DoomLess AI
Digital Nutrition for Your Feed
Analytics
!!!
How to Install the Chrome Beta
Live Instagram extension overlay
Live Instagram extension overlay
Extension attention dashboard
Explained Digital Nutrition score
Try it for yourself
Attention >>
Thank you!!

DoomLess AI: Digital Nutrition for the Modern Web

Executive Summary

DoomLess AI is an AI-powered Digital Nutrition platform that helps people evaluate short-form content before investing their attention.

Unlike tools that block applications or report screen time afterward, DoomLess evaluates individual content while the user browses.

The working MVP is a privacy-first Chrome extension for Instagram Reels on desktop. It detects the active Reel, analyzes available evidence, and displays an explainable Digital Nutrition Label.

Each label provides:

A 0–100 Digital Nutrition Score
Seven explained content metrics
Content classification and evidence confidence
Attention-cost and Return on Attention estimates
Positive signals, warnings, and missing information
A recommendation: Watch, Save, Entertainment, Limit, or Skip

Existing tools measure screen time. DoomLess measures content quality.

Category	DoomLess AI
Problem	Screen-time totals cannot distinguish useful learning from repetitive or manipulative content.
Solution	Explainable Digital Nutrition Labels for individual Instagram Reels.
MVP	Chrome extension, secure AI pipeline, fallback scoring, personalization, caching, history, and dashboard.
Innovation	Evaluates content quality before or during consumption instead of only restricting time afterward.
Vision	A Digital Nutrition Layer for the modern web.

Inspiration and Problem

DoomLess began with a familiar experience: opening Instagram for one useful Reel and unintentionally entering a long scrolling session.

Short-form feeds place education, entertainment, advertising, repetition, and clickbait inside the same frictionless interface. Likes and comments indicate popularity, but they do not explain usefulness, credibility, relevance, emotional effect, or attention cost.

Twenty minutes of intentional learning and twenty minutes of scroll bait therefore appear identical in a screen-time report.

Typical wellbeing tools	DoomLess AI
Measure application-level time	Evaluates individual content
Block access or limit sessions	Supports decisions while browsing
Focus on reducing usage	Focuses on improving attention quality
Treat content inside an application similarly	Separates learning, entertainment, relevance, and risk

The idea follows a familiar analogy: food labels reveal nutritional qualities before eating; Digital Nutrition Labels reveal likely content value before watching.

The goal is not to block social media. It is to transform passive scrolling into informed choice.

The Solution

When a supported Instagram Reel becomes active, DoomLess extracts only available surface evidence, including:

Caption
Creator
Hashtags
Visible text
Accessibility labels
Duration
Displayed metadata

It evaluates seven Digital Nutrition dimensions:

Learning Value — Knowledge, context, insight, or skill gained.
Actionability — Useful steps, advice, frameworks, or decisions.
Personal Relevance — Alignment with local interests, goals, profession, and avoided topics.
Time Efficiency — Useful value relative to duration and filler.
Emotional Impact — Constructive, neutral, stressful, or manipulative framing without diagnosis.
Clickbait Risk — Exaggerated hooks, misleading claims, urgency, or weak delivery.
Addiction Risk — Repetition, looping, overstimulation, urgency, or compulsive engagement patterns.

The interface converts these signals into a score, classification, confidence level, Return on Attention estimate, supporting evidence, and one recommendation:

Watch — Worth immediate attention.
Save — Useful, but better consumed later.
Entertainment — Intentional comedy, dance, creativity, music, or relaxation.
Limit — Mixed value; consume selectively.
Skip — Weak, repetitive, or misleading content.

Personalization affects Personal Relevance without changing intrinsic quality or hiding behavioral risks. Repeated low-value viewing can trigger a gentle pause rather than blocking Instagram.

Transparent Scoring

Let:

(L) = Learning Value
(A) = Actionability
(R) = Personal Relevance
(T) = Time Efficiency
(E) = Emotional Impact
(C) = Clickbait Risk
(D) = Addiction Risk

Positive Value is:

$$ P = 0.25L + 0.20A + 0.20R + 0.20T + 0.15E $$

Risk Penalty is:

$$ Q = 0.10C + 0.15D $$

The raw score is:

$$ S_{\mathrm{raw}} = P - Q $$

The raw range of (-25) to (100) is normalized to a published score from (0) to (100).

Initial recommendation bands are:

(80–100): Watch
(65–79): Save
(45–64): Limit
(0–44): Skip

Evidence confidence, content classification, Return on Attention, and explicit risk rules can override the initial band. For example, very high Addiction Risk combined with low Learning Value can force a Skip recommendation.

