Good answers need
complete control and the right context.

ChatGPT and Gemini hide the settings and mix everything about you into every answer.
With dequid you set the model's memory, knowledge and behavior — separately, for each thing you do.

+Custommake your own
Always Free. Upgrade for More.
Personas

Different worlds. Different personas.

Each persona is the three fully customizable things the in-product editor exposes — Behavior (how it acts), Memory (what it's picked up about you), Knowledge (what you've handed it). Switch personas, and all three switch with you.

Behavior
How it acts.
  • SemanticsDirect · Analytical · Technical
  • Probability · Executiontemp 0.2 · all turns
  • Constraints · SecurityMarkdown · Comprehensive · standard
Memory
What it picked up about you.
  • ProfileSenior engineer · Go + React · 11 yrs
  • MemoriesPostgres-migration playbook · p99-latency wins
  • ContextProject X stack · sprint conventions
Knowledge
What you handed it.
  • VaultADRs folder · RFC archive · code-review notes
  • ChatsProject X retros · sprint planning · 1:1s w/ Sam
  • PeopleSam (IC4 backend) · Priya (Tech Lead) · Marco (VP Eng)
See how it works →
The second problem

Not how it sounds.
What it decides.

Two pieces of work in the same week. Your AI blends them, and both answers look clean.

Every fact votes

A note from the other analysis counts toward this one. Nothing marks it as borrowed.

VS
Only this project

Each persona weighs the files and facts you gave it. Nothing else reaches the answer.

Why this happens →
The framework

LLMs are stateless. So every dial matters.

Four Dials. Eight Layers. Every input, deliberate.

An LLM is stateless — every answer is rebuilt from scratch using only what you send in the prompt and what you set in the API call. Most tools tweak the prompt and call it "prompt engineering." With dequid, every dial in both is set — Four Dials, Eight Layers, per persona.

SSensitivity
Emotional register
EExploration
Reasoning mode
EExpressiveness
Output style
KKnowledge
Domain expertise
two views of the same persona
P
Profile
Who you are — minus the bits that don't belong in this conversation.
R
Role
The expertise you want at the table. Engineer. Strategist. Writer.
O
Output
A paragraph or a checklist. Long or tight. Markdown or plain.
M
Memory
What gets recalled — and what stays out. No leakage from your other lives.
P
Processing
The shape of the thinking. Analytical. Lateral. Step-by-step.
T
Temporal
When and where the question is being asked from. Date, hour, locale.
E
Expressive
The tone. Calibrated for this audience — not the average of everyone you talk to.
D
Domain
The documents and people that apply. Only this slice. None of the others.

Four dials, eight layers — what each controls. The full options for every SEEK dial live on how it works.

And every layer above isn't just set — it's visible. dequid shows you the exact prompt the model got, layer by layer, with the real settings. Every other AI chat interface hides it.

Honest pricing

Two plans.
No maths homework.

Same product either way. Pro is $7.99 a month with ~690K LLM credits included — every feature, and fuel to run it on.

Free

Try dequid

$0

always free

+ top up credits any time

Starter credits on the house so you can try personas, the vault, and memory. No card required.

  • Storage: 100 MB
  • History: Unlimited
  • Starter credits included
  • All features
  • Top up anytime
Start free →

Pro

Made for daily use

MOST PICKED
$7.99 / mo

billed monthly

~690K credits included

~690K credits included every month, the best rate on anything beyond that, and 10× the storage.

  • Storage: 1 GB
  • History: Unlimited
  • ~690K credits every month
  • Best credit rate
  • All features
  • Auto-top-up
See full pricing →

Stop sounding the same to everyone.

Open dequid in your browser. The right context for whoever you're talking to.

Start free →Mac app — join the waitlist →
dequid — Better Questions, Better Answers