Preparing secure sign-in...
A LITTLE SPACE, JUST FOR YOU
Welcome to AskJ.
AskJ is an AI learning companion designed specifically for kids, built with comprehensive adult safety and preference controls. Instead of relying on passive screen time to keep children busy while you manage household chores, parents can start AskJ on a phone or Echo device to turn routine moments into interactive learning opportunities. Whether curious about a new topic, deepening an everyday activity, or exploring in the classroom, AskJ turns screen time into active, supervised growth.
More members
The selected member's name, age group, relationship, additional guardrails and preferences are sent to AskJ when you start. Chat does not change saved profiles. Choose the active card again to end, or another card to switch.
Admin
Modules
Create a custom module tailored to your loved one’s interests. Make it public so others can subscribe—like an alphabet-learning module—or keep it private for your family’s use, such as exploring family traditions. As of today we just support public modules.
Devices
Set up an AskJ device, choose a profile, and load a learning module. Your loved one can then talk, explore, and learn with J directly through the device—no phone needed.
Review refreshed profile
Profiles
About
Chat
UNDERSTAND THE WAIT
Diagnostics
Measure the browser, service and information agent separately.
Private storage
Read-only Google identity and bounded app-data metadata checks. No profile contents, file or account identifiers, agent requests, Azure requests, or consent prompts. Nothing is changed or archived. Success proves metadata readability only, not profile integrity, write permission, or browser write preflight. Results stay only in this page and clear when you leave Diagnostics or the storage grant changes.
Two fixed synthetic cases, separate from your conversation. No microphone or Drive permission needed. One trace at a time; the server applies a 30-second cooldown. Navigating away cancels the request; work already submitted may still use quota.
Readiness
Emma2 streams through AudioWorklet with pitch-preserving playback speed and volume captured from the server at the start of each conversation. Unconfigured playback defaults to 1.0x. The trace never plays audio or requests microphone permission.
Milliseconds are reported only when measured; missing stages are n/a or not run. Components are nested within case and server totals: do not add overlapping timings. Agent HTTP includes the provider request, not pure model compute time. Full synthesis includes Speech credentials and HTTP. First PCM byte, when reported, is a nested milestone, not browser playback. Its aggregate sums case waits, not overall startup. The synthetic trace does not measure STT or play audio.
Your recent conversation turns
Last 16 completed turns, numeric metadata only. Client prepare includes the local voice-stop request and serialization; it never waits for playback ended or microphone context shutdown. WebSocket send means browser enqueue, not server receipt. Render means DOM update, not a measured screen paint. Playback startup requires an actual playing event or PCM render-start acknowledgement. Missing measurements remain n/a. Traces stay in page memory and clear on logoff; no prompts, answers, tokens or audio are recorded here.