The score is a transparent product heuristic—not a scientific, medical, or psychological measurement.

How We Built It

Chrome Extension

The Manifest V3 extension uses TypeScript, React, Vite, and Tailwind CSS.

A MutationObserver detects dynamically loaded Reel containers, while an IntersectionObserver identifies the active Reel.

Analysis runs only when:

A Reel becomes actively visible
The user hovers for at least 800 milliseconds
The user explicitly selects Analyze

Defensive metadata extraction avoids depending on protected media URLs. Stable hashing, request deduplication, caching, cleanup logic, and an isolated Shadow DOM overlay improve performance and resilience.

Service Worker and Backend

The background service worker checks:

User consent
Private-content restrictions
Versioned cached results
Existing in-flight requests
Timeout state

The secure Next.js/Vinext backend validates inputs with Zod, keeps API keys server-side, sends supported evidence to GPT-5.6, requires strict structured output, validates the response, and applies deterministic TypeScript scoring rules.

AI proposes structured analysis; validated application policy owns the published decision.

Reliability and Fallbacks

If evidence is limited, the API fails, the request times out, or the returned output is invalid, DoomLess uses a cautious local surface scorer.

The interface explicitly identifies missing transcripts, frames, audio, pacing, or looping evidence rather than pretending full-video analysis occurred.

Demo mode follows the same schema, scoring policy, overlay, caching, and user-interface pipeline as the live experience.

Local Data and Dashboard

Consent, preferences, cached analyses, history, and user actions remain in Chrome local storage by default.

The dashboard reports:

Reels analyzed
Average Digital Nutrition Score
Useful versus skipped content
Saved content
Clickbait and addiction-risk exposure
Viewing trends
Estimated attention reclaimed
Recent decisions

Building with Codex

Codex acted as an end-to-end engineering partner across:

Product architecture
Reel detection
Defensive DOM extraction
Shared schemas
Scoring policy
React interfaces
Fallback analysis
Personalization
Caching
Dashboard analytics
Testing and accessibility
Documentation and deployment
GitHub and hackathon preparation

During live-browser debugging, Codex helped trace inconsistent Instagram metadata, strengthen selectors, replace unhelpful “unknown content” states, align the extension with the beta website, and document honest evidence limitations.

The experience showed that AI-assisted development is most powerful as a continuous engineering partnership—not only as isolated code generation.

Challenges and Responses

Challenge	Response
Dynamic Instagram DOM	Layered detection, defensive selectors, observer cleanup, stable identifiers, and demo mode.
Incomplete evidence	Evidence-aware local estimates, confidence labels, and explicit missing-data disclosures.
Explainability	Score bands, classifications, reasons, strongest signals, attention-cost language, and “Why This Score?”
Performance and cost	Active-Reel gating, hover delay, duplicate prevention, timeouts, hashing, and caching.
Privacy	Opt-in analysis, local storage, server-side credentials, excluded private data, and deletion controls.
Non-moralizing design	Entertainment is recognized as legitimate value while clickbait and addiction risks remain visible.

Accomplishments and Learnings

The working MVP can:

Detect live Instagram Reels
Extract available surface evidence
Generate structured AI analysis
Display an interactive Digital Nutrition Label
Explain every score
Cache results and prevent duplicate requests
Store user actions
Personalize relevance
Support a reliable demo mode
Detect repeated low-value viewing
Summarize behavior through a dashboard

We learned that confidence must be visible, personalization must remain separate from intrinsic quality and behavioral risk, and human-centered wellbeing should preserve autonomy instead of relying on shame or blanket blocking.

Novelty

DoomLess introduces Digital Nutrition: a content-level language for value, relevance, emotional effect, behavioral risk, and attention cost.

Its novelty includes:

Evaluating individual content instead of entire applications
Measuring content quality rather than only time
Supporting decisions before or during consumption
Separating intrinsic quality, personal relevance, and behavioral risk
Recognizing intentional entertainment as legitimate value
Explaining every recommendation
Making attention cost visible
Interrupting low-value scrolling loops without blocking the platform

Future Vision

Instagram Reels are the current proving ground. The same Digital Nutrition framework could later support:

Other short-form feeds
Articles and news
Podcasts
Educational and professional media
Creator-quality tools
Family and student experiences
Browser-level content evaluation

Future development could add permission-based multimodal analysis, legitimate transcript ingestion, multilingual labels, privacy-preserving on-device analysis, better-content replacement, and a public Digital Nutrition API or standard.

Closing

Food consumption changed when nutrition information became visible. DoomLess applies that principle to the digital information diet while preserving the user’s final choice.