Skip to content

P0–P2 polish: reposition as fundraising / exit-prep document#1

Merged
CreatmanCEO merged 1 commit into
mainfrom
polish-readme-and-meta
May 3, 2026
Merged

P0–P2 polish: reposition as fundraising / exit-prep document#1
CreatmanCEO merged 1 commit into
mainfrom
polish-readme-and-meta

Conversation

@CreatmanCEO
Copy link
Copy Markdown
Owner

Summary

Repositions the showcase from generic "active-development commercial product" framing into a fundraising / exit-prep document that works simultaneously for two paths confirmed by the author:

  • Path A — owner-led growth with investment / co-founder / partnership
  • Path B — strategic acquisition or republishing

Both paths need the same artefacts: honest progress map, gap analysis, competitor positioning, technical credibility, contact channels. License unchanged (Commercial — All rights reserved).

What changed

P0 — credibility & accuracy

  • Hero rewritten with new Project trajectory section spelling out Path A vs Path B honestly
  • 6 hero badges: License, Stars, Validate CI, Status, Platform, Open to discussion
  • Topics + description applied via gh api after merge

P1 — substantial structural additions

  • Why GHOST? — concrete competitor analysis (Cluely $75/mo · Screenpipe 16K stars · Natively · Interview Hunter · Interview Coder · Final Round AI). What was taken from each, what was avoided. Tone: respectful, not aggressive.
  • Latency target — explicit streaming pipeline rationale (Silero VAD onset → streaming partial transcripts → LLM begins on partials → token-by-token overlay → first reply within 1–2 s of question end). Framed as target, not a measured benchmark.
  • Project status — Phase-1 checklist: 8 items implemented (invisible overlay, dual-source audio capture, sidecar WebSocket, EventBus, etc.), 5 in progress (Silero VAD integration, streaming pipeline, multi-document RAG, hotkey UI, perf opt). Replaces the previous abstract "In Progress / Planned".
  • What's needed to ship — honest gap analysis: engineering (macOS / Linux ports, code signing, auto-update, installer, crash reporting, telemetry), product (onboarding polish, pricing page, payment integration, trial/freemium, support docs), business (legal review, trademark, customer support ops, marketing strategy).
  • Investment thesis — 3-line summary: Market / Gap / Why now.
  • How it's built — multi-agent dev workflow (Orchestrator + Coders + Tester + Documentation Agent) with cross-link to the same author's claude-code-antiregression-setup. Positions solo-founder execution as documented process.
  • Limitations — Windows 10 v2004+ only, Phase-1 pre-V1.0, OS-level capture exclusion is not magic.
  • Open to discussion — explicit dual CTA with email / Telegram / site / GitHub.
  • Related — Claude Code ecosystem — cross-links to all 6 sister repos by the same author.
  • docs/architecture.svg — hand-rendered diagram of the actual stack (Electron 33 + React 19 + Vite + koffi for Win32 / WebSocket JSON-RPC port 9876 / Python 3.11 asyncio sidecar with Silero VAD + pyaudiowpatch + deepgram-sdk + faster-whisper + litellm Router + ChromaDB). Replaces the obsolete ASCII diagram.
  • Tech-stack table corrected: Electron 33+ (was 32), React 19 (was 18), asyncio + websockets (was FastAPI). Matches the actual local code.
  • Invisible-overlay code example now shows the real koffi-based Win32 call, not a pseudo-code stub.

P2 — content

  • README.ru.md — full Russian mirror with same structure, badges, dual CTA.
  • Author signature expanded: full name (Nick Podolyak) + Habr / dev.to profile links — consistent with sister repos.

Bug fixes

  • Removed broken README references to docs/TECH_STACK.md and docs/USE_CASES.md (never existed in this repo) — content surfaced inline.
  • Patched docs/ARCHITECTURE.md and docs/FEATURES.md footers to drop same broken cross-refs.

CI

  • .github/workflows/validate.yml — LICENSE / CHANGELOG presence, examples/*.json parses, docs/*.svg well-formed XML, every docs/* and examples/* asset referenced from README exists, internal Markdown links resolve.

What is deliberately NOT in this PR

  • Pricing / equity / valuation specifics — those belong in private discussions with serious parties only.
  • Specific revenue targets / financial projections — same reason.
  • Anything that would leak Phase-1 internal IP — commit SHAs, internal module names beyond architectural overview, exact LLM prompts, performance numbers.
  • macOS / Linux porting work — out of Phase 1 scope.

