about thala

4 min read

Evening Reads

Transforms a literature review into a 4-part article series suitable for publication (e.g., Substack, blog). Produces one overview article plus three deep-dive pieces.

The workflow:

  1. Validates and analyzes the input literature review
  2. Plans content structure and assigns themes to deep-dives
  3. Fetches and enriches content for each article
  4. Writes the three deep-dives in parallel
  5. Writes the overview article (informed by deep-dives)
  6. Generates header images and formats references

When to Use This

This workflow is intended for turning dense research into accessible, publishable content—a series that readers can engage with over multiple sittings.

For creating the underlying research, use Literature Review or Synthesis. For improving existing reports, use Report Enhancement.

How It Works

flowchart TD
    subgraph input["1. Validation"]
        A[Literature Review] --> B[Validate Structure]
        B --> C[Map Citations]
    end

    subgraph planning["2. Content Planning"]
        C --> D[Identify Key Themes]
        D --> E[Assign Deep-Dive Topics]
        E --> F[Choose Narrative Approaches]
    end

    subgraph fetch["3. Content Enrichment"]
        F --> G1[Fetch for Deep-Dive 1]
        F --> G2[Fetch for Deep-Dive 2]
        F --> G3[Fetch for Deep-Dive 3]
    end

    subgraph write["4. Writing"]
        G1 --> H1[Write Deep-Dive 1]
        G2 --> H2[Write Deep-Dive 2]
        G3 --> H3[Write Deep-Dive 3]
        H1 --> I[Write Overview]
        H2 --> I
        H3 --> I
    end

    subgraph finish["5. Finishing"]
        I --> J[Generate Images]
        J --> K[Format References]
    end

    K --> L[4-Part Article Series]

    style input fill:#e8f4f8
    style planning fill:#f0f8e8
    style fetch fill:#fff8e8
    style write fill:#f8e8f4
    style finish fill:#e8f0f8

The Steps Explained

1. Validation Checks that the input is a valid literature review with proper structure. Maps citations to their full references for later use.

2. Content Planning Analyzes the review to identify the most compelling themes. Assigns each deep-dive a distinct topic and narrative approach to ensure variety:

  • Puzzle: Frames the topic as a mystery or open question
  • Finding: Reports on a significant discovery or insight
  • Contrarian: Challenges conventional wisdom

3. Content Enrichment For each deep-dive, fetches relevant content from the source papers. Enriches with additional context and quotes.

4. Writing Writes the three deep-dives in parallel, each with its assigned narrative approach. Then writes the overview, which synthesizes the big picture informed by all three deep-dives.

5. Finishing Generates header images for each article. Formats all references consistently.

Inputs

InputDescriptionExample
Literature reviewMarkdown literature reviewOutput from Literature Review workflow
Editorial stance (optional)Tone guidance”Accessible to general audience”

Outputs

OutputDescription
Overview article2,000-3,000 words, big-picture synthesis
Deep-dive 12,500-3,500 words, first themed exploration
Deep-dive 22,500-3,500 words, second themed exploration
Deep-dive 32,500-3,500 words, third themed exploration
ImagesHeader images for each article
StatusSuccess, partial, or failed

The Four Articles

Overview (Published Last, Read First) The “big picture” article that introduces the series. Synthesizes major themes without going deep. Written last so it can reference and tease the deep-dives.

Deep-Dive 1: Puzzle Frames a fascinating aspect of the topic as an unresolved question or mystery. Draws readers in with intellectual curiosity.

Deep-Dive 2: Finding Reports on a significant insight or discovery. More declarative—“here’s what we learned.”

Deep-Dive 3: Contrarian Challenges assumptions or presents a counterintuitive perspective. Engages readers who think they already know the topic.

Example

Input:

  • Literature review: 12,000-word review on “Sleep and memory consolidation”

Typical output:

  • Overview: “The Science of Sleep: Why Your Brain Works While You Rest” (2,500 words)
  • Deep-dive 1: “The Mystery of Dream Rehearsal” (3,000 words, puzzle approach)
  • Deep-dive 2: “How Sleep Sorts Your Memories” (2,800 words, finding approach)
  • Deep-dive 3: “Why ‘Sleep On It’ Might Be Bad Advice” (3,200 words, contrarian approach)

Each article includes header image and formatted references.

Developer Reference

Entry point: workflows/output/evening_reads/graph.py — evening_reads_graph

Graph construction: Same file — create_evening_reads_graph()

State: workflows/output/evening_reads/state.py — EveningReadsState

Nodes: workflows/output/evening_reads/nodes/

  • validate_input.py — structure validation and citation mapping
  • plan_content.py — theme identification and assignment
  • fetch_content.py — parallel content enrichment
  • write_deep_dive.py — parallel article writing
  • write_overview.py — overview synthesis
  • generate_images.py — header image generation
  • format_references.py — citation formatting

Schemas: workflows/output/evening_reads/schemas.py — structured output models

Prompts: workflows/output/evening_reads/prompts.py — writing prompts for each article type

Editorial: workflows/output/evening_reads/editorial.py — editorial stance configuration

Backlinks