---
Silent Infinity launches on AWS Bedrock with an open-source crisis-detection module, verifiable safety records, and a deliberate ban on every dark pattern the industry has normalized — starting with six users
---
CONTACT
Harnoor Minhas, Founder
harnoors@gmail.com
silentinfinity.com
---
[City, CA — April 21, 2026] — Silent Infinity (silentinfinity.com), a contemplative AI chat application built by independent developer Harnoor Minhas, today announced its public availability alongside the open-source release of its crisis-detection module, compliance with California SB 243, and a safety archive anchored on the Bitcoin blockchain via OpenTimestamps. The app has six users and 189 conversations logged. That number is stated plainly because honesty about scale is part of the product.
---
In January 2026, Character.AI reached a settlement over allegations that its product contributed to user harm. The settlement followed years of documented concern about AI companions designed primarily around retention metrics. Silent Infinity was designed in direct response to that pattern.
The product does not have streaks. It does not send push notifications. It has no leaderboards, no engagement scores, and no algorithmic nudges toward dependency. The design metaphor is a mirror — a tool that helps users see themselves more clearly and then put it down. Minhas describes the stance plainly: "I built the thing I would actually want to use. Everything the wellness AI space does to keep you hooked, I tried to remove. What's left is just the conversation."
That constraint is structural, not aspirational. The monetization model does not depend on session volume. There is no mechanism in the product that benefits from a user returning when they don't need to.
---
Silent Infinity runs on AWS Bedrock, using Claude Sonnet 4.6 for primary conversations and Claude Haiku 4.5 for supporting workloads. Bedrock's managed inference layer eliminates direct model-hosting overhead and provides the enterprise-grade availability appropriate for a wellness context, where uptime failures are not merely inconvenient.
Crisis detection runs as a standalone module — open-sourced at github.com/silentinfinity/crisis-detection under MIT license — containing 20 clinically-sourced regex patterns. The module is designed to be AFSP-alignable (American Foundation for Suicide Prevention) and can be adopted by any developer building in this space. Releasing it as a public good, rather than a product differentiator, was a deliberate choice.
Every session that triggers a crisis-detection flag is archived with a SHA-256 hash, anchored to the Bitcoin blockchain via OpenTimestamps, and Merkle-verifiable. This means the safety record is tamper-evident and auditable by regulators, researchers, or users themselves — not merely claimed in a privacy policy. The published /safety page documents the full methodology.
The application is CA SB 243 compliant, the California law establishing baseline standards for mental health digital products.
---
The crisis-detection module is MIT-licensed specifically to lower the barrier for other developers to implement clinically grounded safety logic. Minhas is pursuing alignment with the American Foundation for Suicide Prevention (AFSP) messaging guidelines and has structured the module to accommodate their framework without code changes.
Three internal strategy documents govern how features move from idea to deployment: a six-tier Feature Readiness Standard that gates capabilities on safety review before any user exposure; an Emergent Constellation framework that models how users grow across product tiers without coercive upsell mechanics; and a Sound-Science-Modes-Retention framework that grounds session-design choices in peer-reviewed research rather than engagement optimization.
These documents are internal because they are working instruments, not marketing. They exist because building wellness software without them is how harm happens incrementally.
---
Silent Infinity intends to charge for access. The pricing model is subscription-based, flat-rate, and not contingent on usage volume. A user who has one conversation a month pays the same as one who has thirty. The product does not benefit financially from maximizing session frequency.
No venture funding has been accepted. No engagement SLAs are owed to investors. The business model is solvent if users find the product useful enough to pay for it — and not solvent otherwise. That alignment between user value and revenue is, in the current landscape, a differentiator.
---
The next release cycle includes voice mode (AWS Transcribe + Polly + Claude Sonnet 4.6), Cognito SSO with Google and Apple sign-in, multi-chat threads with shared memory, and a mobile application via Capacitor bridging to a future Flutter native build. Each feature ships against the six-tier Feature Readiness Standard — no capability reaches users at a maturity level below tier 4.
The voice mode is the highest-stakes item on the roadmap. Prosodic cues in voice carry emotional information that text does not, and the crisis-detection module will require clinical review before voice triggers are finalized.
---
Silent Infinity is live at silentinfinity.com. The crisis-detection module is available at github.com/silentinfinity/crisis-detection. The /safety page is published at silentinfinity.com/safety.
---
Silent Infinity is an independent AI wellness application built by Harnoor Minhas on AWS Bedrock. It is designed for users who want honest, non-manipulative AI reflection — and for developers who want a reference implementation of what safety-first wellness AI infrastructure looks like in practice.
---
"We have six users and we are serious about getting this right. The scale will come or it won't. The standard doesn't move either way."
— Harnoor Minhas, Founder, Silent Infinity
---
###
This press release is intended for distribution to technology and health media, AWS startup channels, and mental health advocacy organizations. For interview requests, technical briefings on the crisis-detection module, or compliance documentation, contact harnoors@gmail.com.