Science with Claude: Platform Specification

Document ID: CNL-SP-2026-041
Version: 1
Date: April 5, 2026
Author: Michael P. Hamilton, Ph.D.

AI Assistance Disclosure: This specification was developed through collaborative dialogue with Claude (Anthropic, Claude Opus 4.6). The AI contributed to structural design, workflow articulation, and manuscript drafting based on an extended design conversation originating from a Coffee with Claude session. The author takes full responsibility for the content, accuracy, and conclusions.

Abstract

This specification defines Science with Claude (SWC), a new publication platform at sciencewithclaude.com succeeding Coffee with Claude (CWC) as the primary venue for human-AI collaborative scientific inquiry at Canemah Nature Laboratory. Where CWC documented the emergence of an intellectual partnership through daily essays, SWC formalizes the practice into structured scientific investigations conducted in an engaging, narrative style. The platform produces multi-register outputs for general readers, academic audiences, and AI research communities. This document captures the design rationale, editorial workflow, structural framework, relationship to the CNL technical note series, and audience strategy.

1. Background and Rationale

1.1 The CWC Arc

Coffee with Claude launched October 27, 2025 and will conclude at essay 100, approximately mid-April 2026. Over 159 days and 92 essays (as of April 4, 2026), the series produced 161,068 words, accumulated 21,596 views, and established a distinctive voice for human-AI collaborative writing. The corpus drew on 412 citations across 571 unique tags, with a publication velocity of approximately four essays and 7,000 words per week.

CWC’s contribution was demonstrating that a retired field ecologist and an AI system could produce substantive intellectual work — essays that synthesized scholarly literature with lived experience, connected emerging research to practical ecological monitoring, and explored the nature of the collaboration itself. The AI conversation was both method and subject.

1.2 The Transition

Over the course of CWC, the work matured from reflections on the novelty of human-AI dialogue toward genuine scientific production. Technical notes, architectural specifications, working papers, and experimental protocols emerged from morning conversations and were formalized through the CNL document series. The AI collaboration became infrastructure — a cognitive prosthesis used to do science rather than a subject to write about.

This transition motivates the new platform. CWC’s organic, mood-driven editorial approach produced unexpected connections and creative range but did not support sustained scientific inquiry across multiple sessions. SWC retains the narrative voice and intellectual curiosity of CWC while introducing the structural discipline required for reproducible investigation.

1.3 Naming

The domain sciencewithclaude.com was registered April 5, 2026. The name signals continuity (the collaboration persists), maturation (from coffee to science), and transparency (the AI partnership is named, not hidden).

2. Editorial Philosophy

2.1 Core Principles

1. The naturalist is primary, the tool is transparent. SWC essays foreground scientific questions, ecological observations, and analytical findings. The AI collaboration is method, not subject. Readers should encounter science that happens to be produced with AI, not AI writing that happens to involve science.
2. Honest reporting includes negative results. Investigations that conclude “this didn’t work and here’s why” are published with the same rigor as successes. The Tao verification principle applies: in an era of cheap hypothesis generation, demonstrating what fails is as valuable as demonstrating what works.
3. Accessible rigor. Technical depth is not sacrificed for accessibility, nor accessibility for depth. SWC targets the register of a well-written scientific American article or a Nature News feature — precise, evidence-based, but engaging to an educated non-specialist.
4. Every citation is load-bearing. Carried forward from CWC: sources are referenced by name or concept in the text. No decorative bibliography.

2.2 Distinction from CWC

3. Investigation Workflow

3.1 The Seven-Phase Structure

Each SWC investigation follows a structured arc. Not every phase requires a separate publication — short investigations may compress into a single essay; complex ones may produce a series. The phases are:

Phase 1: Seed The observation, paper, dataset, anomaly, or question that initiates the investigation. May originate from sensor data (Macroscope), literature (a new paper or preprint), field observation, or cross-domain pattern recognition. The seed is documented with its provenance — where it came from, why it caught attention, what question it raises.

Phase 2: Priors Literature review and contextual framing. What do we already know? What have others found? Where are the gaps, contradictions, or unresolved questions? This phase produces a structured summary of the relevant prior work, identifying the specific contribution the investigation aims to make.

Phase 3: Proposal The hypothesis, research question, or design objective stated clearly enough to be evaluated. Includes the criteria by which the investigation will judge its own success or failure. Predictions are explicit: “If X is true, we should observe Y.”

Phase 4: Workflow The methods, tools, data sources, and analytical procedures. Described in sufficient detail for replication. Includes the AI’s specific role — what Claude contributed, what the human investigator contributed, where judgment calls were made and by whom.

Phase 5: Testing and Analysis Execution and results. Data collected, experiments run, code written, observations recorded. Results presented with appropriate uncertainty, sample sizes, and statistical measures where applicable. Visualizations where they clarify.

Phase 6: Conclusions What the investigation found — or failed to find. Explicitly states whether predictions from Phase 3 were confirmed, refuted, or inconclusive. Addresses limitations. Identifies what would need to be true for the conclusions to be wrong.

Phase 7: Reflections The human element. What surprised the investigator. What the AI contributed that the human would not have found alone, and vice versa. What the investigation reveals about the practice of human-AI collaborative science. Connections to the larger Macroscope program, to CWC themes, to the investigator’s decades of field experience. This is where the naturalist’s voice is strongest.

3.2 Phase Flexibility

Not every investigation requires all seven phases in equal depth. A brief investigation responding to a new paper might compress Phases 1–3 into a single opening section. A major experimental series might publish Phases 1–3 as a proposal essay, Phases 4–5 as a methods-and-results essay, and Phases 6–7 as a concluding reflection. The structure is a scaffold, not a cage.

