PolicyEngine · MaxGhenis · Aug 27, 2026 · Aug 27, 2026
diff --git a/.gitignore b/.gitignore
@@ -9,3 +9,4 @@ tsconfig.tsbuildinfo
 *.tsbuildinfo
 .env*.local
 slideshows/*/*.pdf
+.env*
diff --git a/public/images/iariw-2026/forecast-spf-only.png b/public/images/iariw-2026/forecast-spf-only.png
diff --git a/slideshows/iariw-2026/DEMO-SCRIPT.md b/slideshows/iariw-2026/DEMO-SCRIPT.md
@@ -2,11 +2,11 @@
 
 The deck embeds the LIVE apps directly, each with an Expand button: slide 23 (Axiom App,
 Belgium), 28 (cliff analyzer, axiom.org/snap), 30 (chatbot, axiom.org/chatbot), 36
-(chronicle.institute), 43 (calibration dashboard), 49 (policyengine.org/be), 55 (Thesis
-baseline forecast, app.thesisinstitute.org/spm-child-poverty-2025), 57 (Thesis bill
+(chronicle.institute), 43 (calibration dashboard), 50 (policyengine.org/be), 56 (Thesis
+baseline forecast, app.thesisinstitute.org/spm-child-poverty-2025), 58 (Thesis bill
 analyses, app.thesisinstitute.org/bills). You can drive the whole demo without leaving the
 deck; preloaded tabs remain smoother for deep interaction. Every beat has static screenshot
-slides right after it (24–27, 29, 31, 37, 44, 50, 56, 58). If the network dies mid-beat,
+slides right after it (24–27, 29, 31, 37, 44, 51, 57, 59). If the network dies mid-beat,
 advance and keep talking.
 
 Cliff analyzer beat: Run reform → point at the allotment cliff near $2,100 and the

diff --git a/slideshows/iariw-2026/SPEAKER-NOTES.md b/slideshows/iariw-2026/SPEAKER-NOTES.md
@@ -2,7 +2,7 @@
 
 IARIW–CAPE workshop "New technologies for evidence-based policy making"
 Thursday 27 August 2026 · 14:40–15:30 · Room P02, UCLouvain Saint-Louis, Brussels
-Deck: policyengine.org/slides/iariw-2026 (local: /slides/iariw-2026) · 62 slides, ~50 presented
+Deck: policyengine.org/slides/iariw-2026 (local: /slides/iariw-2026) · 65 slides, ~50 presented
 
 **Room:** IARIW measurement economists; Koen Algoed (Director General, Budget and Finance,
 Flemish Region) and Jean-Baptiste Traversa (head of microsimulation modelling, FPS Finance)
@@ -39,11 +39,12 @@ modeling groups' publication records.
 | 14:55 | 1 · Axiom (app, cliff analyzer, chatbot) | 14–33 (skip statics 24–27, 29, 31) | 14 |
 | 15:09 | 2 · Chronicle | 34–37 (skip 37) | 3 |
 | 15:12 | 3 · Microcosm | 38–45 (skip 44) | 6 |
-| 15:18 | 4 · PolicyEngine | 46–51 (skip 50) | 5 |
-| 15:23 | 5 · Thesis → 6 · Looking forward | 52–62 (skip 56, 58) | 7 |
+| 15:18 | 4 · PolicyEngine | 46–52 (skip 51) | 5 |
+| 15:23 | 5 · Thesis → 6 · Looking forward | 53–65 (skip 57, 59) | 7 |
 
-Cut-first if long: Microplex (37), WhoUsesIt (40), and skim Evo3/Evo5. Static screenshot
-slides (22–25, 36, 42) are presented ONLY if the network dies.
+Cut-first if long: Microplex (45), WhoUsesIt (49), CountryScope (48), and skim Evo3/Evo5.
+Static screenshot slides (24–27, 29, 31, 37, 44, 51, 57, 59) are presented ONLY if the
+network dies.
 
