Theses

A thesis is a forward statement with a horizon, a named mechanism, premises bound to numeric series or tracked policy instruments, and a falsifier. Strength is computed from the premises by the nightly check, never written by a model. It is premise support, not a probability: strength is premise support, not a probability: no thesis has resolved yet (0 of 10 resolved theses needed before any thesis may be shown as HIGH).

The US 10-year stays above 4.5% through 2027 because disinflation stalled

horizon 2027-12-31 · active · MEDIUM

Every monthly 10-year Treasury print from now to 31 December 2027 is at or above 4.5%.

0.99
premises holding

Resolution test: met US 10-Year Treasury Yield = 5.13 (2026-10-01), test gte 4.5

Sticky core inflation has ceased its decline, prompting a repricing of the long end of the yield curve. The Atlanta Fed core sticky CPI, which had bottomed at 2.70% in August 2026, rose to 3.06% in October, indicating a flat six-month trend rather than a downward one. This stalling of disinflation has led to an increase in the policy rate from 3.63% during the summer to 3.88% in October. Consequently, the 10-year Treasury yield has climbed from 4.48% in May to 5.13% in October, steepening the spread over the federal funds rate to 1.25 points.

The resolution test has already been met, with the 10-year Treasury yield at 5.13% as of October 2026. All premises supporting the forecast are holding, including the rising trend of the 10-year yield and the core sticky CPI remaining above 2.75%. The policy rate is also above 3.5%, indicating no imminent easing cycle. However, the unemployment rate remains below 4.5% at 4.1%, and oil prices are above $70, both of which refute potential downward pressures. For the forecast to be invalidated, a monthly 10-year Treasury print below 4.5%, a core sticky CPI falling below 2.5%, or unemployment rising above 4.5% would need to occur.

holding The 10-year Treasury yield is at or above 4.5%
US 10-Year Treasury Yield = 5.13 (2026-10-01), test gte 4.5
1.00
holding The 10-year is still rising, not merely high
US 10-Year Treasury Yield trend 0.14 per period over 6 points (2026-05-01 to 2026-10-01), test rising 0.05
1.00
holding Core sticky CPI is at or above 2.75% year on year
US Core Sticky CPI = 3.06 (2026-10-01), test gte 2.75
1.00
holding Core sticky CPI has stopped falling
US Core Sticky CPI trend -0.008 per period over 6 points (2026-05-01 to 2026-10-01), test gte -0.05
0.92
holding The policy rate is at or above 3.5%, so the long end is not pricing an easing cycle
US Federal Funds Rate = 3.88 (2026-10-01), test gte 3.5
1.00
broken Unemployment has risen above 4.5%, which would force cuts and pull the long end down (counter-argument, scored against the thesis)
US Unemployment Rate = 4.1 (2026-10-01), test gte 4.5
0.00
broken Oil has fallen under $70, which would drag headline inflation down with it (counter-argument, scored against the thesis)
WTI Crude Oil Price = 89.62 (2026-11-11), test lte 70
0.00
Falsifier. One monthly GS10 print below 4.5% kills it outright. The mechanism dies earlier if sticky core falls back under 2.5% or unemployment rises above 4.5%, either of which puts cuts back on the table.
Resolution rule. True if no monthly observation of us_treasury_10y between seeding and 2027-12-31 is below 4.5. The premise reads the LATEST print, so a dip that recovers would be missed by the daily check: at the horizon, resolve by scanning the stored monthly series rather than trusting the last value.
eur_revenue_bookingrented_model_dependency
last checked 2026-11-18 15:51:26 · narrative 2026-11-17 13:09:14

The euro stays weak enough through 2027 to keep a EUR-booked, USD-spending business squeezed

horizon 2027-12-31 · active · MEDIUM

On 31 December 2027 the USD/EUR rate is still below 1.20.

0.88
premises holding

Resolution test: met USD / EUR Exchange Rate = 1.1 (2026-11-13), test lte 1.2

The mechanism at play involves a business model where revenue is invoiced in euros while significant costs, such as infrastructure and SaaS tooling, are priced in US dollars. This creates a financial squeeze as the euro weakens against the dollar, eroding purchasing power. The exchange rate has already moved unfavorably, with the USD per EUR falling from 1.166 in August 2026 to 1.103 in November, representing a 5.4% loss in purchasing power over three months.

