Under the hood.
Your Mac
Sentient runs every night at 3 AM, as long as it's open in your menu bar & your Mac's plugged in
Your Frontier Model
frontier compute · your choice
Sentient's servers
everything we run
Files
iMessage
…
On-device model
gemma 4 E4B · on your silicon
· summarizes & understands your life
· filters out junk & sensitive data
· runs
90% of sentient's compute
your codex cli
sentient sets it up
Your ChatGPT Plus subscription
gpt-5.6 sol
or
Your Claude subscription
claude models
or
Your OpenRouter model
any frontier model
or
Your LM Studio model
runs on your mac
or
Your frontier model
your own endpoint
Your knowledge base
plain markdown · yours to edit
Morning cards
· prepared overnight
· fires only when you tap
Sidekick
· computer use
· any app, any window
Your ChatGPT & Claude
· can read your knowledge base
· only if you turn it on
There's nothing here.
no accounts
no user data
no readable user data
zero-access encrypted
ciphertext only
Hover anywhere to learn more
Our thesis.
All the AI you use today only works when you type in a prompt. Truly proactive AI is different: it has to continually go through your entire life. That takes immense compute, only possible at the frontier of local, on-device inference. And it takes your entire life, not the slice living inside Google's or Apple's walled gardens.
Sentient was born to push this frontier and break those walls. And we're just getting started.
The Inference Engine
a custom LiteRT-LM fork, running Gemma 4 E4B
Inference Optimization
kv cache reuse + flash attention + speculative decoding
Privacy Engineering
no accounts; raw data never leaves
Reading Your Real Life
whatsapp, imessage & apple notes, decoded
The 3 AM Machine
wakes your Mac overnight, lid shut
The Cloud Brain
computer use on the plain Codex CLI
This is the intelligence layer for your entire digital life.