← Back to Obliqo

How Obliqo works

The disagreement model

Four AI agents run in parallel on the same document. They do not share findings with each other. Each agent reads the full draft independently and produces an output shaped by its own critical posture — what it is designed to look for, challenge, or protect.

This isolation is intentional. When agents share context, they converge. When they work independently, they surface different things. The value is in the divergence.

The four voices

Critical Researcher

Checks your facts against the web and finds what you missed — with sources. Looks for unsupported claims, factual gaps, missing evidence, and counterarguments you have not addressed.

Resonance Amplifier

Identifies what actually works and should not change. Not everything in a draft needs fixing. The Resonance voice protects what earns its place — the sentences, arguments, and framing that should survive the revision.

Hidden Assumption

Questions the premise, not the prose. Surfaces the unstated beliefs your writing depends on — assumptions you have not examined, logical gaps in your reasoning, and the uncomfortable angle the reader might take.

Complicit Editor

Cuts what you are too attached to cut. Identifies what weakens the piece, what can be removed without loss, and what stands between your argument and landing cleanly.

Review depth

Review depth (1–5) controls how aggressively the agents analyze your draft and which models are used.

  • Depth 1–3 — available on the free plan. Focused analysis using efficient models. Covers the core critical readings.
  • Depth 4–5 — Pro plan. Uses more demanding models. Produces annotated document view — agent comments anchored to specific passages in your text.

See the full plan breakdown →

What you receive

Each run produces:

  • Four independent voice outputs — one per agent
  • A brief synthesizing the main tensions across all four readings
  • A Revision Score: Urgent, Important, or Nice to have
  • One open question that only the writer can answer

The Revision Score is not a quality rating. It reflects the severity of what the agents found — whether the issues are structural, surface-level, or absent.

Inputs

Paste your text directly into the editor, or provide a URL and Obliqo will fetch the content. Both modes consume one run. Maximum input length varies by plan.

Try it

Start a free review →  ·  See a sample analysis →  ·  What is Obliqo? →  ·  FAQ →