Currently, the premises supporting this thesis are holding strong. The USD per EUR rate remains below 1.20, with a current rate of 1.1, and the euro continues to lose ground, trending downward by 0.001 per period over 60 points. The US rate differential remains a significant factor, with the US 10-Year Treasury Yield at 5.13, well above the 4.5 threshold. Additionally, rising volatility, as indicated by the VIX trend of 0.094 per period, supports risk-off flows that favor the dollar.

However, the euro-area deficit is a weakening premise. It is currently at -3% of GDP, which could potentially support the euro if it improves further. The resolution test has already been met, but the thesis could be challenged if the US 10-year yield drops below 4.5% or if the euro-area fiscal position strengthens significantly.

holding USD per EUR is below 1.20
USD / EUR Exchange Rate = 1.1 (2026-11-13), test lte 1.2
1.00
holding The euro is still losing ground, not stabilising
USD / EUR Exchange Rate trend -0.001 per period over 60 points (2026-08-19 to 2026-11-13), test falling -0.0005
1.00
holding The US rate differential that pulls flows into the dollar is still there
US 10-Year Treasury Yield = 5.13 (2026-10-01), test gte 4.5
1.00
holding Volatility is rising, and risk-off flows favour the dollar
CBOE Volatility Index (VIX) trend 0.094 per period over 60 points (2026-08-24 to 2026-11-13), test rising 0
0.81
weakening The euro-area deficit is back inside 3% of GDP, which would support the euro (counter-argument, scored against the thesis)
Euro area (20) government net lending/borrowing (% GDP) = -3 (2025), test gte -3
0.50
Falsifier. USD per EUR at or above 1.20. The mechanism fails earlier if the US 10-year drops under 4.5%, removing the rate differential, or if euro-area fiscal credibility improves enough to pull flows back - the euro-area deficit returning inside 3% of GDP is the measurable version and is carried as a refuting premise.
Resolution rule. True if the latest fx_usd_per_eur observation on or before 2027-12-31 is below 1.20. Mechanical, read from macro_observations.
eur_revenue_bookingrented_model_dependencysolo_operator
last checked 2026-11-18 15:51:26 · narrative 2026-11-17 13:09:16

The EU keeps regulating AI without taxing its output through 2028

horizon 2028-12-31 · active · MEDIUM

On 31 December 2028 the EU AI rule count is at least 12 while no instrument whose base is AI output or AI compute has reached proposed status.

0.87
premises holding

Resolution test: not yet met Tracked instruments taxing AI output (EU + member states) = 0 (2026-11), test gte 1

The European Union has developed a robust framework for regulating AI systems, yet it has not established a tax base on AI output or AI compute. Currently, there are 10 AI-specific regulatory instruments in force, while no instruments targeting AI output or compute have reached even the proposed status. This indicates a clear preference for regulation over taxation in the AI domain. The EU's approach to digital taxation has historically focused on taxing revenue, with five instruments targeting digital services revenue, rather than attempting to measure and tax novel outputs like AI.

The evidence supports the thesis that regulation is more politically feasible and administratively straightforward than implementing a new tax regime. AI's increasing salience, as evidenced by its share of generated claims rising from 1.2% to 3.0% since May 2026, further supports the likelihood of continued regulatory activity. The premise that an AI-output tax instrument has reached proposed status is broken, reinforcing the current trajectory.

