Anchor

landing_page
stats_and_footer
checking_page
working_feature
cbt_analysis_left_out
cbt_analysis_school_stress
reflecting_in_action

Inspiration

1 in 5 teens experiences a mental-health condition each year, and roughly 50% of all lifetime mental illness begins by age 14. Yet existing mental-health apps demand sign-ups, mood data, and constant connectivity — meaning surveillance, stigma, and a paywall stand between a struggling teen and help.

We asked: what if a teen in a tough moment could get meaningful support in 60 seconds, completely privately, with zero risk of their words leaking?

What it does

Anchor is not a chatbot — it's an on-device reflection engine built on Cognitive Behavioral Therapy (CBT) principles:

You check in — pick a mood and write what's on your mind, like texting a friend.
Anchor reads it — scores the emotional load and matches your words against 8 cognitive distortions from CBT (all-or-nothing thinking, catastrophizing, mind-reading, labeling, "should" pressure, overgeneralizing, discounting the positive, emotional reasoning).
It reflects back — names the thinking trap (not a label about you), offers a gentler reframe, and suggests one small actionable thing to try (box breathing, grounding exercises, reaching out).
It shows your trend — your week's moods, stored only on your device, so you can see the storm pass.

Everything runs 100% in the browser. No server, no API call, no tracking. A teen can use Anchor on a cheap phone, offline, with zero risk of their words ever leaking.

How we built it

Anchor is a single self-contained HTML file with zero dependencies:

Tone Scoring Engine: Weighted positive/negative lexicons (46 negative, 25 positive words) combined with mood selection produce an emotional-load score (0-100).
CBT Distortion Detector: Regex patterns map natural language to 8 cognitive distortions, each with a warm reframe question and kinder thought.
Coping Router: Context-aware action suggestions that match the themes in the user's entry (breathing for anxiety, reaching out for loneliness, grounding for spiraling).
Local Trend Tracker: In-memory 7-day mood chart with seeded demo data.
Design System: A calming "night-sea to dawn" palette with a living breathing orb, glassmorphic header, and staggered card animations.
Accessibility: Full ARIA attributes, `prefers-reduced-motion` support, keyboard focus styling.

The code includes a clearly marked LLM swap-in hook — point it at any endpoint and the UI keeps working while the privacy-first on-device path stays the default.

Challenges we ran into

Crafting regex patterns that accurately detect cognitive distortions without over-matching casual language was tricky — we iterated through many false positive scenarios.
Balancing the tone of reframes to be warm and non-preachy without being dismissive of real feelings.
Designing the emotional-load scoring to feel accurate across very different types of entries (short vs. long, explicit distress vs. subtle).

Accomplishments that we're proud of

A fully functional CBT-based reflection engine that runs with zero dependencies and zero network.
The app detects real cognitive distortions and provides genuinely helpful, therapist-informed reframes.
Privacy isn't a feature toggle — it's the architecture. Nothing can leak because nothing is ever sent.
The UI creates a genuinely calming experience with the breathing orb and night-sea aesthetic.

What we learned

That CBT distortions can be meaningfully detected with carefully crafted heuristics — you don't always need an LLM.
That for mental health tools, privacy IS the product — trust enables vulnerability.
That teens don't need another chatbot — they need something that helps them name what they're feeling and take one small step.

What's next for Anchor

localStorage persistence so mood trends survive browser sessions.
More distortion patterns and richer reframe libraries.
Optional LLM integration using the built-in swap-in hook for deeper personalization.
PWA support for home screen installation and true offline-first experience.
Multilingual support to reach teens globally.