Test plan

  • LICENSE and CHANGELOG.md exist
  • examples/config.example.json parses as JSON
  • docs/architecture.svg is well-formed XML
  • All docs/* and examples/* assets referenced from README exist
  • All internal Markdown links from README / README.ru.md / CHANGELOG / docs/ARCHITECTURE / docs/FEATURES resolve
  • No code in examples/ actually executable (it's a config example) — JSON validity is sufficient

🤖 Generated with Claude Code

The previous showcase README (Feb 26) framed GHOST as "an active-development
commercial product" without articulating the path to commercialisation.
Author confirmed the project sits at a fork: owner-led growth (with
investment / co-founder / partnership) OR strategic acquisition / republishing.
This PR repositions the showcase to work as a fundraising / exit-prep
document for both paths simultaneously, while keeping the License unchanged
(Commercial — All rights reserved).

P0 — credibility & accuracy
- Hero rewritten with Project trajectory section spelling out Path A vs Path B
  honestly — "open to either"
- Description / topics applied via gh api after merge
- Hero badges: License, Stars, Validate CI, Status, Platform, Open to discussion

P1 — structural strengthening (substantial)
- NEW Why GHOST? section with concrete competitor analysis: Cluely ($75/mo),
  Screenpipe (16K stars), Natively, Interview Hunter, Interview Coder, Final
  Round AI — what was taken from each, what was avoided. Tone: respectful,
  not aggressive.
- NEW Latency target section with the streaming pipeline rationale (Silero
  VAD onset → streaming partial transcripts → LLM begins on partials → token
  output → first reply within 1–2 s). Explicitly framed as TARGET, not a
  measured benchmark — Phase 1 is what makes this achievable.
- NEW Project status with concrete Phase-1 checklist (8 items implemented,
  5 in progress) replacing the previous abstract "In Progress / Planned"
- NEW What's needed to ship — honest engineering / product / business gap
  analysis (macOS port, code signing, auto-update, payment integration,
  Trial / freemium, legal review, trademark, marketing strategy)
- NEW Investment thesis (3-line summary): Market / Gap / Why now
- NEW How it's built — multi-agent dev workflow (Orchestrator + Coders +
  Tester + Documentation Agent) with cross-link to claude-code-anti-
  regression-setup. Positions solo-founder execution as documented process,
  not "force of will"
- NEW Limitations section — Windows-only, beta-stage, OS-level capture
  exclusion is not magic
- NEW Open to discussion section with explicit dual CTA (investment /
  partnership / acquisition) and contact channels (email, Telegram, site)
- NEW Related — Claude Code ecosystem section cross-linking to all 6 sister
  repos by the same author (anti-regression-setup, ai-context-hierarchy,
  claude-statusline, notebooklm-claude-workflows, lingua-companion, diabot)
- docs/architecture.svg — hand-rendered system diagram showing the actual
  stack (Electron 33 + React 19 + Vite + koffi for Win32 / WebSocket
  JSON-RPC port 9876 / Python 3.11 asyncio sidecar with Silero VAD +
  pyaudiowpatch + deepgram-sdk + faster-whisper + litellm Router + ChromaDB)
- Tech-stack table corrected to actual versions (Electron 33+ not 32+,
  React 19 not 18, asyncio + websockets not FastAPI)
- The invisible-overlay code example now shows the real koffi-based
  approach (matches local C:/ghost code), not a generic "SetWindowDisplayAffinity"
  pseudo-code
- Author signature expanded with full name and Habr / dev.to profile links

P2 — content
- README.ru.md — full Russian mirror with the same structure, badges, dual
  CTA. Habr-audience entry point.

Bug fixes
- Removed broken README references to docs/TECH_STACK.md and docs/USE_CASES.md
  (those files never existed in this showcase repo) — content surfaced inline
- Patched docs/ARCHITECTURE.md and docs/FEATURES.md footers to drop the same
  broken cross-refs

CI
- .github/workflows/validate.yml — checks LICENSE / CHANGELOG presence,
  examples/*.json parses as JSON, docs/*.svg is well-formed XML, every
  docs/* and examples/* asset referenced from README exists, all internal
  Markdown links resolve

Excluded
- Pricing / equity / valuation specifics (those belong in private
  discussions, not the public showcase)
- Specific revenue targets / financial projections
- Anything that would leak Phase-1 internal IP (commit SHAs, internal module
  names beyond the architectural overview)
- macOS / Linux porting work (out of scope for Phase 1)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@CreatmanCEO CreatmanCEO merged commit d26128f into main May 3, 2026
1 check passed
@CreatmanCEO CreatmanCEO deleted the polish-readme-and-meta branch May 3, 2026 14:12
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