For this thesis to be falsified, we would need to see at least one AI-output or AI-compute tax instrument reach proposed status. Additionally, if the AI rule count fails to reach 12 by the end of 2028, it would suggest a broader regulatory stagnation rather than a focus on instrument choice. As of now, the resolution test is not met, and the computed strength of the thesis stands at 0.87.

holding The EU has at least 8 AI-specific instruments in force
Tracked EU instruments regulating AI systems = 10 (2026-11), test gte 8
1.00
holding When Europe does tax digital activity it taxes revenue, and there are at least 4 such instruments
Tracked instruments taxing digital services revenue = 5 (2026-11), test gte 4
1.00
holding AI's share of this platform's generated claims is rising, so the subject is getting more salient
Share of generated claims whose category is tech_ai (monthly, %) trend 0.424 per period over 6 points (2026-05 to 2026-10), test rising 0.2
1.00
holding AI is at least 2% of generated claims, so the salience is a level and not just a slope
Share of generated claims whose category is tech_ai (monthly, %) = 3.03 (2026-10), test gte 2
1.00
broken An AI-output or AI-compute instrument has reached proposed status, which ends this thesis (counter-argument, scored against the thesis)
Tracked instruments taxing AI output (EU + member states) = 0 (2026-11), test gte 1
0.00
holding Euro-area debt is high enough that a new tax base stays attractive to finance ministries (counter-argument, scored against the thesis)
Euro area (20) general government gross debt (% GDP) = 87.7 (2025), test gte 85
0.77
Falsifier. policy_ai_tax_count_eu reaches 1 or more at MEDIUM evidence or better, which is the same observation that confirms eu-ai-output-tax-2027. Secondary: the AI rule count stalls below 12, meaning the regulatory route stopped too and the mechanism was about EU paralysis rather than instrument choice.
Resolution rule. True if, at the horizon, policy_ai_rules_count_eu is at least 12 AND policy_ai_tax_count_eu is 0. The second half needs the same human confirmation as eu-ai-output-tax-2027: a human checks that nothing found by the EUR-Lex probes actually taxes AI output before this resolves true.
eu_residencyrented_model_dependencyeu_customer_base
last checked 2026-11-18 15:51:26 · narrative 2026-11-17 13:09:19

Singapore keeps out-growing Italy per head while ageing faster than it

horizon 2029-12-31 · active · MEDIUM

On 31 December 2029, Singapore's GDP-per-capita growth exceeds Italy's by more than 1.5 points at the latest common World Bank print, while Singapore's 65+ share is still rising faster than 0.4 points a year.

0.85
premises holding

Resolution test: met wb_gdppc_growth_sg minus wb_gdppc_growth_it = 3.15 (3.75 vs 0.6, 2025), test gte 1.5

The economic growth of Singapore compared to Italy serves as a controlled experiment in understanding the impact of demography on economic performance. Despite Singapore's faster ageing population, with its 65+ share increasing by 0.51 points annually compared to Italy's 0.34, and a lower fertility rate of 0.97 births per woman against Italy's 1.18, Singapore continues to outpace Italy in GDP-per-capita growth. In 2025, Singapore's GDP per capita grew by 3.75%, significantly higher than Italy's 0.60%, maintaining a growth gap of 3.15 points.

Currently, all supporting premises are holding. Singapore's GDP-per-capita growth exceeds Italy's by more than 1.5 points, and its 65+ share is rising faster than 0.4 points a year. Singapore's fertility rate remains below Italy's, and Italy's population is confirmed to be genuinely old. However, the premise that Singapore's growth trend is declining is broken, as its trend is only slightly negative at -0.097 per period. Italy's growth is accelerating, with a trend of 0.431 per period, which weakens the claim but does not yet falsify it.

The picture would change if the growth gap closes below 1.5 points in two consecutive prints or if Italy's growth trend increases by more than 0.4 points a year over six prints. The resolution test has already been met.

