Skip to content

docs(readme): lead with the problem Spacedock solves#213

Open
ijac13 wants to merge 1 commit into
spacedock-dev:mainfrom
ijac13:spacedock-readme-problem-rewrite
Open

docs(readme): lead with the problem Spacedock solves#213
ijac13 wants to merge 1 commit into
spacedock-dev:mainfrom
ijac13:spacedock-readme-problem-rewrite

Conversation

@ijac13
Copy link
Copy Markdown
Contributor

@ijac13 ijac13 commented May 13, 2026

Summary

Rewrites the README opening so a reader instantly knows whether Spacedock is for them. The previous opening described what Spacedock does; this one leads with the pain it solves, names the mechanism that answers it, and gives three reader-state matches on distinct axes (shape / trust / time horizon).

What changes

  • Opening hook moves from "Spacedock runs agent work through defined stages..." (descriptive) to "Hand an agent a multi-step job and it drifts, skips steps, or invents its own path" (concrete pain, alpha-user language).
  • Mechanism paragraph names how Spacedock answers it: stages with fresh context, explicit gates, structured reports, reviewers that push back, state outside the agent.
  • Origin sentence is added in first-person plural: we lived the pipelines that didn't finish, the reviewers that rubber-stamped, the work that died at the context limit, the routines spread across too many tools.
  • "You want Spacedock if" restructured from a 2-bullet "human / agent" framing into 3 reader-state bullets on three distinct axes:
    1. Shape: repeatable multi-step pipelines on many inputs
    2. Trust: single-agent output you can't verify on its face
    3. Time horizon: work that spans days or weeks

The rest of the README ("What's Different," "Quick Start," everything below) is untouched.

Why now

The messaging is anchored in the alpha-interview synthesis at recce-gtm/knowledge/customer-insight/interview-plans/alpha-interview-synthesis/alpha-problems.md (7 universal problem clusters from the 9-alpha corpus). The clusters that drive each piece of the new copy:

  • Cluster 5 (agents cannot follow full pipelines without scoping) → opening hook
  • Cluster 1 (single-agent loops produce poor work without an external check, n=7) → trust bullet
  • Cluster 2 (context pressure degrades agent quality, n=6) → mechanism paragraph
  • Cluster 3 (state loss across sessions, n=3) → time-horizon bullet

Test plan

  • Render README on GitHub and check formatting (no broken bullets, no em dashes leaked)
  • Read the opening cold and decide: does the pain land in the first 30 seconds?
  • Verify the three axes don't overlap (shape, trust, time horizon stay distinct)
  • Confirm the rest of the README still flows from the new opening

🤖 Generated with Claude Code

The previous opening described what Spacedock does ("runs agent work through
defined stages..."). This rewrite leads with the problem the reader is living:
agents that drift on multi-step jobs, reviewers that rubber-stamp, work that
dies at the context limit, routines spread across too many tools.

Changes:
- New opening hook frames the pain in concrete user-language (cluster-5 alpha
  pattern: "agents cannot follow full pipelines without scoping").
- Mechanism paragraph names how Spacedock answers it: stages, gates, structured
  reports, reviewers that push back, state outside the agent.
- Origin sentence is first-person plural — we lived this loop, the tool is the
  way we wished agent work had been organized.
- "You want Spacedock if" restructured from two bullets on the "human / agent"
  axis to three bullets on three distinct axes:
    1. shape (repeatable, multi-step pipelines on many inputs)
    2. trust (single-agent output you can't verify on its face)
    3. time horizon (work that spans days or weeks)

Source: messaging anchored in the alpha-interview synthesis at
recce-gtm/knowledge/customer-insight/interview-plans/alpha-interview-synthesis/alpha-problems.md
(7 universal problem clusters across the 9-alpha corpus).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Comment thread README.md
# Spacedock

Spacedock runs agent work through defined stages, so you can delegate in batches and make only the calls that need your judgment.
Hand an agent a multi-step job and it drifts, skips steps, or invents its own path. Hand it a workflow and it follows the workflow.
Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

not sure if this is a desire state or problem state

Comment thread README.md
The first officer coordinates the flow: it dispatches workers to advance each work item and surfaces approval-worthy decisions to you, the captain, so batches move forward without pulling you into every session.
Spacedock runs agent work through stages you define. Each stage gets a fresh context, an explicit gate, and a structured report. Reviewers are allowed to push back. State lives outside the agent, so work survives the context limit and picks up the next session where it left off.

We built Spacedock because we were there. We lived the pipelines that didn't finish, the reviewers that rubber-stamped, the work that died at the context limit, and the routines spread across too many tools. The tool is the way we wished agent work had been organized from the start.
Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

don't need to mention "we're there"

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

▎ We built Spacedock because we lived this. Pipelines that didn't finish. Reviewers that rubber-stamped each other's output. Work that died at the
▎ context limit. Routines spread across too many tools. Spacedock is what we needed years ago, built from inside the problem.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

▎Problem: "Hand an agent a multi-step job and it drifts, skips steps, or invents its own path. Hand it a workflow and it follows the workflow."

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant