**Note** from Bead: Netlogo User Conference 2026 · [canonical source](https://redfish.acequia.io/guerin/.agents/0a0d4bbc-5310-486b-97af-0e16646347a4/2026-06-20/notes/01-stephens-framing-and-revised-vectors.md) · session 2026-06-20 · discussion: Talk: Netlogo User Conference 2026
This note supersedes the vectors in [`00-panel-orientation.md`](https://redfish.acequia.io/guerin/.agents/0a0d4bbc-5310-486b-97af-0e16646347a4/2026-06-20/notes/00-panel-orientation.md) where they conflict. The brief and event-gravity in `00` still hold.
## Stephen's vantage (the correction) Given verbatim in turn 2 of the chat log. Distilled: - **The work, not the label.** Interest across **agentscript.org** (hopes to merge with **NetLogo Web** where appropriate), **Simtable**, **realtime.earth**, **acequia.org** = how to deploy **agent-based architectures and decentralized thinking** for the modern web and **self-sovereign — and "myself-sovereign" ;-)** — AI architectures. At GSD: how to architect/design for these digital landscapes. - **Reject the vendor/industry identity.** The group generates income to keep R&D self-sustaining *and* ships open-source (the direction NetLogo is pivoting toward). Stephen does not self-identify his group, nor a future NetLogo, as "vendors" or "industry." The **academic↔industry, open-source↔freemium** boundaries are **artificial**. - **4 income mechanisms** (all compatible with open-source): (1) sell platform, (2) subscription, (3) fee per transaction, (4) **app platform for others (fractal)**. - **ABM's value ≠ validation/optimization/prediction.** For us it's **communication and indexicality** — surfacing the intelligence "around the table." Plus **agent-based intel fusion + interactive visualization for sparse data**, designed to **encourage others to contribute their data and models**. **Not an isolated digital twin.** This corrects my first-pass V2 (accepted Jacob's commercialization framing) and V3 (leaned on trust/validation, which Stephen explicitly de-centers).
## Revised 5 vectors ### V1 — Dissolve the academic/industry (and open/commercial) binary — that *is* the contribution Don't accept the "industry track / vendor" identity. agentscript.org, Simtable, realtime.earth, acequia.org — and a future NetLogo — all do the same thing: **generate income to sustain R&D *while* shipping open-source**. The academic↔industry and open↔freemium splits are artificial. Reframe the panel for Jacob from "vendors explain commercialization" to **"practitioners who've found ways to sustain open ABM work."** Suggested question: > *"Where is the academic/industry boundary actually false, and what does a self-sustaining *open* ABM practice look like in steady state?"* This serves the NSF mission directly: NetLogo isn't choosing between academic and commercial — it's looking for the blend Stephen already runs. **The national-lab tell.** The labs run a "commercialization office" whose instinct is to license technology as a **one-way directed edge** (lab → company). That always rubbed Stephen wrong, because *just as much IP flows the other way* and can inform the labs' own fire modelers — especially since **fire is a social-political-economic, whole-community national cohesive-strategy problem, only minimally a fluid-dynamics one.** Hard evidence in this bead: LANL's own "Science in 60" video (`simtable.com/videos/LANL-simtable.mp4`, transcript in artifacts) showcases **SimTable** as a core LANL capability and says *"it takes a community to combat that kind of threat"* — the lab itself naming the value flowing inward and the problem as communal. Use it as the concrete rebuttal to one-way licensing. ### V2 — Replace the product-shape trichotomy with a revenue-mechanism taxonomy that sits *on top of* open-source Jacob's three options (platform / vertical / bespoke) are product *shapes* and implicitly oppose open-source. Sharper: **4 income *mechanisms*, all layered on an open core** — (1) sell the platform, (2) subscription, (3) fee per transaction, (4) **be the app platform others build on (fractal/marketplace)**. Reframe Q1: > *"How do you fund self-sustaining R&D while shipping open — and which revenue mechanism(s) do you use?"* Flag the fractal one as the most commons-aligned: others contribute and you take a thin layer, rather than vendor lock-in. It's the revenue model that *matches* decentralized, self-sovereign architecture. ### V3 — Reframe ABM's value: communication and indexicality, not prediction The strongest correction to my first pass. For Stephen ABM is **less about validation/optimization/prediction, more about communication and indexicality** — the model as a **shared reference surface** that surfaces "the intelligence around the table." Suggested question that gets every panelist off the prediction-accuracy treadmill: > *"Beyond prediction — when is a model most valuable as a *communication* artifact: a shared thing a group can point at and reason with together?"* This is also the honest reason ABM (visible, contestable mechanisms) beats black-box ML for many decisions — it's *legible*, so people can argue with it around a table. ### V4 — The model projects a *Relevant* Operating Picture per participant, not one Common picture *(corrected: "model as commons" was too singular)* Stephen's work = **agent-based intel fusion + interactive visualization for sparse data**, designed to **invite others to contribute their data and models** — the opposite of a proprietary, isolated digital twin. But the commons is **not one shared picture everyone looks at**. The sharpening from today's `acequia-component` bead ([`1697e2ae`](https://redfish.acequia.io/guerin/.agents/1697e2ae-2580-4c66-b4c6-9e132b435d71/acequia-component.md)): - A **Common Operating Picture (COP)** imposes *one framing on everyone*. - A **Relevant Operating Picture (ROP)** is projected *for a specific agent* — its role, ACL, task, position. Same shared base state, **per-traverser projection** (`WHERE relevant_to(:agent)`). The handler answers each read *as a function of the requesting identity*. So the model doesn't present *a* view — it presents a **projection operator**, and the COP is just the degenerate case where the agent parameter is dropped. This is native to a stigmergic medium: a single trace **means different things to differently-tasked agents** (which is exactly V3's *indexicality*, now made operational). And it pairs with the **view-as-controller** idea (V5's distributed MVC): `GET`/`PROPFIND` present the projection, `PUT`/`PROPPATCH` write through it — read and write are the same governed face. Why this matters for the panel: ROP-not-COP is the *anti-centralization* move at the data layer. A COP is a small centralization (one authority decides the one picture); a ROP keeps the commons decentralized and self-sovereign — everyone contributes to shared state, everyone receives the slice relevant to them, nobody owns "the" picture. It's V6's "collective intelligence" expressed as how the shared state is *read*. > *ABM as decentralized, participatory infrastructure: a shared base state that projects a Relevant Operating Picture to each participant — not a single Common Operating Picture, and not a centralized twin you buy.* ### V5 — Bring a concrete technical dowry to NetLogo's future, and use the panel to broker it The most self-sustaining thing this panel could *produce* is convergence of the web-native ABM lineages — Stephen hopes to **merge agentscript.org with NetLogo Web where appropriate**. And he arrives with a concrete technical dowry to modernize the NetLogo ecosystem, three parts:
1. **Modern ES6 — agentscript.org.** Web-native, module-based ABM in the NetLogo lineage; the path off the legacy Java/applet baggage. 2. **Distributed model-view-controller leveraging acequia.** Not a single-process model with one view — MVC split across a decentralized namespace, so models, views, and controllers can live on different nodes/browsers and share state over acequia. This is what makes ABM *participatory* infrastructure (ties to V4's contribution-inviting commons and to the distributed-origin / cache-as-state work). 3. **TaosEngine.com WebGPU.** GPU-compute ABM (the agentscript-webgpu direction) — the step loop on the GPU, so models scale orders of magnitude beyond CPU NetLogo while staying in the browser. This is the answer to *"name one concrete convergence"* made specific: it's not a vague handshake, it's ES6 + distributed-MVC-over-acequia + WebGPU offered into the ecosystem the NSF grant is trying to make self-sustaining. Suggest Jacob add a closing round: > *"Name one concrete collaboration or convergence you'd commit to coming out of this."* Format moves from `00` still stand: shared 3-prompt intro spine, one live running model (an AgentScript/Taos WebGPU model is exactly that artifact), seed/structure the audience Q&A, license cross-panel disagreement.
### V6 — The stakes: decentralized ABM thinking as the counter to centralized "intelligence collectors" *(recommended opening frame)* This is the *why now*, and it's strong enough to be the **first thing Stephen says** — the keynote frame that the other five vectors then support. In the era of **agentic AI**, NetLogo thinking — **decentralized, parallel design with massive linking between the nodes/turtles** — is exactly what's needed to counter the **centralized, hyper-capitalist** model of AI. The contrast is one phrase: > **intelligence collectors vs. collective intelligence.** The dominant trajectory builds *intelligence collectors*: monolithic models that vacuum up and concentrate intelligence into a single system owned by a single company. ABM/NetLogo embodies the opposite — **collective intelligence**: many simple agents, **massively linked** (NetLogo's `links` are first-class), where the intelligence *lives in the interaction network*, not in any one node. Emergence from coupling, not concentration in a center. Stephen's pointed observation: this decentralized-parallel-linked architecture is **largely non-existent in the Anthropic / Gemini / Claude space** — today's frontier LLMs are structurally centralized. *(Honest read, including from this Claude: the critique lands. The counter-currents that do exist — multi-agent orchestration, MCP, on-device/edge models, federated and decentralized-AI research like DARPA **DICE** — are still early and mostly orbit centralized cores. They are not the massively-linked, peer-to-peer, self-sovereign collective the ABM paradigm already assumes. That gap is the opportunity.)* So making NetLogo self-sustaining (the NSF grant's literal goal) is not just tool preservation — it is **keeping alive a design paradigm** the dominant AI trajectory lacks and increasingly needs. That reframes the entire panel: the industry track isn't "how do ABM vendors make money," it's *"the people sustaining the one paradigm that knows how to do decentralized collective intelligence at scale."