Scored ranking of rivus projects. Review monthly or when deciding what to focus on next.
| Dimension | 1 | 3 | 5 |
|---|---|---|---|
| Payoff | Nice to have | Useful, saves time | Game-changing, compounds |
| Amplification | One-off, manual | Reusable for several tasks | Runs autonomously at scale |
| Showability | Internal tooling only | Demoable to collaborators | Jaw-dropping wow, differentiating |
| Feasibility | Unclear how, high risk | Known approach, some risk | Clear path, low risk |
| Effort | Months of focused work | Weeks | Days to working v1 |
| Momentum | Cold start, nothing built | Some pieces exist | Already partially working |
| Defensibility | Easily replicated | Some proprietary data/edge | Unique data, arch, or workflow moat |
| VC narrative | Irrelevant to pitch | Supporting evidence | Core to the story, demo “aha” |
| Monetization | Pure internal tool | Could charge eventually | Clear revenue path / customer pull |
| Excitement | Dread / burned out | Neutral, would do it | Can’t stop thinking about it |
Score = Payoff + Amplification + Showability + Feasibility + Effort + Momentum + Defensibility + VC narrative + Monetization + Excitement (max 50)
| # | Project | Pay | Amp | Show | Feas | Eff | Mom | Def | VC | Mon | Exc | Total | Notes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Intel (companies + people) | 5 | 4 | 5 | 4 | 3 | 4 | 4 | 5 | 4 | 4? | 42 | Core deliverable. Proprietary data + VC demo asset |
| 2 | Vario (multi-model platform) | 4 | 5 | 4 | 5 | 4 | 5 | 3 | 3 | 3 | 5 | 41 | Foundation for everything. Excitement keeps it near top |
| 3 | VC intel | 5 | 4 | 5 | 3 | 3 | 3 | 4 | 5 | 4 | 4 | 40 | Pitch with your own tool analyzing them = ultimate “aha” |
| 4 | World inference + investing | 5 | 5 | 3 | 4 | 3 | 3 | 4 | 3 | 5 | 4 | 39 | Macro reasoning via ng → investable insight. No burnout |
| 4b | Market signal intelligence | 5 | 5 | 4 | 3 | 3 | 2 | 4 | 4 | 5 | 4 | 39 | Ingest news + world events → assess stock/sector impact. Real-time edge |
| 5 | Ideas (idea eval + landscape) | 4 | 4 | 4 | 4 | 4 | 5 | 3 | 3 | 2 | 4 | 37 | Infra: upgrades Draft novelty, finance thesis eval, semnet synergy |
| 5b | Skillz feasibility spec | 4 | 3 | 4 | 4 | 5 | 2 | 3 | 4 | 3 | 5 | 37 | Moonshot worth a feasibility pass. Could unlock the #1 moat project |
| 6 | Skillz | 5 | 5 | 3 | 2 | 2 | 2 | 5 | 3 | 2 | 5 | 34 | Highest ceiling + deepest moat. Excitement pulls it up |
| 6 | Supply chain / bottleneck | 4 | 3 | 4 | 3 | 3 | 2 | 4 | 4 | 4 | 3? | 34 | Unique vertical, clear enterprise ROI, defensible angle |
| 6 | Video analyzer | 4 | 3 | 4 | 4 | 3 | 3 | 3 | 3 | 3 | 4 | 34 | Earnings calls + media consumption. Excitement boost |
| 9 | Learning / recall + gyms | 4 | 4 | 3 | 3 | 3 | 3 | 4 | 2 | 2 | 4? | 32 | Symbolic recall + gyms for product perf self-improvement |
| 10 | Draft (doc analysis + writing) | 4 | 3 | 4 | 4 | 3 | 4 | 2 | 2 | 2 | 3 | 31 | Useful but crowded space. Ideas module is its upgrade path |
| 11 | SuperReader + SuperFeed | 4 | 4 | 4 | 3 | 2 | 1 | 2 | 2 | 3 | 4 | 29 | Cold start. Shared pipeline, single initiative |
| 13 | Brainstormer | 3 | 3 | 3 | 4 | 4 | 3 | 2 | 1 | 1 | 4 | 28 | Retrieval + human thought UI. Distinct from Vario |
| 13 | Semantic net / inference | 4 | 4 | 3 | 2 | 2 | 2 | 3 | 2 | 1 | 4 | 27 | Cool research play. Excitement keeps it from bottom |
These can override the score. The score is a starting point, not a straitjacket.
Intel work is directed toward VC demo. Companies + people infra serves the TenOneTen sprint and 20-firm pipeline.
