Models

// Innovation data with correlated lognormal distributions
techData = [
  {name: "SHR Geothermal", weight: 117, min: 0.43, max: 8.82},
  {name: "SMR", weight: 966, min: 0.29, max: 7.64},
  {name: "EGS", weight: 170, min: 0.22, max: 4.41},
  {name: "Plant-Based Protein", weight: 153, min: 0.18, max: 4.04},
  {name: "Cultivated Protein", weight: 117, min: 0.00001, max: 0.07}, // Avoid zero for lognormal
  {name: "Nat Gas + CCS", weight: 313, min: 0.21, max: 5.21},
  {name: "Solar", weight: 497, min: 0.15, max: 4.19},
  {name: "Green Cement", weight: 136, min: 0.13, max: 2.42},

// Cost-effectiveness of messaging campaigns directed at elected representatives. 
// I created the core model and then ran it through Squiggle AI to create the boilerplate. Then I wrote all the docstrings.

// ==== Inputs ====

inputs = {
  @name("Proportion of Votes Within 1% (%)")
  @doc(
    "Proportion of vote outcomes that fall within one percentage point of the pass threshold. Based on historical voting records. The number has varied over time—it was 2.2% from 2000 to 2010, and 5.1% from 2015 to 2025.
  

voll = lognormal({p5: 10.6, p95:18.7}) // $/kWh
//via ERCOT/Brattle — their 95CI fitted to 90% CI due to squiggle idiosyncracies
threshold = 1*10^11 // $
lost_load_kWh = threshold / voll
lost_load_TWh = lost_load_kWh / 10^9
total_us_consumption_TWh = 4070


duration_multiplier = lognormal({p25:2.8, p75:3.2})
//via 

/*
Describe your code here
*/

a = normal(2, 5)

crises__per_decade = 1 to 7
chance_sentinel_identifies_a_crisis_a_week_to_two_months_beforehand = beta(6,10)

// Large scale catastrophe
chance_crisis_is_catastrophic_but_not_existential = beta(3, 100) 
badness_of_catastrophic_risk_vs_existential = 0.1

// Existential catastrophe
chance_crisis_is_existential = beta(1, 100)

/*
Describe your code here
*/

a = normal(2, 5)

questions = 10
forecaster_time_for_intial_forecast = 4 to 20
forecaster_hourly_rate = 100 to 200

initial_forecast_cost = questions * forecaster_time_for_intial_forecast * forecaster_hourly_rate

forecast_update = questions * (0.3 to 1)
update_cost = forecast_update * forecaster_hourly_rate

annual_cost_to_deliver = initial_forecast_cost + update_cost * 52

// How many AI digital workers?
// =====================================================
// Key question: on tasks that AI can currently solve, how many digital workers could OpenAI deploy given their current inference compute stocks?


// Method:
// 1. We start with estimates we've made of OpenAI's compute stocks (currently around 1M H100-equivalents, split roughly equally between Hoppers and GB200s)
// 2. From this, we account for utilization + only some fraction of compute going to inference, and get a daily amount of inference FLOP
// 3. We then make assumptions about of GPT-5 active params, and based on this estimate (a) the number of GPT-5 inference tokens OpenAI could generate in a day. We ensemble this with another estimate of (a), based on OpenAI's announcement that they process 3 billion messages a day.
// 4. We then estimate (b) the number of "tokens" spent by a human employee in a day, based on estimates of human thinking speed, as well as some token usage data from METR

/*
Generated by Squiggle AI. Workflow ID: 168d0158-1934-49d1-9972-32851ae39b32
*/
import "hub:ozziegooen/sTest" as sTest

// Democracy Preservation vs. AI Safety Funding Analysis
// This model examines the relative cost-effectiveness of funding
// democratic election interventions vs. AI safety interventions

// == Core Parameters ==

/*
Cost to morally offset the harm of an LLM subscription by donating to AI safety
organizations.

