64: The AI Assistant That Lives in Your Notifications

Jan 30, 2026

Dola is a floating AI that waits inside your notifications, ready to answer, summarise, translate, or even book dinner while you stay inside whatever app you’re using. No account, no ads, no subscription, just a persistent bubble that turns every text field into a ChatGPT prompt.

First Launch & Setup

Install is 28 MB. Grant Notification Access and Draw Over Other Apps; a 20-second tutorial shows how to summon the bubble. You’re done. no email, no phone number, no cloud sign-in.

How It Works

Summon: long-press text, tap “Ask Dola” bubble, or swipe the notification tile.
Ask: type, speak, or paste a question.
Answer: Dola replies in a floating card; tap to copy, insert, or share.
Stay Invisible: no “sent” status, no read receipt, no app-switching.

Core Talents

Chat: general Q&A, creative writing, code snippets, math.
Summarise: walls of text, PDFs, web pages (via share-sheet).
Translate: 100+ languages; replies auto-detect input.
Search: live web results with source cards.
Book: tables, flights, tickets via plug-in APIs (OpenTable, Skyscanner).
Voice: two-way speech; answers spoken aloud when headphones are connected.

AI Engine

Default: GPT-4o mini via Azure East Asia.
Optional: Gemini Pro, Claude 3.5, local Llama-3 8B (downloaded on-device).
Switch engines per query; no account needed.

Privacy Posture

No analytics SDKs detected.
Queries are sent to the chosen engine; Dola does not store them.
Local Llama runs fully offline; cloud models delete context after 30 min.
Notification text is parsed on-device; only the user-selected snippet is sent.

Performance & Battery

30 s voice chat consumed 12 MB data.
Background service uses 90 MB RAM; no measurable battery drain over 8 hrs.
Offline Llama adds 3 GB download but answers without data.

Customization

Bubble color, size, opacity, keyboard height.
Quick-tile for instant summon.
Auto-translate in specific apps (WhatsApp, Telegram, IG DM).
Schedule replies (premium).

Monetization

Completely free; no ads, no watermark.
GPT-4 full instead of mini
Schedule send
Unlimited web search
Custom plug-in hooks (Zapier, Make)
Free tier is not throttled; premium is purely optional extra power.

Limitations

Notification access can conflict with battery optimisers on MIUI/One UI.
Live web search needs data; offline Llama is general-knowledge only.
Scheduled replies require premium.
No desktop companion yet.

Pros:

System-wide AI bubble without account or ads.
Multiple engines (cloud + local) switchable per query.
Voice in/out, live web, translation, booking in one float.
Tiny background footprint, no battery drain.
Free tier is full-featured; premium is optional turbo.

Cons:

Notification permission is a red flag for some users.
Live features need data; offline model is 3 GB.
Scheduled replies locked behind weekly subscription.
No tablet or desktop client.
Occasional bubble overlap over full-screen games.

Who Should Install

Heavy chatters who want instant translation/replies without blue ticks.
Students who summarize articles on the fly.
Travellers booking tables or flights while inside Maps.

Who Can Skip

Users who refuse Notification Listener on principle.
People who need desktop round-trip or collaborative docs.
Those on capped data who binge voice queries.

Final Frame

Dola is the closest thing to having J.A.R.V.I.S. inside your notification shade, free, frictionless, and increasingly indispensable. If you’re comfortable granting one more accessibility toggle, it turns every text field into a super-power. Just remember: once you’ve asked AI to book dinner mid-Instagram, going back to copy-paste feels like dial-up internet.