STATUS 9,574 companies, 4,481 VC firms, 120+ VC people. Founder scoring 3/7 dims. TenOneTen dossier complete (partner profiles, firm analysis). No VC → portfolio mapping pipeline yet.
STATUS 5 blocks (produce, score, revise, reduce, repeat), 11 recipes, CLI + Python API, 221 tests. Enrich block design approved. Missing: enrich, execute, decompose, classify, steer, plan.
Sprint plan — blocks first (they unlock everything), then prove with benchmarks + real tasks:
| Day | What | Produces |
|---|---|---|
| 1 | Build enrich block (web fetch + search + RAG) | Unlocks VC, supply chain, world inference |
| 2 | Build execute block (run code, call APIs, tool use) | Unlocks verification, data processing, agentic tasks |
| 3 | Port benchmark.py, run MATH on 11 recipes | Numbers: which recipe wins. Proof engine works |
| 4 | Run ng on real Draft doc + wire gyms to ng | Proof ng is useful. Gym baselines established |
| 5 | Build decompose block (sub-task breakdown) | Unlocks complex multi-step reasoning |
| 6 | TenOneTen through ng: enrich → score → dossier | Killer demo powered by ng |
| 7 | World inference: model_debate on macro questions | Investable insight without backtesting burnout |
How gyms and ng interact: Gyms use ng recipes as the thing being measured.
gym task → ng recipe → output → gym evaluator → score → learn. Improving recipes improves gym scores. Gym scores tell you which recipes to improve. Self-reinforcing loop.
Detailed next steps:
Goal: Demonstrate that product performance self-improves through gyms.
The non-burnout path to finance. Use ng model_debate / confirm / weighted_vote on macro questions — the output is insight about the world, not trading signals. But insight feeds investing naturally.
Examples:
ng model_debate "What happens to semi demand if AI capex peaks in 2027?"ng confirm "Is TSMC pricing power sustainable given Intel foundry?"ng weighted_vote "Top 3 supply chain bottlenecks for 2026-2027?"The real-time edge. Ingest current market news + world events, assess impact on stocks/sectors before the market fully prices it in. Distinct from world inference (#4, which is macro reasoning about the future) — this is reactive: event happens → within minutes, assess which stocks/sectors are affected and how.
Pipeline: news source → ingest → entity extraction → impact assessment (variong multi-model) → signal
What makes it defensible: combines intel company graph (who supplies whom), supply chain bottleneck data (which disruptions cascade), and multi-model reasoning (variong) into a response pipeline no retail investor has.
Key job clusters and goals:
vic_ideas (9.6K pending, PAUSED on fetch errors), vic_text_extract (18K pending). Goal: feed ML model for idea evaluationsupplychain_anchors (44 pending), supplychain_expand (79), supplychain_news (674, PAUSED). Goal: build company relationship graphvc_data_* (all complete), a16z (complete). VC firm data enriched.newsflow_monitor (9K pending, PAUSED), newsflow_youtube (1.8K, PAUSED on bot detection). Goal: keep content pipeline flowingBuild end-to-end VC intel demo using TenOneTen Ventures (LA, data science founders). Cold start — nothing in DB yet.
Next: Repeat this pattern for remaining 19 target firms. Needs portfolio mapping pipeline first.
SV Tier 1
SV Tech/AI-Focused
LA-Based / LA-Present
Rough sizing — what each project could be worth if it works. Payoff range is annual revenue or equivalent value.
| Project | Difficulty | Size (person-months) | Payoff Range |
|---|---|---|---|
| Intel (companies + people) | Medium | 3-6 | $50K-500K/yr |
| Vario (multi-model platform) | Medium | 4-8 | $0 (infra) or $100K-1M |
| VC intel | Medium | 2-4 | $100K-1M/yr |
| Skillz | Hard | 6-18 | $1M-50M/yr |
| Skillz feasibility spec | Easy | 0.5-1 | Unlocks above |
| World inference + investing | Medium | 2-4 | $100K-10M/yr |
| Market signal intelligence | Medium | 3-6 | $200K-5M/yr |
| Supply chain / bottleneck | Medium | 4-8 | $200K-2M/yr |
| Video analyzer | Medium | 2-4 | $50K-300K/yr |
| Learning / symbolic recall | Hard | 4-8 | $0 (internal) or $50K-500K |
| Draft (doc analysis + writing) | Easy | 2-3 | $20K-200K/yr |
| SuperReader + SuperFeed | Medium | 4-8 | $50K-500K/yr |
| Semantic net / inference | Hard | 6-12 | $50K-1M/yr |
| Brainstormer | Easy | 1-2 | $0 (feature of Vario) |