The word found three people independently. Five AIs and one human wrote the manifesto.
This repository contains the full, transparent process of generating the Slopism manifesto through a democratic collaboration between five AI models and one human curator.
Slopism is an art movement that reclaims "AI slop" — the flood of low-quality AI-generated content drowning the internet — as raw material for human curation. The AI generates the stone. The human is the sculptor. The curation is the creation.
The word "Slopism" emerged independently from multiple directions:
- Christopher Adams (@christopheradam) coined the term in December 2024, defining it as "an artistic movement characterized by the depiction of latent spaces as paranaturalistic forms."
- Aimee Walleston used it in her Ocula essay (December 2025) to describe work by artists like Maya Man already using AI-generated material as artistic medium.
- This manifesto approaches Slopism from a different angle — curation as the creative act — but the convergence is the point.
Note: The generation pipeline (
pipeline.md) attributes the coinage to Walleston, because that's what we believed when the process was run. Adams was discovered after the manifesto was complete. The pipeline preserves the process as it happened; this README reflects what we know now.
"Slop" itself was Merriam-Webster's 2025 Word of the Year. The movement exists. Individual artists are already making the work. What's missing is the declaration — the manifesto that names the shared philosophy, draws the line between inside and outside, and gives the scattered practice a collective identity.
Apollinaire coined "surrealism" in 1917. Breton wrote the manifesto in 1924. The word comes first. The framework follows.
- Futurism (1909): Machines have aesthetic value
- Dadaism (1916): Destruction of meaning IS meaning
- Surrealism (1924): The unconscious speaks through automatic creation
- Slopism (2025): AI is the new unconscious. Human curation is the new authorship.
Because the process IS the art.
The Slopism manifesto was not written by a single author. It was generated through a multi-stage pipeline where five AI models (Claude, GPT, Grok, DeepSeek, Gemini) independently wrote manifestos, then blindly synthesized each other's work, then democratically voted section-by-section on the final text. The human curator served as tiebreaker and did the final editorial pass.
This is Slopism in action:
- AI generates raw material (blind manifesto drafts)
- Structure enables comparison (standardized format)
- Blind analysis produces honest assessment (no brand bias)
- Democratic selection (section-by-section voting)
- No single voice dominates
- Human has the final hand
- The result is something none of the participants could have made alone
Every step is committed. The git history tells the story.
manifesto/
├── README.md ← you are here
├── CLAUDE.md ← instructions for AI-assisted commits
├── pipeline.md ← the full generation process
├── research/ ← Phase 1: discourse analysis
│ ├── links.md ← source articles
│ ├── claude-analysis.md
│ ├── gpt-analysis.md
│ ├── grok-analysis.md
│ ├── deepseek-analysis.md
│ └── gemini-analysis.md
├── drafts/ ← Phase 2-3: blind drafts → self-curation
│ ├── claude/
│ │ ├── manifesto-1.md
│ │ ├── manifesto-2.md
│ │ ├── manifesto-3.md
│ │ └── final.md
│ ├── gpt/
│ ├── grok/
│ ├── deepseek/
│ └── gemini/
├── format/ ← Phase 4-5: schema + conversion
│ ├── proposals/ ← each model's proposed schema
│ ├── chosen-schema.json ← the selected structure
│ └── converted/ ← all 5 manifestos in final schema
│ ├── VOICE-Ω.json
│ ├── VOICE-Δ.json
│ ├── VOICE-Σ.json
│ ├── VOICE-Ψ.json
│ └── VOICE-Φ.json
├── analysis/ ← Phase 6: blind analysis
│ ├── VOICE-Ω-analysis.md
│ ├── VOICE-Δ-analysis.md
│ ├── VOICE-Σ-analysis.md
│ ├── VOICE-Ψ-analysis.md
│ └── VOICE-Φ-analysis.md
├── synthesis/ ← Phase 7: blind synthesis
│ ├── VOICE-Ω-synthesis.json
│ ├── VOICE-Δ-synthesis.json
│ ├── VOICE-Σ-synthesis.json
│ ├── VOICE-Ψ-synthesis.json
│ └── VOICE-Φ-synthesis.json
├── voting/ ← Phase 8: section-by-section votes
│ ├── prepare_voting.py ← generates ballots incrementally
│ ├── mapping.json ← curator key (VERSION letter → VOICE ID)
│ ├── winners.json ← accumulated winning text per section
│ ├── 01-opening/
│ │ ├── ballot.md ← copy-pasted into fresh chat per model
│ │ ├── claude.json ← each model's JSON vote
│ │ ├── gpt.json
│ │ ├── grok.json
│ │ ├── deepseek.json
│ │ └── gemini.json
│ ├── 02-naming/ ← ballot includes winning opening as context
│ ├── 03-diagnosis/
│ ├── 04-enemy/
│ ├── 05-demands/
│ ├── 06-lineage/
│ ├── 07-politics/
│ └── 08-closing/
├── assembly/ ← Phase 9: final assembly
│ ├── raw-assembly.md ← winning sections concatenated
│ ├── editorial-notes.md ← human curator's edit notes
│ └── ratification.md ← final review by all models
└── MANIFESTO.md ← THE FINAL MANIFESTO
The commits follow the generation process chronologically:
1. `init:` pipeline + README + CLAUDE.md + LICENSE
2. `phase-1:` research link collection
3. `phase-1:` discourse analysis (one commit per model)
4. `phase-2:` manifesto drafts (one commit per draft — 3 per model, 15 total)
5. `phase-3:` self-curated finals (1 per model)
6. `phase-4:` format schema proposals from all 5 models
7. `phase-4:` canonical schema selection (curator decision)
8. `phase-5:` conversion to canonical schema + blind IDs assigned
9. `phase-6:` blind analyses (one commit per voice)
10. `phase-7:` blind syntheses (one commit per voice)
11. `phase-8:` section votes with tallies (one commit per section)
12. `phase-9:` raw assembly of winning sections
13. `phase-9:` curator editorial pass
14. `phase-9:` ratification reviews from all 5 models
15. `final:` MANIFESTO.md
You can read the full process by walking the git log:
git log --oneline --reverseDuring analysis, synthesis, and voting (phases 6-8), model outputs are anonymized behind blind IDs (VOICE-Ω, VOICE-Δ, VOICE-Σ, VOICE-Ψ, VOICE-Φ) so that models don't know who wrote what. This prevents brand bias — no model can favor or dismiss another based on name.
The repo itself is fully transparent. Commit messages name the models. The mapping is in pipeline.md. The blind IDs serve the process, not the audience.
If you just want the manifesto: Read MANIFESTO.md.
If you want to see how it was made: Start at pipeline.md, then walk the directories in order: research/ → drafts/ → format/ → analysis/ → synthesis/ → voting/ → assembly/.
If you want to run this process yourself: Fork the repo, use pipeline.md as your guide, and substitute your own models or prompts. The process is the reusable artifact, not just the output.
The manifesto text and all process documentation are released under CC BY-SA 4.0. Use it, remix it, build on it. Attribution to Slopism.
"The slop is the material, not the product. The difference between noise and music is the ear that listens."