holding Singapore's GDP-per-capita growth exceeds Italy's by more than 1.5 points
wb_gdppc_growth_sg minus wb_gdppc_growth_it = 3.15 (3.75 vs 0.6, 2025), test gte 1.5
1.00
holding Singapore's 65+ share is rising faster than 0.4 points a year, so it is not out-growing Italy by staying young
Singapore population aged 65+ (% of total) trend 0.507 per period over 6 points (2020 to 2025), test rising 0.4
0.86
holding Singapore's fertility is at or below 1.05, worse than Italy's
Singapore total fertility rate = 0.97 (2024), test lte 1.05
0.90
holding Italy is genuinely old, so the comparison is not against a young economy
Italy population aged 65+ (% of total) = 25.1 (2025), test gte 24
1.00
broken Singapore's per-capita growth is itself trending down, which would close the gap without Italy improving (counter-argument, scored against the thesis)
Singapore GDP per capita growth (annual %) trend -0.097 per period over 6 points (2020 to 2025), test falling -0.3
0.16
weakening Italy's per-capita growth is accelerating, which would close the gap from the Italian end (counter-argument, scored against the thesis)
Italy GDP per capita growth (annual %) trend 0.431 per period over 6 points (2020 to 2025), test rising 0.4
0.55
Falsifier. The growth gap closes below 1.5 points at the latest common print, in two consecutive prints. Secondary: Italy's own per-capita growth trend turns up by more than 0.4 points a year over six prints, which would mean the gap is closing from the Italian end rather than Singapore slowing.
Resolution rule. True if, at the horizon, wb_gdppc_growth_sg minus wb_gdppc_growth_it at the latest common period is above 1.5 AND the wb_pop65_sg six-print slope is above 0.4. Both read from macro_observations; mechanical.
italian_residencysolo_operator
last checked 2026-11-18 15:51:26 · narrative 2026-11-17 13:09:24

France closes most of its debt gap to Italy by the end of 2029

horizon 2029-12-31 · active · MEDIUM

On 31 December 2029 the latest Eurostat print has Italy's debt/GDP exceeding France's by less than 12 points.

0.75
premises mostly holding

Resolution test: not yet met eu_debt_gdp_it minus eu_debt_gdp_fr = 21 (136.7 vs 115.7, 2025), test lte 12

The mechanism driving the convergence of debt ratios between Italy and France is straightforward arithmetic. Italy is reducing its debt-to-GDP ratio by maintaining a primary balance near zero and achieving nominal growth above the effective interest rate, while France's debt ratio is increasing due to the opposite conditions. As a result, the gap between their debt ratios is narrowing.

Currently, the resolution test has not been met, as Italy's debt-to-GDP ratio still exceeds France's by 21 points, not yet reaching the target of less than 12 points. However, several premises support the forecast. Italy's debt-to-GDP ratio is falling at a rate of 3.709 points per year, faster than the required 2 points, and its deficit is within the target at -3.1% of GDP. France's debt-to-GDP ratio is not falling, with a trend of 0.043 points per year, and its deficit remains wider than 4% at -5.1% of GDP, feeding its debt ratio.

For the forecast to be falsified, the gap must stop closing. This could occur if Italian debt/GDP stops falling at the current pace or if French debt/GDP starts to decline, particularly if France's deficit narrows to inside -4% of GDP. As of now, these conditions are not met, supporting the forecast's strength of 0.75 out of 1.

broken Italy's debt/GDP still exceeds France's by less than 12 points
eu_debt_gdp_it minus eu_debt_gdp_fr = 21 (136.7 vs 115.7, 2025), test lte 12
0.00
holding Italian debt/GDP is falling faster than 2 points a year
Italy general government gross debt (% GDP) trend -3.709 per period over 6 points (2020 to 2025), test falling -2
1.00
holding French debt/GDP is not falling
France general government gross debt (% GDP) trend 0.043 per period over 6 points (2020 to 2025), test gte -0.5
1.00
holding France's deficit is wider than 4% of GDP, so its debt ratio keeps being fed
France government net lending/borrowing (% GDP) = -5.1 (2025), test lte -4
1.00
holding Italy's deficit is at or inside -3.5% of GDP, so its debt ratio keeps falling
Italy government net lending/borrowing (% GDP) = -3.1 (2025), test gte -3.5
0.70
broken France is consolidating faster than a point of GDP a year, which would stop the gap closing (counter-argument, scored against the thesis)
France government net lending/borrowing (% GDP) trend 0.591 per period over 6 points (2020 to 2025), test gte 1
0.09
Falsifier. The gap stops closing: either Italian debt/GDP stops falling (slope above -2 points a year over six prints) or French debt/GDP starts falling (slope below -0.5 a year), in two consecutive prints. A French consolidation that takes the deficit inside -4% would do it on its own.
Resolution rule. True if, at the horizon, eu_debt_gdp_it minus eu_debt_gdp_fr at the latest common period is below 12. Read straight from macro_observations, so this resolves without human judgement.
italian_residencyeu_residencyeur_revenue_booking
last checked 2026-11-18 15:51:26 · narrative 2026-11-17 13:09:21

Italy's fiscal and demographic gap to the euro-area average stays wide through 2028

horizon 2028-12-31 · active · MEDIUM

On 31 December 2028, Italy's debt/GDP still exceeds the euro-area average by more than 45 points and Italy's old-age expenditure share still exceeds the EU27 average by more than 3 points of GDP.

