Animation Concept

Understanding the stochastic Petri net model — and how to read the animation

1. What the Animation Shows

The animation simulates 2 years of research activity comparing two labs:

The research activity is visualized as a growing network of nodes and connections. Each node represents a research element (idea, method, result, publication) and edges show how they build on each other. The networks grow at different rates — the Augmented Lab's network is denser and more connected because the model predicts higher productivity.

Each simulation run is stochastic — events occur probabilistically, so the exact growth pattern varies between runs. The expected output factor converges to ~1.5x over time, not the 3-5x that vendors claim.

2. The Animation Concept Inspired by Petri Nets

Stochastic Petri Net

A formalism for modeling distributed, concurrent processes with probabilistic outcomes:

The Augmented Lab's model has different transition probabilities (faster boilerplate, faster writing, slower validation) — producing a different marking trajectory than the Traditional Lab. The ~1.5x output factor is the ratio of final markings after equivalent simulated time.

The stochastic nature is essential: a deterministic Petri net would always produce the same result, but real research is full of contingency. The simulation's variability mirrors this.

3. Network Visualization

3.1 Node Types

PI (white, large)Principal Investigator — the central hub. Every network grows outward from here.
PhD researchersEach has a unique color (cyan, pink, yellow). They spawn new research directions and connect to the PI.
Publication draft (green)High-value. A research direction completed and written up — rare in the simulation.
Derivation (purple)Mathematical progress — proofs, derivations, theoretical work.
Numerical result (blue)Computational or experimental results generated.
Consistency check (orange)Validation, debugging, cross-checking — including hallucination checking overhead from AI.
Dead end (red)Abandoned or failed research direction — surrounded by a red aura.
In progress (faded)Research direction still active — not yet resolved as success or failure.

3.2 Node Intensity — What Is Being Modeled

Modeling: Why Nodes Have Different Intensities

The intensity of a node — how bright or dark it appears — is not cosmetic. It models the confidence level of a research element.

Newly spawned nodes appear faint (low confidence). As work progresses and a node is validated by subsequent results (consistency checks, derivations building on it), its intensity increases — it becomes more "real" in the network. Highly connected nodes that have survived multiple checks are rendered with stronger glow.

This mirrors how real research works: early-stage ideas are uncertain and tentative, while results that have been checked and built upon carry more weight. The simulation models this explicitly — nodes are not born equal; they earn their confidence.

Dead-end nodes are rendered with a red aura because they represent a strong negative signal: the research direction was tried and found to fail. The aura marks them as rejected, even after the node is "settled."

3.3 Edges

Lines connect nodes to show dependencies — a new derivation building on a previous result, a publication draft connecting multiple numerical results, and so on. Denser connections mean more cumulative research progress. Edges are rendered with low opacity so they don't obscure the nodes they connect.

3.4 Zoom Behavior

Both sides zoom out over time as the network grows. The Augmented Lab's network is larger, so it zooms out faster to keep everything visible. The relative zoom reflects the productivity gap — you can see the right side's network becoming visually busier faster.

3. Header Elements

3.1 Lab Labels

Each panel is labeled: TRADITIONAL LAB and AUGMENTED LAB. Below each label:

3.2 Event Counter and Progress Bar

Below the lab labels, a small progress bar shows the cumulative number of events logged. The bar fills as events accumulate, giving a sense of ongoing activity in each lab. Each event is counted and displayed as events: N below the bar.

3.3 PhD Color Legend

A small inline legend shows the color coding for each of the three PhD researchers. Their nodes appear in their respective colors throughout the network.

4. Output Factor

The output factor is the ratio of events logged by the Augmented Lab vs. the Traditional Lab. It is displayed between the two panels (or between sections on mobile).

Expected Value: ~1.5x

The model predicts approximately 1.5x output factor — not 3-5x as vendors claim. This is grounded in published RCTs (Peng et al., Noy & Zhang) and independent studies (METR).

5. Productivity Graph

The graph in the center plots the productivity ratio over simulated time:

On the left side of the graph, time is labeled in years (Traditional Lab). On the right side, it is labeled in weeks (Augmented Lab) — though they share the same vertical position, reflecting the different durations. The graph is removed on mobile.

6. Progress Tracker

At the bottom of each panel, the progress tracker shows categorized research activity as filled cells. Each cell represents a logged event — brighter cells indicate higher confidence (more evidence accumulated around that event), fainter cells indicate more recent or uncertain events. When a row's bar fills, it wraps and clears. The tracker rows:

Publication draft
highest value · rarest events
Derivation
mathematical progress · proofs
Numerical result
computational / experimental results
Consistency check
validation · debugging · hallucination checks
Dead end
abandoned · ~33% of directions (OSC 2015)
Modeling — Cell Intensities: Bright cells in any row represent well-established events (high confidence, built upon by subsequent work). Faint cells are recent or tentative. Dark cells represent events that occurred early in the timeline or have been superseded. The orange row (consistency check) tends to accumulate faster in the Augmented Lab because AI hallucination checking adds overhead. The red row (dead end) may show slightly more activity in the Augmented Lab due to AI-generated incorrect results that must be caught and discarded.

7. Related Pages