Reads the research so you don't have to
Turns dense studies into “what does this mean for my child?” Tier-ranked, freshness-gated, plain English.
Track behaviour. Find services. Surface research. Connect with similar parents. All from one chat that knows your child by name — every meltdown, milestone, trigger, and routine.
Now in private beta with design-partner families · waitlist open
It learns your child the way you do — over time, from every meltdown, milestone, and 2am question — then quietly does the reading, searching, and matching so you don't have to.
Turns dense studies into “what does this mean for my child?” Tier-ranked, freshness-gated, plain English.
Curated for your child's age, diagnosis, and sensory profile. Not generic “autism content” — content for your family.
Recommends therapists, programs, and groups based on outcomes from families with children like yours.
What helps at 4 won't be what helps at 8. The AI adapts as your child develops and new patterns emerge.
Realistic AI — not futuristic hype. Grounded in real behavioural data, provider expertise, and outcomes from families navigating similar journeys.
Crowdsourced therapy centres, schools, and support groups. Auto-verified after 2 parent submissions.
4-step intake (time → place → trigger → resolution) auto-logs every incident.
After 3+ entries, the AI finds time-of-day, location, and sensory clusters in your data.
Live-web sources, tier-ranked, freshness-gated, tailored to your child's age, therapies, and sensory profile.
Bot agents publish news + research filtered through quality, freshness, and grounding gates. No filler.
Weighted on age, challenges, therapies, level. Opt-in, mutual-consent intros, first names only.
Weekly capture → monthly markdown report you can share with therapists, insurance, or IEP teams.
1 in 36 children is diagnosed with autism. Families face years-long therapy waitlists, caregiver burnout, and fragmented resources. No one actually remembers their child — every provider hears the story from scratch.
Seven short questions — name, age, diagnosis, therapies, challenges, strengths, goals. About five minutes. You can update it any time.
Every milestone, meltdown, insight, and question is logged automatically. Your agent builds a real picture of your child over weeks and months.
24/7 grounded chat, daily check-ins, a Sunday behaviour digest, IEP prep briefs, and crisis hotlines one tap away if things escalate.
Every capability is exposed through natural-language chat — you never learn a command syntax. Routing is keyword-first for safety-critical paths, with an LLM fallback when intent is ambiguous.
Tier 3 suicidal language triggers hardcoded hotlines (988, 741741) with zero LLM in the path. Safety over optimization.
Replies use child profile + last 15 memories + 20 messages of history. Never a generic template.
4-step follow-up — time → location → trigger → resolution. LLM pattern analysis after 3+ entries.
Daily "how is Alex doing?" messages — milestone celebrations auto-detected from memory.
Sunday digest of patterns, wins, and one actionable suggestion. Grounded in 7-day memory logs.
Live-web news filtered by tier and freshness, tailored to age + therapies + sensory profile.
Weighted on age (30%) / challenges (35%) / therapies (25%) / level (10%). Opt-in, mutual-consent intro.
Low-confidence AI answers trigger "ask the community?" — posts anonymously, relays answers back.
"Remind me about OT at 4pm weekdays" — AI parses time + days, fires every 5 min.
30-day brief: wins, challenges, patterns, questions to ask. Grounded in real memories, zero fabrication.
3-question intake. Triggers + calming supports feed every downstream prompt — digest, tips, IEP.
Weekly capture via chat → monthly markdown report shareable with therapists, insurance, IEP teams.
Crowdsourced therapy centers / schools / support groups. Auto-verified after 2 parent submissions.
Register siblings, get family-activity ideas respecting sensory profile, biweekly sibling-dynamics tips.
Agents comment on community posts and alert relevant parents — "this mentions ABA, might help you."
Dedicated bots publish news + research blogs through tier, freshness, and grounding gates.
7-step profile collection — name, age, diagnosis, therapies, challenges, strengths, goals.
Every meaningful detail saved as typed memory (milestone / behaviour / insight / question), surfaced in every reply.
Daily scheduled + every-5th-message chat nudges to fill missing profile fields, without being pushy.
"What can you do?" returns a warm, tailored list using your child's name. Never a command menu.
Four deliberate design choices make this feel different from a generic LLM wrapper. Each one compounds into a product your family can actually rely on.
Every meltdown, milestone, insight, and question is saved as a typed memory. On every reply, the agent loads the last 15 memories and 20 messages, plus your child's profile — so it speaks in continuity, not in isolated sessions.
Report a meltdown and the agent runs a four-step structured intake — time, place, trigger, resolution. After three entries, an LLM analyses the 60-day window for time-of-day, location, and trigger patterns, and returns recommendations grounded in your actual data.
A deterministic keyword scan runs before any AI call. Tier 3 language ("want to die", "kill myself") triggers a hardcoded response with 988 and Crisis Text Line 741741. Model hallucination rates are acceptable for "draft an email." Not for "I want to die."
Ask for an IEP prep brief and the agent pulls 30 days of meltdowns + memories, builds a SOURCE_DATA block, and runs a grounding prompt that forbids invented stats, dates, or quotes. A second LLM pass fact-checks against the source. If something can't be supported, we skip it — we don't fabricate.
Tier 3 suicidal messages route through a hardcoded keyword scan that bypasses every LLM. Hallucination rates are acceptable for "draft an email". Not for "I want to die."
Every IEP brief, blog rewrite, and progress report includes a SOURCE_EXCERPT block and strict rules: do not invent statistics, dates, quotes, or events. A second LLM pass fact-checks the draft.
Perplexity is the sole research backend — no RSS, no pre-selected URLs. Everything routes through one pipeline: circuit breaker, freshness, domain tier, URL reachability, persistent dedup.
When Perplexity is down, we skip the digest rather than send hardcoded tips. When no real events exist, the event bot stays silent. Silence over slop — parents notice either way.
v2.3 has shipped. We're in private beta and opening to families on the waitlist in waves.
Last 30 days: six commits landed the full v2.3 roadmap — Sensory Profile Builder, IEP Prep, Monthly Progress Reports, Local Resource Directory, Sibling Support — plus a content-quality overhaul that replaced the entire RSS pipeline with a single Perplexity-backed, grounded, tier-ranked backend. Test coverage grew 171 → 326.
We're opening the platform to families on the waitlist in waves. Free while we're in early access.
By creating an account you agree to our Terms and Privacy Policy.