Wiki Log

Append-only log of every ingest, query, and lint pass. Every entry starts with ## [YYYY-MM-DD] <op> | <target> so it is grep-able:

grep "^## \[" log.md | tail -10

[2026-04-28] ingest | Multica + Hercules (dual batch, mixed clipped + web-fetched sources)

• Multica sources: links.md + Welcome.md (docs landing) + 'Multica — Project Management for Human + Agent Teams.md' (homepage clip) + 'multica-aimultica … managed agents platform.md' (GitHub README clip) + image.png (board screenshot).
• Hercules sources: 'AI App and Website Builder.md' (dashboard clip) + 6 screenshots (uncategorized) + web-fetches of hercules.app, hercules.app/docs, hercules.app/pricing.
• Created competitor pages for both (Overview, ICP, Pricing, Marketing, Tech): 10 pages.
• Created feature pages: Multica × 3 (Agents as Teammates, Reusable Skills, Unified Runtimes), Hercules × 3 (AI App Builder, Built-in Commerce, Templates Marketplace): 6 pages.
• Created new canonical: Agent Orchestration Platform — Multica-only in this vault, flagged as Tier 2 BD partnership target rather than competitor.
• Category reframe to 4 lanes:
• Builder lane (3): Replit + Anything + Hercules (third instance — convergence is now a trend, not coincidence)
• Runtime / sandbox lane (1): Fly.io
• Orchestration lane (1, new): Multica — open-source, vendor-neutral, 21.3k★
• Agent-economy lane (CreateOS-only): MPP + router + sandbox + marketplace
• Dashboard updates:
• Home dashboard.html: hero text reframed (3 → 5 players, 4 lanes); 2 new positioning cards (Hercules amber, Multica indigo); horizontal scroller now overflows; ICP SVG gained Hercules + Multica arcs + 2 new seg cards; stats counter goes 3 → 5 competitors.
• product.html: feature matrix expanded 5 → 7 columns (50 rows × 2 new cells = 100 new cells, plus 8 group <tr>s updated to colspan="7").
• strategy.html: insights 8 → 10 (added #09 Multica-as-partner, #10 builder-convergence-reinforces-our-thesis); insight #01 reframed lanes count 3 → 4.
• marketing.html: positioning side panel updated to 5-player landscape.
• bd.html: Multica added as Tier 2 partnership target with named pitch (cloud-runtime adapter); Hercules added to Tier 4 (track-only — same lane as Anything).
• Cross-cutting impact: updated index; updated CSS tokens (added --multica indigo + --hercules amber + matching *-wash).
• Gaps flagged:
• Multica: license, full Cloud pricing, MCP / x402 integration status.
• Hercules: pricing tier names + $ amounts (page didn't render), full integration catalog, agent reasoning UX, AI Gateway internals.

[2026-04-23] ingest | Fly.io (batch, 4 sources — web-fetched since raw/ only had link anchors)

• Sources: fly.io homepage, fly.io/docs, fly.io/pricing, phoenix.new (all fetched via WebFetch 2026-04-23).
• Created competitor pages: Overview, ICP, Pricing, Marketing, Tech.
• Created feature pages: Sprites, Fly Machines, Fly GPUs, Managed Postgres, Phoenix.new.
• No UX flow pages — no screenshots in raw/Flydotio/.
• Cross-cutting impact:
• Updated AI App Generation Agent — added Fly column (Phoenix.new as narrow Elixir beachhead).
• Updated Built-in Backend Services — added Fly column (primitives-direct, not bundled).
• Created Hardware-Isolated Sandboxes — new canonical feature, Fly-only in this vault, flagged as direct CreateOS competitor.
• Updated Enterprise Developers — added Fly section (light: HIPAA $99/mo + support tiers, no logo wall).
• Category reframe: Fly.io is NOT a direct Replit/Anything.com competitor. Real bet is Sprites (agent runtime), overlapping E2B / Daytona / Modal / Cloudflare / Vercel Sandbox / Blaxel — and CreateOS. Phoenix.new is a community beachhead, not a mass-market builder.
• Gaps flagged: no UX screenshots, no Sprites API/SDK shape, no per-GPU pricing, no free-tier scope, no Sprites-in-OpenAI-Agents-SDK confirmation, no customer logos.

