Hard questions.
Honest answers.

You can get a response from ChatGPT about autism. But ChatGPT does not know your child.

Every time you open ChatGPT, you start from zero. It has no memory of what your child's triggers are, what their communication level is, what happened last week, or what strategies have worked before. You have to re-explain everything every time — and then hope the response accounts for the full picture.

AutiPower is built around a persistent child profile. It knows your child's age, communication level, sensory sensitivities, interests, triggers, and behavioral history. Every response is generated in the context of that specific child — not a generic autistic child, not a Western textbook example, not a statistical average.

The AI engine underneath is Claude (by Anthropic) — one of the most capable models available today. But the engine is not the product. What AutiPower adds on top is the structure, the child context, the behavioral frameworks, and the memory. That layer is what general-purpose AI completely lacks.

This is one of the most honest and important criticisms of AI-based tools — and you are right to raise it.

Most autism research, therapy frameworks, and published behavioral literature is dominated by Western, English-language sources. That bias exists in AI training data too.

Here is what AutiPower is doing about it:

First, the platform is designed to receive parent input as ground truth. You are the expert on your child and your cultural context. The AI's job is to reason with what you provide — not to impose a Western framework on top of it.

Second, this is exactly why the platform is open. We are actively inviting therapists, educators, parents, and ABA practitioners from India, the Middle East, Southeast Asia, and Africa to contribute real-world behavioral knowledge. The system improves as more cultural and contextual knowledge is added.

Third, responses are generated in your language — not translated after the fact, but reasoned and structured natively in your preferred language. A parent in Telugu or Arabic receives guidance in that language, framed for that context.

We are not claiming to have solved this. We are claiming to have built the infrastructure to solve it together — openly, with contributions from the global community.

Yes — and this is a concrete example of where general-purpose AI fails and AutiPower is designed to do better.

When you set up your child's profile, you can specify the school system, class structure, the level of support available, and what the school expects. That context is passed to the AI with every response.

A child navigating a 3-tier school system in India — with different expectations at home, school, and tuition — faces transition challenges, social dynamics, and routine disruptions that are very different from a child in a single-institution Western system. AutiPower can be calibrated for that reality.

This is also an area where parent input shapes the platform. If your child's situation is not well-handled today, tell us. The system is built to evolve from exactly these real-world gaps.

This is the right question to ask, and you deserve a direct answer — not a marketing statement.

Here is exactly what happens with your data:

Your child's profile is stored in Supabase — a Postgres database with row-level security. This means the database is structurally enforced so that your data can only be accessed by your account. Not by other users, not by the platform owner, not by any admin without deliberate database-level access.

When you send a message, the child profile context is sent to the Claude API (Anthropic) to generate a response. Anthropic's API does not use your data to train models — this is stated in their API usage policy. The data goes in, a response comes out, and it is not retained for training.

We are not selling your data. There is no advertising model. There are no third-party analytics trackers. The platform is not built on a monetization model that treats your child's information as an asset.

The codebase is open. You can read exactly what is sent to the AI and what is stored. That is the transparency that "trust us" can never replace.

Fair observation — and yes, technically accurate at a surface level.

The frontend is Next.js (React). The AI responses come from the Claude API. The database is Supabase (Postgres). None of that is hidden.

But the same observation could be made about most software products. Gmail is "just a JavaScript frontend with email servers behind it." What matters is what the system is structured to do.

What AutiPower adds on top of raw AI access:

— A persistent, structured child profile injected into every AI request — A behavioral prompt engine that frames questions using autism-specific frameworks — A 5-part structured response format (Behavior Understanding → Why it Happens → Immediate Actions → Learning Opportunities → Long-term Development) — Quick action triggers designed for crisis moments (meltdown, sensory overload, transition refusal) — Behavioral event logging and pattern tracking over time — Language-native responses in 10+ languages — Row-level security so your data is structurally protected

You could replicate parts of this yourself with API access and prompt engineering. But most parents are not software engineers — and even engineers do not want to rebuild this every time they need support at 2am.

The platform exists so that the infrastructure is already built, open, and available.

This is a technical issue we are actively fixing, not a security risk.

During early development, the authentication system (Google login) requires the callback URL to be registered in both Supabase and Google's OAuth configuration. When running locally (localhost), that URL must be explicitly whitelisted.

The redirect to an unexpected URL is the auth system failing gracefully and returning to the login page with an error code — it is not a redirect to a third-party or malicious site.

Once the platform is fully deployed at autipower.com, this issue disappears entirely. The production domain will be whitelisted, verified, and the authentication flow will work cleanly end to end.

If you see this error on the live site after deployment, please report it immediately and we will treat it as a critical fix.

Yes. This is not just allowed — it is the entire point.

The platform is designed on the assumption that the most valuable autism parenting knowledge does not sit in research papers. It sits with parents who have spent years in the trenches — figuring out what works for their specific child, in their specific culture, in their specific school system.

There are several ways to contribute:

— Share behavioral scenarios and what worked, so we can expand the AI's knowledge base — Report gaps where the AI's guidance did not match your real-world situation — Contribute language translations and culturally specific context — If you are a therapist, ABA specialist, or educator, contribute structured frameworks directly

The codebase is open on GitHub. The behavioral knowledge layer is designed to grow. Every real-world experience shared makes the system more accurate for the next parent who faces the same situation.

Yes — openly and without restriction.

AutiPower is built as public infrastructure, not a proprietary product. The codebase is available on GitHub. Any developer, school, non-profit, therapy organization, or government program can take this foundation and build on top of it.

You can: — Fork the project and deploy your own instance for your community — Extend the child profile schema for your specific therapeutic framework — Add new languages and cultural behavioral contexts — Build specialized interfaces for schools, clinics, or community centers — Integrate the AI engine into your own existing platform

There is no licensing fee. No partnership agreement required. No approval process. If you are building something that helps autism families, take it and build.

We only ask one thing: if you build something valuable, share it back so others can benefit too.

The honest answer is that the roadmap is shaped by parents, not by a product team working from assumptions.

The immediate priorities are:

1. Stable authentication and deployment at autipower.com 2. Reliable AI chat with full child profile context 3. Behavioral event logging and pattern tracking 4. Expanding language support beyond the current set 5. Incorporating real parent feedback into the AI knowledge base

Beyond that, the direction depends on what the community identifies as the highest need. Features suggested by parents in the field — like the 3-tier school system context, or culturally specific behavioral frameworks — are exactly the kind of input that shapes what gets built next.

This is not a product on a fixed roadmap. It is a platform on a mission — and the mission is defined by the families it serves.