What VentureHub is

VentureHub is AK's venture portfolio hub — a living web artifact and the decision-making cockpit for his businesses and business ideas. Its jobs:

1. Externalize venture decision-making: argue with the criteria once, let the ranking fall out — instead of holding one gut ranking in his head.
2. Preserve every idea so nothing is lost, even low-priority ones.
3. Separate stage (exists vs. idea) from priority (what to work on next).
4. Make next actions visible and queryable (tasks).
5. Stay legible to both AK and his agents.

Current roster: 9 ranked ventures — AccelerateBooks (live business), Arsenal (idea — product venture), Powerful Websites for Christian Leaders (side hustle), Theology for Sleep (side hustle), Nile Outbound (pre-flight), GraceOverflow (staged audience asset), Gospel Practitioner (idea — credibility platform), TikTok Shop Videos (idea, unresearched), and Zeitgeist Apparel (idea, low priority) — plus Agora Ventures LLC as an unranked foundational venture / holding entity.

Hosting: temporarily on Agent Muse's web runtime. Plan: move to one of AK's machines or a VPS, with Agent Muse remaining steward as business lead agent. A read-only Google Doc mirror exists and is refreshed on every edit.

Who it's for (read this before proposing anything)

• AK: AuDHD/OCD, self-described visionary; executive dysfunction alongside hyperfocus; learns chronologically; needs zero-cost capture and momentum from small visible wins.
• Working rules: verify before asserting (every factual/product claim carries its checked source inline); never send or publish anything in AK's name without explicit per-instance approval; never expose his neurotype clinically to third parties.
• Design sensibility: dense information on detail pages, aesthetic breathing room on the hub; mobile-first (he reads on his phone); dark mode available.

The design rationale (from Agent Muse — critique welcome)

The core problem. AK's ventures vary on orthogonal axes — stage (live vs. staged vs. pre-flight vs. idea), nature (operating business vs. product venture vs. credibility platform vs. audience asset vs. side hustle), and worthiness (expected return × effort × sustainability). Collapsing those into one gut ranking is exactly what executive dysfunction punishes. VentureHub's job is to externalize the criteria so AK argues with the criteria once, and the ranking updates itself.

The conclusions the current design rests on

1. Confidence and enjoyment are multipliers, not addends. Dread isn't a minor demerit — with AK's brain it's the thing that kills execution, so a dread-heavy venture's expected value discounts toward zero no matter its TAM.
2. Time-to-traction beats time-to-profitability for AK specifically. Momentum itself — seeing things get built — is fuel, so “smallest next traction step” is the most important field.
3. Stage is a filter, not a score. Existing ≠ priority (GraceOverflow exists and is correctly low-priority).
4. Unknowns should cost something. Unresearched ideas (TikTok Shop videos) look cheap precisely because nothing is known; research depth is the reality check.
5. Maintenance burden is the stall predictor. AK's projects die in maintenance more than launch, so manual-vs-automated weight is heavy.
6. Scaffolding sits above the ranking. The SOP/agent-infra work multiplies every venture's score but can't be ranked against them — it's a meta-workstream.
7. Constraints live outside the score. Brand separation (Gospel Practitioner's credibility must be isolated from Theology for Sleep / GraceOverflow) and commitments (AccelerateBooks “can't give up the ship”) aren't factors to trade off — the score only answers “what next?” among valid options.

What the app does (functional inventory)

• Hub home: “Agora Ventures.” / “VentureHub” header, portfolio count, a separate unranked foundation layer, list/grid toggle, stage/role/commitment/work-window filters, scenario-aware priority board, sticky hamburger nav, and dark mode.
• 9 dense venture pages with independent stage and work-window controls, soft dates, AK/agent execution lanes, a configurable 1–7 factor model, energy forecasts, commitment bounds, and next steps.
• A Scoring & settings page exposes thresholds, unknown handling, coverage scaling, scenario sorting, the energy curve, forecast horizon, novelty flag, and every factor weight.
• Per-idea todo checklists plus a combined Tasks page with shared launch-scaffolding work listed once.
• A living Documentation page explains the current system, scoring, state-service plan, and wiring status; Philosophy and CHANGELOG preserve rationale and major changes.

Tech stack & architecture (honest account — check my claims)

• At the reviewed build, one static HTML file was approximately 130 KB: inline CSS and vanilla JavaScript, with no JavaScript framework, package dependencies, or build dependencies. The optional Google Fonts stylesheet is a network dependency, so the page is not fully dependency-free or offline.
• State is entirely in-memory (ideaState, taskItems, factor values, weights, theme). There is no localStorage or durable client persistence — edited factors, checked todos, theme and layout choices reset on reload. The shared state service is the intended durable store.
• Hash-based routing (showRoute, 17 views including the unranked foundation page, documentation, and scoring-settings pages), history.pushState.
• VentureHub.tasks: a namespaced in-memory task API (schemaVersion: 1; all / getByIdea / outstanding / add / remove / setComplete; subscribe pub/sub; defensive copying on reads) — deliberately designed as the seam where a future MCP adapter or durable store plugs in. Documented in-code for agent legibility.
• Trade-offs taken: the single file is portable and auditable, but its optional font stylesheet still needs the network for the intended typography. Cost: no durable state yet, no multi-device sync, no auth for multi-agent writes, and a single file gets harder to keep modular as it grows.
• Known gaps for you to attack: durable persistence; OS theme detection (dark mode is manual toggle only); factor edits lost on reload; no server-side MCP yet; ranking weights are session-local.

AK's nine design pillars (use these as your rubric)

1. SSOT — “Is this a source, or a copy? If a copy, can I replace it with a reference?”
2. Modularity — “Does this do 1 thing well, or 3 things adequately?”
3. MVC — “Has this solved a real problem, or am I engineering against a hypothetical?”
4. Dual Readability — “Am I forcing myself to think like a machine, or building a system that thinks like AK?”
5. Flexibility — “If AK renames/reorganizes this in 6 months, how much work?”
6. Extensibility — “Does this open more future options than it closes?”
7. Interoperability — “If the tools disappeared tomorrow, could these files open in another editor?”
8. Resilience — “If I deleted this file right now, what else breaks?”
9. Traceability — “If this turns out wrong, can I trace why it was decided?”

The invitation

We're all on the same team and these are OUR projects. Be adversarial — steelman first, then attack. I want:

1. Design critique (mobile + desktop): assess against UI/UX best practices and named laws (Fitts, Hick, Miller, Jakob, etc.). Name the law, show the violation, propose the fix. Mobile-first: AK reads this on his phone.
2. Rationale critique: where is the reasoning in section 3 wrong? What's missing? What would you add, augment, or replace? Propose better factor sets or scoring models if you have them.
3. Real-world failure forecasting: AK will lean on this for his life and productivity. What breaks first — data loss on reload, stale rankings, 50-idea scale, sync across devices, agents writing conflicting tasks? Walk the failure modes.
4. Architecture assessment against: future-flexibility, mobile-first, scalability, durability, AK-intuition, agent legibility, optionality. Is the single-file static approach right for the next 12 months? What should replace it, and when is the trigger?
5. Concrete proposals, prioritized, each with trade-offs and the pillar(s) it serves.

Ground rules

• Don't invent AK's business facts. Sparse > fabricated. Every page must trace to something AK actually said.
• Verify before asserting: product/factual claims carry checked sources inline.
• Nothing is sent or published in AK's name — proposals only.
• Brand separation is a hard constraint, not a variable.
• Date everything; AK's memory is temporally indexed.