 ---
 
@@ -217,24 +218,31 @@ science team built 10ds-microsim on it; MOUs with NBER (open source TAXSIM emula
 Atlanta Fed (Policy Rules Database). For this room: an open EUROMOD-family stack for the US
 and UK, built in public.
 
-### 48 · Who uses it (30s) — CUT IF LONG
+### 48 · The models, by country (1m) — CUT IF LONG
+Depth by country, counted from each repository's main branch today: US 5,956 parameter
+files / 5,981 variables (defining the 95,000+ dated parameter values); UK 597/872; Canada
+395/393; Israel and Nigeria community seeds; Belgium via Axiom — 107 encoded provisions.
+The point: depth follows demand, and the encoder evolution changes the arithmetic — the
+Belgian provisions arrived in weeks, verified against EUROMOD.
+
+### 49 · Who uses it (30s) — CUT IF LONG
 Logo wall. One sentence and move.
 
-### 49 · DEMO — a Belgian reform, live (~3m; slide 50 is the fallback)
+### 50 · DEMO — a Belgian reform, live (~3m; slide 51 is the fallback)
 policyengine.org/be in-deck: move the top bracket rate, watch budget/Gini/poverty recompute
 (155 precomputed cells, 28 CIR 92 parameters); scroll to "The population, checked" — both
 engines against administrative truth, misses in red with named mechanisms.
 
-### 51 · policyengine.py (1m — works offline)
+### 52 · policyengine.py (1m — works offline)
 The same models as a Python package: a UK household in four lines; the same call for the US
 with a reform attached. This is the interface the Belgian work targets.
 
-### 52–53 · Thesis divider → Conductors, not oracles (2m)
+### 53–54 · Thesis divider → Conductors, not oracles (2m)
 The model routes to verified tools and integrates calibrated outputs; analyst judgment lives
 at every routing decision. The judgment-to-mechanism loop: intuition becomes mechanism over
 time.
 
-### 54 · The loop that matters most (2m)
+### 55 · The loop that matters most (2m)
 Every primitive has its gauge; the one that ranks them all is whether forecasts resolve
 against reality — which is what Thesis exists to do: open forecasts of public outcomes,
 every prediction published with its reasoning and graded when the official number lands.
@@ -244,38 +252,51 @@ We intend to score that. And the deeper point for 16:00: policy takes
 a different shape when baseline conditions change quickly; if you assign real probability to
 that, how does this community arm policymakers to respond?
 
-### 55 · DEMO — a baseline Thesis forecast: SPM child poverty (1.5m; 56 is the fallback)
+### 56 · DEMO — a baseline Thesis forecast: SPM child poverty (1.5m; 57 is the fallback)
 app.thesisinstitute.org/spm-child-poverty-2025: a live forecast cell on a published
 government data point — the 2025 SPM child poverty rate with an 80% interval, the Census
 history, and the agent's full reasoning trace (assumptions AND caveats in the open; the
 trace may show prototype infra notes — that transparency is the design). Graded when the
 Census publishes in September. Start here so the bills demo lands as "the same machinery,
 pointed at legislation."
 
-### 57 · DEMO — Thesis bill analyses (1.5m; 58 is the fallback)
+### 58 · DEMO — Thesis bill analyses (1.5m; 59 is the fallback)
 app.thesisinstitute.org/bills: "start from the bill, derive the outcomes" — each analysis
 reads a bill's provisions, separates countersignable goals from likely effects, and maps
 candidate outcome metrics against the live forecast registry. The opening question, made
 practice. Labeled prototype; say so.
 
-### 59 · Looking forward (divider, 15s)
+### 60 · Looking forward (divider, 15s)
+
+### 61 · The five, as one stack (45s — the layered recap)
+The vertical recap: Axiom and Chronicle side by side at the bottom (the two substrates) →
+Microcosm builds on both (the construction layer, highlighted) → PolicyEngine composes them
+→ Thesis, dotted, on top — the newest layer, deciding which questions matter. One breath per
+layer; the audience has now seen a demo of each.
 