[2026-04-22] init | vault scaffolded

• Created CLAUDE.md with schema + workflows.
• Created index.md, log.md.
• Created directories: raw/, wiki/competitors/, wiki/features/, wiki/ux-patterns/, wiki/icp-segments/, wiki/marketing-plays/, wiki/comparisons/.
• Ready for first competitor ingest.

[2026-04-22] ingest | Replit (batch, 2 sources)

• Sources: Replit Docs, Replit Product.
• Created competitor pages: Overview, ICP, Pricing, Marketing, Tech.
• Created feature pages: Agent 4, Parallel Agents, Multiple Artifacts, Infinite Canvas, Built-in Services, Import Wedges, Mobile App.
• No UX flow pages — no screenshots in raw/.
• Cross-cutting impact: created AI App Generation Agent, Parallel Agent Execution, Multi-Artifact Output, Built-in Backend Services, Competitive Migration Imports, Enterprise Developers.
• Gaps flagged: mid-tier pricing, full integration catalog, agent-credit economics, customer case studies, architecture specifics, UX screenshots.

[2026-04-22] ingest | Anything.com (batch, 2 sources)

• Sources: Anything.com Docs, Anything - AI app builder.
• Created competitor pages: Overview, ICP, Pricing, Marketing, Tech.
• Created feature pages: Conversational Builder, Max, Multiple Artifacts, Accept Payments, Databases, Templates, Custom Instructions, Domains.
• No UX flow pages — no screenshots in raw/.
• Cross-cutting impact: created Autonomous Testing and Fixing, Built-in Payment Acceptance, Creator Education Content, Indie Makers and Solopreneurs.
• Gaps flagged: pricing detail, security posture, integration catalog, GitHub/open-source status, core architecture, UX screenshots.

[2026-04-22] synthesis | Replit vs Anything.com

• Created Replit vs Anything.com flagship comparison.
• Key findings: Replit = platform depth + dual audience; Anything.com = sharp maker-monetization focus. Sharpest differentiator is Built-in Payment Acceptance (Anything.com first-class vs. Replit integration-only).
• Updated index — ~30 new page entries.

[2026-04-22] dashboard | v0.1 — 2-competitor brief

• Built dashboard.html — single-file HTML brief, opens in browser.
• Sections: positioning spread, 5 headline findings, 25-row feature matrix, ICP coverage SVG (3-segment axis), gaps + next-moves.
• Editorial visual treatment: warm cream bg, competitor accent colors (Replit orange / Anything teal), serif display type. Self-contained, no external deps.
• Designed to absorb a 3rd competitor as another matrix column + ICP arc when the wiki grows.

[2026-04-22] build | HTML rendering for all wiki pages

• Reason: clicking wiki links from dashboard dumped raw markdown with mojibake.
• Installed markdown 3.9 via pip (user site). Wrote build.py converter.
• Converter handles: YAML frontmatter → styled metadata block; Obsidian [[wiki-links]] → relative .html hrefs (depth-aware, URL-encoded, with fallback unresolved-link span); > 💭 / > ❓ / > ⚠️ blockquotes → styled callouts; tables + fenced code via markdown extensions; protects code blocks from wiki-link rewriting.
• Generated 44 HTML pages (36 wiki + 4 raw sources + CLAUDE, Welcome, index, log). Shared stylesheet at assets/wiki.css.
• Re-run after every ingest: python3 build.py.
• Updated dashboard.html: all 12 .md hrefs → .html.

[2026-04-23] build | dashboard auto-refresh (mechanical stats)

• Added <!-- BUILD:KEY -->...<!-- /BUILD --> sentinels to dashboard.html around 9 mechanical fields: topbar meta line, 5 stat counters, 3 footer placeholders.
• Extended build.py with compute_dashboard_stats() + update_dashboard_stats(). Counts competitors from wiki/competitors/*/ dirs, wiki pages from wiki/**/*.md, sources from raw/**/*.md, UX flows from wiki/competitors/*/ux-flows/, and features-in-matrix by parsing the dashboard's own <tbody> (so the counter stays in sync when Claude edits matrix rows).
• Updated CLAUDE.md: added steps 9–10 to the batch-ingest workflow ("rewrite dashboard.html synthesis sections" + "run python3 build.py"), plus a new ## Dashboard section documenting the auto/manual split, what to edit when adding a competitor, and how to introduce a new competitor color.
• Result: counters + dates auto-refresh on every build.py run; synthesis stays Claude-authored and stays on-model because CLAUDE.md now enforces regeneration as part of ingest.
• Verified: build output 2 competitors · 24 matrix features · 36 wiki pages · 4 sources · 0 UX flows. (Corrected "25 features tracked" → actual 24-row matrix count.)

