skeptic.
MODEL-INDEPENDENT BY DESIGN V1 / PROTOTYPE

Your work. Your memory.
Any AI.

Switch AI agents without starting over.

Try the setup preview See how it works →
01 / THE FOUNDATION

User-owned knowledge

  • ObsidianYour vault
  • MarkdownPortable notes
  • ProjectsGoals & progress
  • FilesYour source of truth
02 / THE CONTINUITY LAYER

Skeptic OS

PRE-ALPHA

The intelligence and continuity layer.

  • Knowledge discoveryImplemented prototype · synthetic data

    Find relevant passages within your chosen scope.

  • Session handoffsDocumented workflow

    Carry verified progress into the next session.

  • Preference learningPlanned · approval required

    Keep working preferences consistent across models.

  • Model onboardingPlanned · no live connections

    Give your next model a clear starting point.

  • Memory updatesPlanned · explicit permission

    Save accepted knowledge in user-owned files.

  • Context optimizationImplemented prototype · bounded passages

    Select useful context without a bulk vault upload.

03 / YOUR NEXT CONVERSATION

Your choice of AI

  • Claude CodePlanned adapter
  • CodexPlanned adapter
  • Local AIPlanned adapter
Architecture preview: your files remain the source of truth. Arrows show the intended flow, not live integrations. No automatic learning or memory updates.

Pre-alpha preview. No installation, file access or live provider connections.

THE IDEA

Your knowledge.
One lasting home.

Skeptic is not another AI. It is the independent continuity layer for supported AI agents. Own your knowledge, connect your AI tools, switch with context, and return caught up. Claude → Codex → Claude: the returning agent receives the latest accessible context, with verified Git evidence separated from documented claims. Obsidian is preferred, not required; ordinary Markdown works too. Hidden session state cannot be recovered.

WHY SKEPTIC

Less repetition. More continuity.

Keep ownership.

Readable files you control, in the tools you already use.

Repeat yourself less.

Shared preferences help the next AI understand how you work.

Keep moving.

Clear checkpoints preserve progress when you switch models.

HOW IT WORKS / THE PLANNED EXPERIENCE

Three simple steps.

01

Start with your files.

Keep your notes where they are. Markdown and Obsidian are the starting formats.

02

Choose your boundaries.

Decide what the AI may read. Reading never grants permission to edit or publish.

03

Carry progress forward.

Use a project checkpoint to show what’s done and what comes next.

The preview uses invented examples. Provider setup and live retrieval remain experimental. See what’s available →

FOUR STARTING PATHS / PREVIEW ONLY

Begin where you are.

The walkthrough adapts to your starting point. It shows what setup could involve without connecting an account or opening your files.

Claude-first

Bring your existing Markdown instructions.

Codex-first

Start from shared rules and project progress.

Local-model-first

Explore a path for an AI running on your device.

Fresh start

See how a simple Markdown workspace could begin.

Choose your path in the preview →
YOUR NEXT STEP

Your AI can change.
Your knowledge stays.

See the eight-step preview. Keep your real files where they are.

Explore the preview Read the guide →