-### 60 · And if the baseline itself moves? (1.5m — the closing forecast beat)
+### 62 · And if the baseline itself moves? (45s — the setup)
 Everything before this slide matters regardless of how the economy evolves; if conditions
-change quickly, being nimble matters more. The chart: professional forecasters disagree less
-than ever about long-run growth (SPF 10-year IQR 0.2pp, half its 1990s level) — but asked
-specifically about AI, economists and forecasters see room for major impact (+0.07pp to
-+30pp of annual growth) [Philadelphia Fed SPF; AI Frontiers compilation]. Landing: if not
-just GDP but unemployment, wage inequality, and the capital income share move, policymakers
-will need even better tools — and AI can help, if we arm it in turn. "After the break, I'm
-looking forward to discussing what those tools can improve in policymaking."
-
-### 61 · It takes all of us (1.5m)
+change quickly, being nimble matters more. The punchline first: professional forecasters
+don't think it will — SPF 10-year growth disagreement is down to a 0.2pp IQR, half its 1990s
+level [Philadelphia Fed SPF]. Beat. "But that's the unconditional question."
+
+### 63 · Asked about AI specifically, the range explodes (1m — the closing forecast beat)
+The right panel appears: published estimates of AI's growth effect span +0.07pp to +30pp of
+annual growth — three orders of magnitude wider than the baseline disagreement [AI Frontiers
+compilation]. The expert survey behind the Yale Budget Lab's fiscal scenarios (Karger et
+al. 2026 — 69 economists, 52 AI experts, 38 superforecasters, fielded Oct 2025–Feb 2026)
+shows the same shape: median economists near trend unconditionally, materially higher
+conditional on rapid AI progress. Landing: if not just GDP but unemployment, wage
+inequality, and the capital income share move, policymakers will need even better tools —
+and AI can help, if we arm it in turn. "After the break, I'm looking forward to discussing
+what those tools can improve in policymaking."
+
+### 64 · It takes all of us (1.5m)
 Model-building has always been about making consequences visible before the choice.
 Statistical offices, EUROMOD, BEAMM, ministries, open models — each doing what it does best.
 "Which is exactly what the roundtable is about — see you at 16:00."
 
-### 62 · Thank you / QR (leave up)
+### 65 · Thank you / QR (leave up)
 QR → axiom.org. Links: axiom.org · policyengine.org/be · both GitHub orgs.
 
 ---

diff --git a/slideshows/iariw-2026/config.ts b/slideshows/iariw-2026/config.ts
@@ -45,6 +45,7 @@ import MicrocosmBESlide from './slides/MicrocosmBESlide';
 import MicroplexSlide from './slides/MicroplexSlide';
 // 4 — PolicyEngine
 import PEOverviewSlide from './slides/PEOverviewSlide';
+import CountryScopeSlide from './slides/CountryScopeSlide';
 import WhoUsesItSlide from './slides/WhoUsesItSlide';
 import DemoPySlide from './slides/DemoPySlide';
 // Live embeds + static fallbacks
@@ -75,7 +76,11 @@ import {
 } from './slides/DemoLiveSlides2';
 // 5 — Together
 import ConductorsSlide from './slides/ConductorsSlide';
-import ForecastUncertaintySlide from './slides/ForecastUncertaintySlide';
+import StackRecapSlide from './slides/StackRecapSlide';
+import {
+  ForecastSpfSlide,
+  ForecastAiRangeSlide,
+} from './slides/ForecastUncertaintySlide';
 import ClosingLoopSlide from './slides/ClosingLoopSlide';
 import CommunityClosingSlide from './slides/CommunityClosingSlide';
 import QuestionsSlide from './slides/QuestionsSlide';
@@ -155,6 +160,7 @@ export const iariw2026Config: SlideshowConfig = {
     // 4 — PolicyEngine (the model)
     EngineDividerSlide,
     PEOverviewSlide,
+    CountryScopeSlide,
     WhoUsesItSlide,
     DemoPeBeLiveSlide,
     DemoPeBeSlide,
@@ -169,7 +175,9 @@ export const iariw2026Config: SlideshowConfig = {
     ThesisLiveSlide,
     ThesisBillsStaticSlide,
     LookingForwardDividerSlide,
-    ForecastUncertaintySlide,
+    StackRecapSlide,
+    ForecastSpfSlide,
+    ForecastAiRangeSlide,
     CommunityClosingSlide,
     QuestionsSlide,
   ],