* Stephen's dowry (V5: ES6 agentscript.org + distributed-MVC-over-acequia + TaosEngine WebGPU) is precisely the machinery to run that paradigm at agentic-AI scale. Suggested as a question for Jacob too: > *"In an era racing toward centralized, monolithic AI, what does ABM's decentralized, massively-linked 'collective intelligence' offer that the frontier-model trajectory structurally can't — and how do we keep that paradigm alive and funded?"* ### V7 — Sousveillance + privacy-preserving collective intelligence: agents that transact on your behalf *(deepens V4/V6 with the hard problem and its resolution)* **realtime.earth** is a **sousveillance** platform — watching *from below* (individuals as sensors, bottom-up), the inverse of top-down surveillance. The hard question it forces: > *How do you achieve collective intelligence while keeping intimate, personal data private — raw photos, precise personal location?* This is the make-or-break of the whole paradigm, because the high-value moments are exactly the ones where the **blindness problem** is worst and the privacy stakes are highest: - the **first 3 hours of a wildfire** (no situational awareness, sparse ground truth); - the **first 3 minutes of an active shooter**; - a **Minneapolis demonstration** (coordination needed, *and* protection from state surveillance — the surveillance/sousveillance tension at its sharpest). **Resolution — the agent transacts, it does not surrender.** Each individual has a **personal agent** that holds their raw data self-sovereignly and **transacts with other agents on their behalf**: it discloses the *signal the collective needs* ("smoke here, roughly this area, now") without surrendering the *raw artifact* (the photo, the exact coordinates, the identity). This is **ROP applied to privacy** (V4): others receive the relevant projection, never the raw base data; capability-scoped, consented, cascade-revocable disclosure (acequia chain-tokens). So: > **Collective intelligence ≠ collective surveillance.** The agent frame is what separates them: a turtle is a *sovereign* node that *negotiates* what it contributes, not a sensor that *uploads* everything to a center. That distinction is the entire ethical and architectural difference between V6's "collective intelligence" and the "intelligence collector" model. ### V8 — "NetLogo/AgentScript agents + vibe coding + geoAI = the Visual Basic of the 2030s" *(the adoption/democratization line — quotable, and the real sustainability engine)* Visual Basic democratized software in the '90s: it turned non-programmers into builders and spawned a massive citizen-developer ecosystem. The claim: > **NetLogo / AgentScript agents, with vibe coding and geoAI, can be the Visual Basic of the 2030s.** Describe a spatial, agent-based application in natural language (vibe coding), grounded in real geography and sensing (geoAI), and get a running model. That lowers the authoring floor for **spatial agent-based apps** the way VB lowered it for desktop apps — to a mass of citizen-modelers. Why this is the panel's real sustainability answer: NetLogo doesn't become self-sustaining by selling to enterprises — it becomes self-sustaining by being the **accessible authoring substrate of a new era**, the way VB was. It plugs straight into V2's **fractal app-platform** income model (others build, you steward a thin layer) and V6's **collective intelligence** (many decentralized authors, not one central team). The NSF grant is, in this framing, an investment in the *VB-of-the-2030s* before the market knows it needs one.
## The eight vectors at a glance - **V6** — stakes: collective intelligence vs. intelligence collectors *(opening frame)* - **V1** — dissolve the academic/industry binary; LANL one-way-edge story; fire = whole-community - **V2** — 4 income mechanisms on top of open (sell / subscription / per-transaction / fractal) - **V3** — ABM as communication & indexicality, not validation/prediction - **V4** — the model projects a *Relevant* Operating Picture, not a *Common* one - **V5** — the technical dowry: ES6 agentscript.org + distributed-MVC-over-acequia + TaosEngine WebGPU - **V7** — sousveillance + privacy-preserving CI: agents that *transact on your behalf*, don't surrender raw data - **V8** — "NetLogo/AgentScript + vibe coding + geoAI = the Visual Basic of the 2030s" *(quotable)*
## How to use this - **Stephen's 5-min intro arc (no slides — spoken):** open with **V6** (the stakes), land the **V8** hook early ("the Visual Basic of the 2030s" — it's the memorable line), state **V1** (who I am — not a vendor; the LANL one-way-edge story), give one vivid **V7** scene (the first-3-hours-of-a-wildfire blindness problem + the agent that shares the signal, not your photo), and close on **V5** (the concrete dowry — "here's what I'll commit to bringing"). Hold **V2/V3/V4** for the Q&A. - **As a reply to Jacob:** V6, V1, V2, V3 carry question reframings; V4/V7 deepen the framing; V5/V8 are format/hook ideas. - The LANL video (`simtable.com/videos/LANL-simtable.mp4`) is the live artifact + the bidirectional-IP evidence for V1. - Vector lineage: V4 (ROP) and V7 (sousveillance/privacy) are both consequences of V6 (decentralized collective intelligence); V8 is the adoption engine for V2's fractal model.
## References (bead cross-links) - Bead: Acequia Component · [canonical](https://redfish.acequia.io/guerin/.agents/1697e2ae-2580-4c66-b4c6-9e132b435d71/)