[2026-04-23] ingest | UX screenshots (18 files, both competitors)

• User added 18 screenshots (9 per competitor) covering: landing, signup, workspace, agent chat, code/preview, publish, settings, version history, canvas (Replit), task board (Replit), command palette (Replit), Core upgrade modal (Replit).
• Corrections surfaced (major):
• Anything.com has a full in-browser code editor with Next.js file tree + JSX editing. Prior wiki marked "code editor: not advertised" ❌ — was wrong.
• Anything.com has Figma import as a first-class prompt-input affordance. Prior wiki marked "no designer on-ramp" ❌ — was wrong.
• Replit has Autonomous Build Mode (Core-gated). Prior Autonomous Testing and Fixing marked autonomous loop as Anything-only ❌ — was wrong.
• New features captured:
• Anything.com: Named Design Systems ("High-Fidelity SaaS", "Ghost Structure", Action Blue), Assets, Secrets, app-level Auth Providers, App Store Marketing, Monetization, Social Share, Invite/Collaboration, Publishing (project-wide), Contextual Max invocation.
• Replit: Task Board, Plan Mode, Connectors, Mobile Simulators, Background Tasks, Security Scanning + SBOM, Referral Badge ("Made with Replit" + 50 credits/click), Feedback Widget, Command Palette, Library Panel, Usage Widget, Economy Mode.
• New pages: 8 Anything.com UX flows, 9 Replit UX flows, 11 Anything.com feature pages, 13 Replit feature pages, 7 cross-competitor UX pattern pages, 1 canonical feature rewrite.
• Updates to existing pages: Anything.com - Overview, Anything.com - Custom Instructions, Replit - Overview, Replit - Marketing, Autonomous Testing and Fixing, Competitive Migration Imports, Enterprise Developers.
• Dashboard rewrite: new hero + lede ("same shell, different audiences"), both positioning cards rewritten with post-ingest evidence, insights grown 5 → 7 (added: Anything-is-quietly-a-dev-tool, Security+SBOM is Replit's hardest moat, Made-with-Replit viral loop), feature matrix grown 24 → 47 rows with new sections (Trust & compliance, Publish + growth, Pricing), ICP notes updated with enterprise proof + hidden-code-editor framing, gaps + next-moves rewritten (removed stale "no UX flows" gap).
• Build verified: python3 build.py → 92 HTML pages, stats 2 competitors · 47 matrix features · 84 wiki pages · 4 sources · 17 UX flows. UX-flow counter correctly moved from 0 to 17.

[2026-04-23] playbook | prescriptive recommendations (McKinsey-style)

• User asked: "separate pages for optimum UX + what works / what doesn't / how to grow acquisition, retention, revenue."
• Created 7 pages under wiki/playbook/:
• 00 - Executive Summary — top 5 moves, 2×2 by objective, top-level metrics.
• 01 - Acquisition — pre-signup prompt, chips, imports, social proof, distinctive brand.
• 02 - Retention — persistent usage widget, visible agent reasoning, durable version history.
• 03 - Revenue — credit metering, gate-at-desire paywall, bundle-vs-unbundle, viral loops.
• 04 - UX Recommendations — adopt patterns (7 shell + 5 differentiators) + avoid anti-patterns.
• 05 - Strategic Bets — pick one audience + one autonomy + one signature + one loop.
• 06 - Quick Reference Card — print-and-pin cheat sheet with 10 target metrics.
• Added new Playbook section to dashboard.html — 4-card strip (Acquisition / Retention / Revenue / Strategy), each with ✓ do / ✕ avoid bullets + link to full page. CSS accent color per topic (purple / blue / green / amber). Dark footer strip with hardest-truth callout + links to exec summary, UX patterns, quick ref.
• Updated footer "Go to" links + index.md with Playbook section.
• Format: each page uses "3 moves to make / 3 things to avoid / how to measure." Opinionated and prescriptive.