Variables and their values were written by me; scaffolding and docs were initially written by AI and then heavily edited by me.

Note: Any reference to "AI companies" refers specifically to frontier AI companies that are working toward AGI/ASI.
*/

@name("Revenue to Valuation Ratio")

/*
Generated by Squiggle AI. Workflow ID: 583117f3-e55f-4334-86f4-4d094f193e4e
*/
import "hub:ozziegooen/sTest" as sTest

// Youth Programs Cost-Effectiveness Model
// Comparing Leaf (large, low-cost with select Fellows) vs NT (selective with scholarships)

@name("Model Inputs")
inputs = {

p =.3% to 3% // Percentage
pop = 2M to 16M // Species population
a = 1/(0.2 to 0.3125) // index -- typically 0.25 is used, ie. f^0.25 of species live in fraction 0.5 of the habitat.
total_area = 10M to 15M // global rainforest area in sqkm
fraction = (1-(1-p)^a)/(p*pop) // fraction of forest that must be destroyed to extinct one species
area_per_extinction = total_area * fraction // in sqkm

/*
relevant variables here
p misaligned
p control works
timeline to agi

*/

agi_arrival_year = 28 to 39

// Estimate person-hours to first-draft this supplement, combining inside/outside views
// Results are distributions; key outputs appear in the notebook at the end.

// ===== Inputs =====

@name("Pages per item by section")
@doc("Fixed page targets per question in each section")
pagesPerItem = {
  sec1to3 = 1.5
  sec4 = 0.8

//Sensitivity Analysis Examples in Squiggle

// ===== Basic Model: Investment Returns =====

@name("Base Case Inputs")
baseInputs = {
  @name("Initial Investment ($)")
  @format("$,.0f")
  initialInvestment = 100k
  

// TESTING: is multiplying EVs a good approximation of the EV of a multiplication?

a = 1 to 1000
b = 1 to 1000
c = 1 to 1000
d = 1 to 1000
e = 1 to 1000

// EV of mult
mult = a*b*c*d*e

// Assumptions
total_campaign_benefit = beta({mean: 0.07, stdev: 0.04}) / 2 // div by 2 because we are only increasing half the voters' turnout
labor_share = beta({mean: 0.01, stdev: 0.01})
labor_benefit = total_campaign_benefit * labor_share
default_margin = normal(-0.07, 0.07)  // dan osborn's margin was -7%

// Calculations
n = 10000
labor_samples = sampleN(labor_benefit, n)
margin_samples = sampleN(default_margin, n)

/*
Generated by Squiggle AI. Workflow ID: ec3fdb2e-104c-4847-bbfa-ad46559d669f
*/
import "hub:ozziegooen/sTest" as sTest

// Cost-Benefit Analysis of Prostate Orgasm Exploration
// This model evaluates the expected costs and benefits of pursuing prostate orgasms

// == Input Parameters ==

/*
For an AI coup to happen by 2035:
* Develop very powerful AI systems
* AI systems are catastrophically misaligned
* AI systems are deployed in a coup enabling context
* AI systems succesfully pull off a coup
*/

p_dev_powerful_ai_2035 = 0.8
p_cat_misalignment = 0.5

/*
Generated by Squiggle AI. Workflow ID: ca63c333-eaa6-40be-a4b5-798a2ba5e8d4
David Reinstein -- edited from this starting point. The original model started with https://acbmcostcalculator.ucdavis.edu/
*/
/// Initial code generated by GPT5 based on conversation at https://chatgpt.com/share/68b9b37b-810c-8002-9300-1f6c6a8da252 on 4 Sep 2025.
/// Cultured Meat Cost (CM-COGS) — Simple Scenario Model (v0.1)
/// Audience: economists & non-engineers
/// Epistemic status: starter scaffold with reasonable ranges; replace with your own data.
/// Conventions: all costs in USD; “kg” means edible wet weight. [?of pure CM cells?]
/// cultured meat cost model (simple starter)