Title: Annoymail Updated: A Modern Re-Architecture of Intentional Notification Friction for Digital Well-Being

Authors: A. Developer, B. Researcher
Affiliation: Applied Human-Computer Interaction Lab
Date: April 12, 2026

Abstract Email remains a primary source of both critical communication and cognitive distraction. Originally conceived as a joke or anti-productivity tool, early versions of Annoymail introduced deliberate friction (e.g., typing delays, captchas, mandatory re-reading) to discourage reactive email checking. This paper presents Annoymail Updated, a complete re-architecture that transforms the original proof-of-concept into a production-ready, cross-platform email middleware. The updated system introduces adaptive friction scoring, contextual awareness, and positive reinforcement mechanics. Empirical benchmarks show a 47% reduction in non-urgent email checks and a 31% increase in perceived message retention among beta users.

1. Introduction Conventional email clients optimize for speed: zero latency, swipe-to-archive, and push notifications. This optimization often encourages compulsive, habit-driven checking. The original Annoymail (circa 2020) inverted this logic by deliberately annoying the user before displaying a new message. However, the original implementation suffered from high user abandonment (62% within 48 hours) due to static, non-negotiable friction.

Annoymail Updated addresses three core limitations:

1. One-size-fits-all annoyance – not all emails deserve equal friction.
2. No learning capability – the system never adapted to user behavior.
3. Negative-only reinforcement – annoyance without reward led to uninstallation.

2. System Architecture

The updated system operates as an IMAP/SMTP proxy layer between the mail server and the client (mobile/desktop).

2.1 Adaptive Friction Engine (AFE) Instead of applying the same delay or puzzle to every email, AFE calculates a Friction Score (0–100) based on:

• Sender reputation (internal user history: always spam? always urgent?).
• Semantic analysis (presence of keywords like "invoice," "emergency," or "meeting confirmed" reduces friction).
• Time of day (higher friction during deep work hours, lower friction during scheduled email windows).

2.2 Intervention Modalities The system no longer relies solely on typing delays. It now supports four pluggable friction types:

• Type-to-unlock (user must type a random 6-character string).
• Single-captcha (simple visual or audio task).
• Wait-n-seconds (progressive backoff for repeated checking).
• Reflection prompt ("What is the likely intent of this email before opening?").

2.3 Positive Reinforcement Loop Crucially, when a user resists opening an email for 15 minutes after it arrives, the Annoymail Updated client displays a small reward (e.g., "Focus saved: 2 distraction credits"). Accumulated credits unlock "express mode" where friction is temporarily disabled.

3. Implementation Details

• Backend: Python 3.11 + FastAPI, Redis for stateful friction scores.
• Email handling: aioimaplib and aiosmtplib for async operations.
• Client plugins: Safari/Chrome extension, Thunderbird add-on, and a standalone macOS status bar app.
• Privacy: All text analysis is performed locally via ONNX runtime (no email content leaves the user's machine).

4. Evaluation

We conducted a 14-day field study with 120 knowledge workers (60 control using standard email client, 60 using Annoymail Updated).

| Metric | Control | Annoymail Updated | Change | |--------|---------|------------------|--------| | Daily email checks (self-initiated) | 42.3 | 22.4 | -47% | | Time spent in email (min/day) | 118 | 79 | -33% | | Reported stress (1–10) | 6.2 | 3.8 | -39% | | Correct recall of subject line after 1h | 68% | 89% | +31% |

User qualitative feedback:
"The reflection prompt stopped me from opening 10 marketing emails I didn't actually care about."
"I hated the captchas at first, but after a week I started batching my email reading."

5. Limitations and Future Work

• False positive friction: 8% of urgent emails were incorrectly delayed (mitigation: emergency unlock phrase "URGENT_NOW").
• Desktop-only for now: Mobile push notifications bypass the proxy; iOS version requires MDM profile.
• Future: Integration with LLM-based summarization to replace opening emails entirely.

6. Conclusion Annoymail Updated demonstrates that intentional, adaptive friction—combined with positive reinforcement—can significantly improve email habits without forcing abstinence. By annoying the user intelligently, the system reclaims attention for deep work. The updated architecture is stable, open-source (MIT license), and available at https://github.com/annoymail/updated.

References [1] Newport, C. (2016). Deep Work. Grand Central Publishing.
[2] Harris, T. (2019). "Time Well Spent: Reforming engagement metrics." Interactions, 26(4), 32-37.
[3] Annoymail Original. (2020). GitHub repository (archived).
[4] Lukoff, K., et al. (2021). "Designing friction for intentional mobile use." CHI Conference Proceedings.

3. The Security Flaws: Why the "Update" Isn't Enough

This is where the deep review becomes critical. Anonymail suffers from structural flaws that updates have not fixed:

• The IP Address Problem: Most web-based anonymailers do not route traffic through the Tor network or a VPN automatically. If you use the service from your home computer, the Anonymail server knows your real IP address. If the server is compromised or seized by law enforcement, your anonymity is gone.
• Sender Reliability (SPF/DKIM): Modern spam filters are incredibly aggressive. Emails sent through public relay services like Anonymail often lack proper authentication headers (DKIM signatures matching the sender). Consequently, emails sent via Anonymail frequently land in the recipient's Spam folder or are blocked entirely by corporate firewalls.
• Trust Model: You are trusting a third party completely. With ProtonMail or SecMail, the architecture is end-to-end encrypted. With Anonymail, the admin could read the message if they chose to, as the message is decrypted and re-encrypted on their server.

1. The "Laser Looper" (Auto-Follow Up)

The original Annoymail had a "Nudge" feature that sent a follow-up after 72 hours. The updated version introduces the Laser Looper. This AI-driven scheduler creates an unbreakable chain of follow-ups that increase in urgency and absurdity.

• Tier 1 (Day 1): "Checking in on this."
• Tier 2 (Day 2): "Circling back."
• Tier 3 (Day 3): "Per my email below (attached for convenience)."
• Tier 4 (Day 4): Sends a calendar invite titled "Closure: Your response to my query."
• Tier 5 (Day 5): Emails your recipient’s manager with a subject line: "Is [Name] okay?"

Early testers report a 400% increase in reply rates, but a 200% increase in HR complaints.

How to Update to Annoymail 3.7.2

If you are a current user, the update is rolling out today.

1. Desktop: Refresh your Chrome or Edge extension. Look for the little green "Enraged" icon in your toolbar.
2. Mobile: Update the Annoymail Keyboard on iOS/Android. Warning: Do not use this during a performance review.
3. Enterprise: IT admins must approve the new "Shadow CC" feature, which secretly adds a dummy BCC to your own "Smug Folder."

Pro Tip from the developers: After updating, run the "Onboarding Rage Assessment." It’s a 5-question quiz that calibrates your Annoyance Threshold. Answering "I immediately lose respect for anyone who sends a PDF when text would do" will set the AI to maximum toxicity.