[2026-04-23] ingest | CreateOS Internal-Research (5 docs)

• User added raw/Internal-Research/: CreateOS_Growth_Playbook.md (503 lines), CreateOS_Playbook_Addendum_Inflection_Points.md (172), CreateOS_Playbook_Addendum2_Agent_Infra.md (221), CreateOS_Playbook_Addendum3_Crypto_Agent_BD.md (238), CreateOS_Playbook_OnePage_Reference.md (51). Total: 1,185 lines of internal strategy research.
• Critical context shift: the competitive research is for CreateOS (NodeOps, founder NK). Dashboard + playbook reframed from "recommendations for anyone building in this space" to "recommendations for CreateOS specifically."
• Key findings from Internal-Research:
• CreateOS has MPP Gateway (HTTP 402 agent payments, live), router.nodeops.network (inference routing, live), marketplace with creator revenue share, AI Sandbox — an unassembled stack no competitor has.
• CreateOS is NOT on OpenAI's Agents SDK default provider list (E2B, Modal, Daytona, Cloudflare, Vercel, Blaxel, Runloop are). This is the #1 BD priority.
• Cloudflare declared itself "definitive platform for the agentic web" on April 13, 2026 (Agents Week). Category consolidating real-time — window measured in months.
• Recommended positioning pivot: from "prompt-to-live-app" to "the only agent sandbox where the agent can earn."
• x402 is now $600M annualized, co-governed by Coinbase + Cloudflare + Google + Visa. MPP must align publicly within 30 days (facilitator / extension / compatible — pick one).
• Three-factor hockey-stick formula: (new primitive) × (narrow product decision) × (distribution event). Bolt, Lovable, Cursor, Replit, Base44, Hugging Face, v0 all combined all three in 1-3 month windows.
• Named BD targets with contact paths: Karan Sharma (OpenAI Agents SDK), Rohan Varma (OpenAI Codex), Harrison Chase (LangChain), Sri Viswanath (Sycamore), Jansen Teng (Virtuals Protocol), Nish (HydraDB), x402 Foundation via Nemil Dalal.
• New wiki section: 7 pages under wiki/createos/ — Overview, Positioning & Wedge, The 10 Gaps, Three-Factor Inflection, BD Targets, 30-60-90 Plan, Risks & Mitigations.
• Dashboard rewrite:
• New hero/lede framing ("three players, one unclaimed wedge").
• Prominent CreateOS strategic brief section at top with: unassembled-stack table, critical-gap callout (OpenAI Agents SDK), 3 top moves (reframe positioning / build Agents SDK integration / Agent Economy Week), three-factor hockey-stick formula applied, links to CreateOS pages.
• All 7 headline findings updated with CreateOS-specific implications.
• All 4 Playbook cards rewritten as CreateOS-specific moves (Acquisition = ship no-signup demo + SDK inclusion; Retention = per-agent spend caps + usage widget; Revenue = MPP↔x402 alignment + agent-economy tier; Strategy = Agent Economy Week in Month 2-3).
• Dark CTA strip updated.
• Footer "Go to" links prioritize CreateOS pages.
• Added > 🎯 For CreateOS specifically: callouts at the top of each of the 7 playbook pages.
• Updated index.md with new CreateOS section (pinned above Competitors).
• Purple accent color (#5B21B6) introduced for CreateOS surfaces, distinct from Replit orange + Anything teal.

