Platforms & Utilities
Starting development

Peiron

A planned, user-funded search engine that keeps sources visible and makes substantive disagreement easier to inspect.

Category
Platforms & Utilities
Platform
iPhone + Safari
Status
Starting development

Search the web without surrendering the frame.

Conventional results come first. When a query contains meaningful disagreement, Peiron can add an optional source-grounded map of materially different arguments without imposing a default AI conclusion.

User value

Sources first. Perspective when useful.

Find relevant results quickly, then inspect how serious arguments differ.

01

Use ordinary search

Straightforward queries remain straightforward. Perspective tools appear only when they add value.

02

Inspect the evidence

Every visible argument stays connected to sources the user can open and assess.

03

Keep editorial agency

Peiron maps disagreement without turning one model-generated synthesis into the destination.

Features

The planned first experience.

A focused search product for finding sources and exploring meaningful disagreement.

  1. 01

    Conventional results

    Retrieve, normalize, deduplicate, and rank useful web results with direct source access.

  2. 02

    Search lenses

    Focus a query on the web, news, forums, recent material, or primary sources.

  3. 03

    Perspective overview

    See materially different claims, assumptions, explanations, priorities, or proposed actions.

  4. 04

    Missing-perspective search

    Ask what the initial result pool may be missing through a bounded follow-up retrieval.

  5. 05

    Ranking explanations

    Inspect the material factors that helped a result appear.

  6. 06

    Explicit source controls

    Block or favor domains through reversible local preferences rather than inferred ideology.

How it works

From query to evidence.

The shortest useful path remains available at every step.

  1. 01

    Ask

    Enter a query and, when useful, choose a search lens.

  2. 02

    Find

    Receive conventional, source-oriented results before slower analysis completes.

  3. 03

    Compare

    Open a perspective overview when the evidence supports meaningful disagreement.

  4. 04

    Inspect

    Move from an argument to representative sources and the evidence beneath it.

  5. 05

    Decide

    Refine, search for missing coverage, open a source, share it, or leave.

Product concept

Arguments, not tribes.

A perspective is defined by what it argues about this query, not by a permanent label attached to its publisher.

01

Query-relative

The same source can contribute differently depending on the question and the evidence it presents.

02

Relevance before diversity

Different does not mean useful. Weak material is not promoted to manufacture balance.

03

AI remains subordinate

Models may help extract and group arguments, but sources remain visible and abstention remains valid.

Architecture

One service, replaceable providers.

This is the frozen target architecture for implementation, not a deployed system.

  1. 01iPhone app or Safari extension
  2. 02Versioned Peiron API
  3. 03Bounded retrieval adapters
  4. 04Normalization and ranking
  5. 05Optional perspective processing
  6. 06Progressive source-oriented response

Modular monolith

Clear internal modules keep deployment manageable while preserving system boundaries.

Provider-independent

Search and model vendors sit behind replaceable adapters rather than defining the product.

Graceful reduction

Conventional search should remain usable if AI processing or one retrieval provider fails.

Technology

A planned, replaceable stack.

Apple clients

  • Swift
  • SwiftUI
  • StoreKit
  • Safari Web Extensions

Backend

  • Python
  • FastAPI
  • Pydantic
  • SQLAlchemy
  • Alembic

Data

  • PostgreSQL
  • Valkey
  • iOS Keychain
  • Local device storage

Retrieval

  • Self-hosted SearXNG
  • Replaceable search adapters
  • Replaceable model adapters

Development

Definition complete. Engineering next.

Phase 0 established and froze the product, experience, perspective model, privacy boundary, target architecture, economics, and validation strategy. No executable product exists yet.

Stage
Starting development
Current milestone
Phase 0 complete
Implementation
Not started
Product media
Not available

Completed definition work

  • Thirty-four authoritative product and system definitions
  • Nine recorded definition batches with an integrity-checked archive trail
  • Frozen user experience, perspective model, privacy boundary, and first-release scope
  • Target architecture, data model, API contract, security model, and test strategy
  • Development plan and chronological journal for the transition into engineering

Next phase

Create the reproducible engineering foundation, then prove conventional retrieval, normalization, deduplication, ranking, and cost measurement before adding the perspective layer.