knows their name, their stories, their preferences, their patterns.
Right model for each task. Local LLMs for routine analysis (cost-optimised). Frontier models for advanced reasoning. Proprietary voice intelligence for emotion detection.
Amodei identifies five areas where AI could compress a century of progress into a decade. CarePlans AI sits at the intersection of the two most human.
AI as a "virtual biologist" — accelerating disease treatment, drug discovery, and preventive care. CarePlans contributes: early detection of cognitive decline, health trend monitoring, and proactive intervention through daily voice data.
Curing or treating most mental illnesses through behavioural interventions and neural measurement. CarePlans contributes: real-time emotional detection, loneliness intervention, WHO-5 scoring, and daily companionship that measurably reduces isolation.
AI accelerating health interventions globally and improving food security and climate outcomes. Adjacent: our multi-model cost-optimised architecture makes daily care calls economically viable at population scale.
Strengthening institutions and expanding citizen participation. Adjacent: automated compliance evidence reduces regulatory burden while improving care quality transparency.
Meaning derives from human relationships, not economic labour. This is our core thesis. Care workers shouldn't spend their days filling forms. They should spend them caring. Kate handles the rest.
Amodei wrote his essay from the perspective of a researcher imagining what's possible. Andrew Payne built CarePlans AI from the perspective of a son who watched his father age and wished technology could have helped.
Every feature exists because he asked: "What would I have wanted for Dad?"
That question became a company. The company became a platform. The platform is now live — making daily calls at a national healthcare provider and supporting at-risk students at Western Chances.
Not a concept deck. Not a pitch. Real calls. Real people. Real care.
If you're in aged care, NDIS, health, education, or government — let's talk about what a daily caring voice could do for the people you serve.
Start a conversation