[2026-04-23] app | Flask web UI for competitor + resource management

• User requested a holistic product: UI for adding competitors, uploading files, updating resources, viewing current resources, getting insights.
• Built app.py (Flask, port 5001) with endpoints: / (competitors list), /add (add form), /c/<name> (detail), /c/<name>/upload, /c/<name>/replace/<path>, /c/<name>/delete-resource, /c/<name>/delete, /c/<name>/build (insight generation), /c/<name>/preview/<path> (file preview). Plus pass-through serving of dashboard.html, index.html, log.html, wiki/*.html, assets/*, raw/*.
• Templates under templates/: base.html, competitors.html, add_competitor.html, competitor_detail.html, 404.html — Jinja2. Aesthetic matches existing dashboard (cream bg, serif display, competitor accent colors).
• Stylesheet assets/app.css — form controls, resource table grouped by folder, insight panel, danger zone.
• Insight generation flow (kept human-in-the-loop for quality): click "Regenerate + get Claude prompt" → runs build.py + returns a paste-ready prompt the user copies into Claude Code to trigger the ingest workflow from CLAUDE.md.
• File upload: multi-file, auto-classifies by extension (.png → screenshots/, .md → pages/, .pdf → docs/). Max 25 MB per file. Extensions restricted.
• Security: path-traversal guards on all file operations (safe_resource_path helper validates target is inside competitor's raw dir).
• Installed Flask 3.1.3 via pip (user site).
• Updated README.md with setup, workflow, and architecture notes.
• Updated CLAUDE.md preamble to note the app exists + what form the Claude-triggered prompts take.
• Smoke-tested all endpoints: home (3 competitors), add form, detail pages, asset serving, wiki passthrough, dashboard passthrough, 404. All return expected status codes.

[2026-04-23] refactor | React frontend + Flask JSON API (backend/frontend split)

• User asked to make the frontend React-based with the Python backend, proper folder structure, without breaking anything.
• New layout:
• backend/ — Python package with api.py (all routes refactored to return JSON for CRUD; static HTML passthroughs retained).
• frontend/ — Vite + React + TypeScript SPA. Components: Layout, StatBar, FlashStack, UploadForm, ResourceRow, InsightsPanel, DangerZone. Pages: CompetitorsList, AddCompetitor, CompetitorDetail, NotFound. Shared: api.ts (typed fetch client), types.ts, flash.ts, styles.css.
• app.py — thin launcher at vault root: from backend.api import app. Keeps python3 app.py working.
• build.py — unchanged at root.
• URL design: / = React SPA, /api/* = JSON, /static/* = Vite bundle, /dashboard.html + /wiki/* + /assets/wiki.css = existing static HTML passthroughs. Renamed old /index.html → /wiki-index.html to avoid collision with React's index.html at /.
• Flask static folder: pointed at frontend/dist/static/ so Flask's built-in /static/ handler serves Vite-bundled assets without a custom route.
• Removed: templates/ (old Jinja templates), assets/app.css (old app stylesheet), empty flydotio stale folder.
• Dev mode: Vite dev server at :5173 with proxy for /api, /dashboard.html, /wiki/*, etc. to :5001. Optional — only when iterating on the frontend.
• End-to-end smoke test:
• SPA at /, /add, /c/:name → all 200, all contain id="root".
• API: /api/stats = {competitors: 3, files: 26, wiki_pages: 34, ux_flows: 9}. /api/competitors returns list. /api/competitors/replit returns detail with 13 resources + 34 wiki pages.
• Static assets: JS (~184KB), CSS (~13KB) both 200.
• Passthroughs: dashboard.html, wiki-index.html, log.html, assets/wiki.css, wiki/*.html — all 200.
• CRUD: create cursor-test via POST /api/competitors → folder created with links.md + initial_notes.md → delete via DELETE /api/competitors/cursor-test?confirm=cursor-test → gone. Clean round-trip.
• Updated docs: README.md rewritten with new folder structure, URL layout, setup + run commands, workflow diagram.
• Installed: flask 3.1.3 (Python), react 18.3.1 + react-router-dom 6.28 + vite 5.4 + typescript 5.6 (Node). 47 modules in the Vite build.

[2026-04-23] cleanup | remove unused files

• Audited the vault for orphan files (unreferenced anywhere in code or content).
• Removed: Welcome.md + Welcome.html (Obsidian's default welcome note, never linked from anywhere) and the now-dead /Welcome.html Flask route.
• Verified no .DS_Store, no __pycache__, no empty directories.
• Left alone (legitimately in use): .obsidian/ (Obsidian config), .git/ (git history), .claude/ (Claude Code state), CLAUDE.html (built output, served + proxied), node_modules/ (required for Vite build), package-lock.json (npm lockfile).