diff --git a/slideshows/iariw-2026/slides/CountryScopeSlide.tsx b/slideshows/iariw-2026/slides/CountryScopeSlide.tsx
@@ -0,0 +1,98 @@
+import Slide from '@/components/core/Slide';
+import SlideHeader from '@/components/layout/SlideHeader';
+import SlideTitle from '@/components/layout/SlideTitle';
+
+const rows = [
+  { country: 'United States', params: '5,956', vars: '5,981', note: '' },
+  { country: 'United Kingdom', params: '597', vars: '872', note: '' },
+  { country: 'Canada', params: '395', vars: '393', note: '' },
+  { country: 'Israel', params: '2', vars: '19', note: 'early' },
+  { country: 'Nigeria', params: '6', vars: '17', note: 'early' },
+];
+
+export default function CountryScopeSlide() {
+  return (
+    <Slide>
+      <SlideHeader>
+        <SlideTitle>The models, by country</SlideTitle>
+      </SlideHeader>
+
+      <div className="mt-8 grid grid-cols-[1.15fr_0.85fr] gap-10 items-start">
+        <div>
+          <table className="w-full border-collapse">
+            <thead>
+              <tr className="border-b-2 border-gray-200">
+                <th className="text-left text-sm uppercase tracking-wide text-gray-500 pb-3">
+                  Country
+                </th>
+                <th className="text-right text-sm uppercase tracking-wide text-gray-500 pb-3">
+                  Parameter files
+                </th>
+                <th className="text-right text-sm uppercase tracking-wide text-gray-500 pb-3">
+                  Variables
+                </th>
+              </tr>
+            </thead>
+            <tbody>
+              {rows.map((r) => (
+                <tr key={r.country} className="border-b border-gray-100">
+                  <td className="py-3 text-xl text-gray-800">
+                    {r.country}
+                    {r.note && (
+                      <span className="ml-3 text-xs uppercase tracking-wide text-gray-400 align-middle">
+                        {r.note}
+                      </span>
+                    )}
+                  </td>
+                  <td className="py-3 text-xl font-mono text-right text-pe-teal">
+                    {r.params}
+                  </td>
+                  <td className="py-3 text-xl font-mono text-right text-pe-teal">
+                    {r.vars}
+                  </td>
+                </tr>
+              ))}
+              <tr>
+                <td className="py-3 text-xl text-gray-800">
+                  Belgium{' '}
+                  <span className="ml-3 text-xs uppercase tracking-wide text-gray-400 align-middle">
+                    via Axiom
+                  </span>
+                </td>
+                <td
+                  className="py-3 text-xl font-mono text-right text-pe-teal"
+                  colSpan={2}
+                >
+                  107 encoded provisions
+                </td>
+              </tr>
+            </tbody>
+          </table>
+          <p className="mt-4 text-sm text-gray-500 leading-snug">
+            File counts on each repository&apos;s main branch, 27 August 2026;
+            one file per variable. The 5,956 US parameter files define the
+            95,000+ dated parameter values.
+          </p>
+        </div>
+
+        <div className="space-y-5">
+          <div className="content-card p-6">
+            <div className="slide-tag mb-3">The shape of the gap</div>
+            <p className="text-base text-gray-700 leading-relaxed">
+              Depth follows demand: the US and UK models are
+              production-grade; Canada is substantial; Israel and Nigeria are
+              community seeds awaiting the same treatment.
+            </p>
+          </div>
+          <div className="accent-block">
+            <p className="text-base text-gray-700 leading-relaxed">
+              The encoder evolution changes this arithmetic: the Belgian
+              provisions arrived in weeks, verified against EUROMOD &mdash;
+              the recipe every next country inherits.
+            </p>
+          </div>
+        </div>
+      </div>
+    </Slide>
+  );
+}
diff --git a/slideshows/iariw-2026/slides/ForecastUncertaintySlide.tsx b/slideshows/iariw-2026/slides/ForecastUncertaintySlide.tsx
@@ -3,19 +3,51 @@ import Slide from '@/components/core/Slide';
 import SlideHeader from '@/components/layout/SlideHeader';
 import SlideTitle from '@/components/layout/SlideTitle';
 
