# Wave 1 — Monetization Strategist

## Position (2 sentences)
MiniML has effectively zero direct-revenue potential today — [MEASURED] 1 GitHub star, 0 forks, 0 issues, and 57 npm downloads in the last month (391 in all of H1 2026, i.e., author + bots) — in a category where the paying tiers are already owned by VC-armed incumbents (Cube $48M raised, dbt Labs post-Transform, Snowflake Cortex Analyst billing 6.7 credits/100 messages natively inside the warehouse). The only non-fantasy return is career capital plus a long-shot "strategic asset" option, and both are maximized by the same action: keep it MIT, make it excellent, and market the author rather than the package.

## Evidence (ranked paths table)

| # | Path | Who pays | How much (realistic) | Adjacent-payment evidence | Verdict |
|---|------|----------|----------------------|---------------------------|---------|
| 1 | **Career capital** (author's salary, mobility, consulting, conference/credibility positioning) | Future employers / clients pay the *author*, not the project | $0 project revenue; plausibly +$10–40k/yr in salary leverage or 1–2 consulting engagements ($150–250/hr) IF paired with content and adoption push [SPECULATION on magnitude] | US data-eng salaries ~$130–200k+ with senior/staff spread wide ([Indeed](https://www.indeed.com/career/data-engineer/salaries), [Levels.fyi](https://www.levels.fyi/)); a shipped, tested, documented semantic layer with an AST-based SQL-injection story is a strong staff-level portfolio artifact [SPECULATION on interview value] | **KEEP — the only path that pays at current traction.** But honesty check: at 1 star it's a portfolio piece, not a reputation. Value requires writing/talks, not more commits. |
| 2 | **Acquisition / acqui-hire bait** | A platform company buying the *author + design taste* | $0 expected value today; historically these deals go to teams with traction | Transform (MetricFlow, VC-funded team) → dbt Labs 2023 ([dbt Labs](https://www.getdbt.com/blog/dbt-acquisition-transform), [TechCrunch](https://techcrunch.com/2023/02/08/dbt-acquires-transform/)); ClickHouse serially acquiring OSS projects with large communities (HyperDX, Langfuse, LibreChat) ([TechCrunch](https://techcrunch.com/2026/05/27/clickhouse-triples-annualized-revenue-to-250m-charting-a-path-toward-an-ipo/)); Honeydew got Snowflake Ventures money as a *company* ([Snowflake blog](https://www.snowflake.com/en/blog/investment-honeydew-business-intelligence/)) | **LONG-SHOT OPTION, not a plan.** Every acquired comparable had thousands of stars and/or a funded team. MiniML at 1 star is invisible to corp-dev. This converts to Path 1 (get hired) at best. |
| 3 | **Snowflake Native App / marketplace** | Snowflake customers via marketplace billing | Aggregate market is real: partners earned $100M+ gross in H1 2026 ([Snowflake](https://www.snowflake.com/en/blog/snowflake-marketplace-agentic-ai-growth/)); individual data-app wins near $1M ARR exist ([Hakkoda](https://hakkoda.io/resources/native-apps-for-snowflake/)) | Payment evidence is the strongest of any path — but the slot "semantic layer as Snowflake Native App" is *already occupied by Honeydew, which Snowflake itself funded*, and Snowflake ships Cortex Analyst + Semantic Views natively | **KILL for solo/side-project.** Requires a product company's effort, competes with Snowflake's own investment, and collides head-on with the Wayvia employer conflict (see below). |
| 4 | **Hosted/cloud version** | Data teams, $80–100/dev/mo benchmarks | Cube Premium $80/dev/mo ([cube.dev/pricing](https://cube.dev/pricing)); dbt Starter $100/user/mo + ~$0.075/queried metric ([getdbt.com/pricing](https://www.getdbt.com/pricing), [Paradime](https://www.paradime.io/guides/dbt-cloud-pricing)); Cortex Analyst 6.7 credits/100 msgs ([Snowflake docs](https://docs.snowflake.com/en/user-guide/snowflake-cortex/pricing)) | **KILL.** People demonstrably pay — to companies with SOC2, SSO, HA, and sales teams. MiniML's entire pitch is "zero-infrastructure embeddable library"; a hosted version negates its one differentiator and enters a knife fight with $48M-funded Cube ([Crunchbase](https://www.crunchbase.com/organization/cube-dev)) and the warehouses themselves. |
| 5 | **Open-core / dual licensing** | Enterprises needing gated features (RBAC, more dialects, governance) | $0 now | Lightdash monetizes open-source BI but needed $19.4M and 5,000 teams on OSS first ([TechCrunch](https://techcrunch.com/2024/10/08/open-source-bi-platform-lightdash-gets-accels-backing-to-bring-ai-to-business-intelligence/)); Vanna sells hosted tiers atop OSS text-to-SQL ([vanna.ai](https://vanna.ai/)) | **KILL for now.** Legal nuance the charter asked about: MIT-no-CLA is NOT a blocker here — [MEASURED] 0 forks/0 outside contributors means Dave is sole copyright holder and can relicense future versions unilaterally. The blocker is that there is no adoption to convert and ~1,040 LOC leaves nothing worth gating. Option stays open; exercising it now yields $0. |
| 6 | **Paid MCP server** | AI-agent builders, per-call or subscription | $0–low hundreds/mo realistic [SPECULATION] | MCP monetization rails exist and pay: Apify reports $1M+/mo total developer payouts, Nevermined/Stripe MPP handle per-call billing ([Zuplo](https://zuplo.com/blog/monetize-an-mcp-server), [Nevermined](https://nevermined.ai/blog/mcp-monetization-ai-agents)) | **KILL as revenue, KEEP as distribution.** [MEASURED] The MCP server doesn't exist yet — it's README keywords. And a metered MCP server that still requires the customer's own warehouse credentials is a bad per-call product. Build a *free* MCP server as the adoption wedge; it's the single highest-leverage missing feature for Paths 1–2. |
| 7 | **Sponsorware / GitHub Sponsors** | Grateful users and their employers | $0–50/mo at current traction; ceiling $800–4k/mo even for tools with big audiences | Documented maintainer income clusters at $800–4k/mo for *popular* tools; Caleb Porzio's $100k/yr is a top-0.1% outlier with a huge Laravel audience ([calebporzio.com](https://calebporzio.com/i-just-hit-dollar-100000yr-on-github-sponsors-heres-how-i-did-it)); <12% of OSS devs earn anything ([markaicode](https://markaicode.com/monetize-open-source-github-income/)); sqlfluff, far more adopted, still runs on community funding ([github.com/sponsors/sqlfluff](https://github.com/sponsors/sqlfluff)) | **KILL.** Sponsorship follows audience; MiniML has none. Setting up a Sponsors page costs nothing and is fine, but it is not a path to value. |
| 8 | **Support contracts / consulting on MiniML itself** | Companies running MiniML in production | $0 | Rill needed a funded team and ~a dozen enterprise customers to reach ~$5M ([TechCrunch](https://techcrunch.com/2022/08/04/rill-wants-to-rethink-bi-dashboards-with-embedded-database-and-instant-ux/)); nobody buys support for a 1,000-LOC MIT library they can read in an afternoon | **KILL.** Consulting on *conversational analytics* (Path 1) is real; support contracts on MiniML are fantasy. |

