# llm-seq

**Repository Path**: phyzy/llm-seq

## Basic Information

- **Project Name**: llm-seq
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-05-29
- **Last Updated**: 2026-05-30

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# LLM Remote Sensing Segmentation Harness

This repository contains a supervised research harness for remote sensing image
interpretation and segmentation with existing large models. It is designed so a
configured Claude Code environment can clone the repository, verify readiness,
and begin the Phase 1 literature survey with explicit user gates.

## Quick Start

From the repository root:

```bash
ruby scripts/check_ready_for_supervised_run.rb
```

Expected output:

```text
READY for supervised Phase 1 run
Next gate: user confirmation to begin the systematic literature survey sprint.
```

After readiness passes, ask the user for confirmation to begin Phase 1
`survey_sprint`.

## Main Files

- `CLAUDE.md`: project instructions for Claude Code supervised execution.
- `harness/remote_sensing_segmentation_research.yaml`: source-of-truth control
  file for phases, gates, artifacts, and iteration policy.
- `docs/harness/remote_sensing_segmentation_research.md`: operator guide.
- `research/iteration_1/`: Phase 1 survey and deep-dive artifacts.
- `designs/iteration_1/`: Phase 2 technical scheme artifacts.
- `experiments/iteration_1/`: Phase 3 experiment artifacts.
- `reviews/iteration_1/`: iteration review artifact.
- `paper/`: Phase 4 outline and claim-evidence artifacts.

## Validation

Run the full local validation suite:

```bash
ruby test/validate_research_harness_test.rb
ruby test/check_ready_for_supervised_run_test.rb
ruby scripts/validate_research_harness.rb --check-artifacts
ruby scripts/check_ready_for_supervised_run.rb
```

## Supervision Gates

The harness requires user confirmation before:

- updating seed literature or anchor papers,
- selecting the primary technical scheme,
- starting a major pivot,
- framing final paper claims,
- deciding final submission or archive readiness.

## Current Status

The repository is ready to start Phase 1 after user confirmation. The first
action is an agent-led systematic literature `survey_sprint`, followed by a
high-interaction `interactive_deep_dive`.