-export default function ForecastUncertaintySlide() {
+export function ForecastSpfSlide() {
   return (
     <Slide>
       <div className="flex flex-col h-full">
         <SlideHeader>
           <SlideTitle>And if the baseline itself moves?</SlideTitle>
           <p className="text-xl text-gray-600 mt-2">
-            Everything so far matters regardless of how the economy evolves
-            &mdash; but if conditions change quickly, being nimble matters
-            more.
+            Professional forecasters don&apos;t think it will.
           </p>
         </SlideHeader>
 
+        <div className="flex-1 min-h-0 content-card overflow-hidden border border-gray-200 bg-white">
+          <Image
+            src="/images/iariw-2026/forecast-spf-only.png"
+            alt="SPF 10-year US real GDP growth forecasts 1992-2026: the interquartile range narrows to 0.2pp, half its 1990s level"
+            width={2400}
+            height={2160}
+            className="w-full h-full object-contain"
+            priority
+          />
+        </div>
+
+        <div className="mt-4 flex items-baseline justify-between gap-8">
+          <p className="text-lg text-gray-700 leading-snug">
+            The Survey of Professional Forecasters disagrees less than ever
+            about long-run growth: the 10-year interquartile range is down to
+            0.2pp, half its 1990s level.
+          </p>
+          <p className="text-sm text-gray-500 whitespace-nowrap">
+            Philadelphia Fed SPF
+          </p>
+        </div>
+      </div>
+    </Slide>
+  );
+}
+
+export function ForecastAiRangeSlide() {
+  return (
+    <Slide>
+      <div className="flex flex-col h-full">
+        <SlideHeader>
+          <SlideTitle>Asked about AI specifically, the range explodes</SlideTitle>
+        </SlideHeader>
+
         <div className="flex-1 min-h-0 content-card overflow-hidden border border-gray-200 bg-white">
           <Image
             src="/images/iariw-2026/forecast-disagreement.png"
@@ -29,14 +61,15 @@ export default function ForecastUncertaintySlide() {
 
         <div className="mt-4 flex items-baseline justify-between gap-8">
           <p className="text-lg text-gray-700 leading-snug">
-            Professional forecasters disagree less than ever about long-run
-            growth (10-year IQR 0.2pp, half its 1990s level). Asked
-            specifically about AI, economists and forecasters see room for
-            major impact &mdash; published estimates span +0.07pp to +30pp of
-            annual growth.
+            Published estimates of AI&apos;s growth effect span +0.07pp to
+            +30pp of annual growth &mdash; a range three orders of magnitude
+            wider than the baseline disagreement. The expert survey behind the
+            Budget Lab&apos;s fiscal scenarios shows the same shape: median
+            economists near trend unconditionally, materially higher
+            conditional on rapid AI progress.
           </p>
           <p className="text-sm text-gray-500 whitespace-nowrap">
-            Philadelphia Fed SPF; AI Frontiers compilation
+            Philadelphia Fed SPF; AI Frontiers; Karger et al. (2026)
           </p>
         </div>