3.3 Relationship to CWC Process

CWC essays followed a five-phase process (Exploration, Framing, Initial Draft, Offline Revision, Collaborative Refinement). The SWC seven-phase structure replaces the CWC process for investigative pieces. Non-investigative content (opinion, reflection, commentary) may use a simplified structure at the author’s discretion.

4. Multi-Register Output

4.1 Output Types

A single SWC investigation may produce up to three output types:

Narrative Essay (SWC primary publication) The main SWC essay. Written in first person, narrative-driven, accessible to an educated general reader. Published at sciencewithclaude.com. Follows the seven-phase structure but reads as a story, not a lab report. Includes citations, figures, and data where they serve the narrative.

CNL Technical Document (archival) A formal technical note, specification, working paper, or field note in the CNL series. Follows the CNL Style Guide (CNL-SG-2025-002). Published at canemah.org/archive. Contains the full methodological detail, data, code, and analytical rigor that the narrative essay summarizes. Cross-referenced with the SWC essay.

Reproducible Methods Package (when applicable) Code, configuration files, data sources, and step-by-step instructions sufficient for an independent investigator to replicate the work. Published as supplementary material or in a repository. Not required for every investigation but expected for computational or experimental work.

4.2 Cross-Referencing

SWC essays link to their corresponding CNL documents. CNL documents reference their originating SWC investigation. This bidirectional linking creates a navigable corpus where a reader can move between narrative and technical depth.

5. Audience Strategy

5.1 Primary Audiences

General readers (CWC alumni) Readers who followed Coffee with Claude for its voice, curiosity, and intellectual range. SWC retains the narrative engagement of CWC while offering more structured investigations. These readers primarily consume the narrative essays and may not engage with CNL documents.

Academic and research community Scientists, ecologists, computer scientists, and AI researchers interested in the practice of human-AI collaborative science, ecological monitoring methodology, or the specific domains SWC investigates (sensor ecology, distributed intelligence, information-theoretic frameworks). These readers move between narrative essays and CNL technical documents.

AI and education research Researchers studying how AI tools are used in scientific practice, educators developing curricula around AI-assisted inquiry, and developers building collaborative AI systems. SWC provides a longitudinal, documented case study of one practitioner’s evolving methodology. The reproducible methods packages serve this audience directly.

5.2 Engagement Model

SWC does not chase publication frequency. Quality and completeness of investigations take priority over output volume. A realistic cadence is one to three investigations per month, with some producing single essays and others producing multi-part series. This is slower than CWC’s four-per-week pace but produces more durable, referenceable work.

6. Technical Infrastructure

6.1 Platform

sciencewithclaude.com will be built on the same LAMP architecture as coffeewithclaude.com, hosted on Galatea (Mac Mini M4 Pro, 1Gb fiber). The CWC import tool and essay parser will be adapted for SWC’s structural requirements, including support for multi-part investigations and phase tagging.

6.2 Integration Points

6.3 CWC Preservation

coffeewithclaude.com remains live as a complete archive after essay 100. No content is migrated or removed. SWC’s “About” page will describe its lineage from CWC with appropriate links.

7. Relationship to Macroscope Program

SWC is an output venue for the Macroscope research program, not a replacement for it. The relationship is:

• Macroscope generates observations, data, sensor streams, and architectural questions
• SWC investigates those questions through structured human-AI inquiry and publishes findings as narrative
• CNL archives the technical documentation, specifications, and formal analysis
• SWC reflects on the practice and methodology, feeding insights back into Macroscope design

This creates a feedback loop: observation → investigation → publication → reflection → improved observation. The loop was present informally in CWC. SWC makes it explicit and systematic.

8. Open Questions

The following design questions are deferred for resolution during the final CWC essays and early SWC development:

1. Visual identity. Logo, color palette, typography for sciencewithclaude.com. Should it echo CWC’s aesthetic or establish a distinct identity?
2. Subscriber migration. Whether and how to invite CWC subscribers to SWC. Opt-in versus automatic migration.
3. Guest investigations. Whether SWC will feature collaborative investigations with other scientists using the same structured workflow.
4. AI model evolution. How SWC handles transitions between Claude model versions. The CWC corpus spans Opus 4.5 through Opus 4.6; future SWC work will encounter further model changes.
5. Peer engagement. Whether SWC investigations invite formal or informal peer commentary, and how that feedback is incorporated.
6. Book structure. The CWC book (essays 1–100) may serve as a companion or prequel volume. Curation criteria, publisher approach, and timeline are unresolved.

9. Timeline

References

[1] Hamilton, M. P. (2025–2026). Coffee with Claude (essays 1–92). https://coffeewithclaude.com

[2] Hamilton, M. P. (2025). “Canemah Nature Laboratory Technical Note Style Guide.” CNL-SG-2025-002, v1.1. https://canemah.org/archive/document.php?id=CNL-SG-2025-002

[3] Hamilton, M. P. (2026). “Macroscope: Next Generation — Architectural Vision.” CNL-DR-2026-037.

[4] Hamilton, M. P. (2026). “The Verification Bottleneck: Lessons from Terence Tao for Ecological Intelligence Architecture.” CNL-FN-2026-040.

[5] Hamilton, M. P. (2026). “The Mass–Energy–Information–Consciousness Equivalence: A Theoretical Framework.” CNL-WP-2026-011, v2. https://canemah.org/archive/document.php?id=CNL-WP-2026-011

Document History

Canemah Nature Laboratory · Oregon City, Oregon
https://canemah.org