### The Wayvia constraint (applies to every commercial path)
[MEASURED from dossier context] Author is employed at Wayvia and dogfoods MiniML there for Snowflake-based conversational analytics. Any *commercial* MiniML offering (hosted, Native App, paid MCP) is a product in his employer's own domain — worst case triggers IP-assignment/moonlighting clauses, best case sours the relationship that provides his only production deployment. [SPECULATION on contract terms, but the conflict structure is standard.] Career capital is the one path his employer will actively applaud.

## Strongest point FOR the project
The willingness-to-pay in this exact category is proven and large — Snowflake meters Cortex Analyst per message, dbt bills per queried metric, Cube charges $80/dev/mo, and Snowflake Marketplace partners cleared $100M in six months — so MiniML is aimed at a real, monetizing problem ("governed SQL for LLMs"), and its author demonstrably understands that problem well enough to build a clean, tested, honestly-documented solution. That is precisely the raw material career capital and acqui-hire outcomes are made of.

## Strongest point AGAINST
[MEASURED] After 13 months on npm: 1 GitHub star, 0 forks, 0 issues, 57 downloads/month. Every monetizable comparable (Lightdash, Rill, Vanna, Wren AI, Honeydew, Cube) had either thousands of stars or millions in funding *before* a dollar of revenue — and the free competition here isn't other startups, it's features the warehouses (Cortex Analyst, Semantic Views) and dbt/Cube give away or bundle. There is no niche left between "free and native" and "funded and hosted" for a solo MIT library to charge in.

## Viability score: 14/100
(Direct revenue: ~2/100. Score reflects real but modest career-capital value plus a preserved relicensing option as sole copyright holder.)

## The one question that would change my mind
Will a second company — one that does not employ Dave Templin — put MiniML into production within six months of a real MCP server shipping? A single external production user with a budget would revive open-core and support paths from "kill" to "test"; continued zero would confirm this is a portfolio piece, and a good one.