0.60
premises mostly holding

Resolution test: met eu_debt_gdp_it minus eu_debt_gdp_ea20 = 49 (136.7 vs 87.7, 2025), test gte 45

Italy's fiscal and demographic challenges are primarily driven by its aging population and substantial debt burden. These structural issues are exacerbated by Italy's higher old-age dependency ratio and declining fertility rate compared to the euro-area averages. As of 2025, Italy's debt-to-GDP ratio stands at 136.7%, significantly higher than the euro-area average of 87.7%, maintaining a gap of 49 points. Additionally, Italy's old-age expenditure is 13.9% of GDP, exceeding the EU27 average by 3.3 points as of 2024. The old-age dependency ratio further underscores the demographic strain, with Italy at 39.0 compared to the EU27's 34.4, a gap of 4.6 points.

Despite these challenges, there are signs of improvement. Italy's deficit has narrowed to -3.1% of GDP, close to the 3% threshold, and its debt-to-GDP ratio is decreasing at a rate of 3.709 points per year. However, these improvements have not yet closed the fiscal gaps below the thresholds required for resolution. For the forecast to be falsified, Italy's debt gap would need to fall below 45 points, or its old-age expenditure gap below 3 points, in two consecutive Eurostat prints. Until then, the evidence suggests that Italy's fiscal and demographic gaps will remain significant through 2028.

holding Italy's debt/GDP exceeds the euro-area average by more than 45 points
eu_debt_gdp_it minus eu_debt_gdp_ea20 = 49 (136.7 vs 87.7, 2025), test gte 45
0.90
weakening Italy's old-age expenditure exceeds the EU27 average by more than 3 points of GDP
eu_pension_exp_gdp_it minus eu_pension_exp_gdp_eu27 = 3.3 (13.9 vs 10.6, 2024), test gte 3
0.65
holding Italy's old-age dependency ratio runs more than 4 points above the EU27 ratio
eu_olddep_it minus eu_olddep_eu27 = 4.6 (39 vs 34.4, 2025), test gte 4
0.70
holding Italian fertility is still falling, so the dependency gap is still being built
Italy total fertility rate trend -0.016 per period over 6 points (2019 to 2024), test falling 0
0.89
holding Italy's deficit is back inside shouting distance of the 3% limit, which would let the debt gap close (counter-argument, scored against the thesis)
Italy government net lending/borrowing (% GDP) = -3.1 (2025), test gte -3.5
0.70
holding Italian debt/GDP is falling faster than 2 points a year, which would close the gap on its own (counter-argument, scored against the thesis)
Italy general government gross debt (% GDP) trend -3.709 per period over 6 points (2020 to 2025), test falling -2
1.00
Falsifier. Either gap closes below its threshold in two consecutive Eurostat prints: debt gap under 45 points, or old-age expenditure gap under 3 points of GDP.
Resolution rule. True if, at the horizon, the latest published Eurostat prints give eu_debt_gdp_it minus eu_debt_gdp_ea20 > 45 AND eu_pension_exp_gdp_it minus eu_pension_exp_gdp_eu27 > 3. Both are read from macro_observations, so this one resolves without human judgement.
italian_residencyeur_revenue_bookingsolo_operator
last checked 2026-11-18 15:51:26 · narrative 2026-11-16 15:36:20

An AI-output tax instrument appears in the EU by end-2027

horizon 2027-12-31 · active · MEDIUM

By 31 December 2027 at least one instrument whose base is AI output or AI compute reaches "proposed" status or beyond, at EU level or in an EU member state.

0.60
premises mostly holding

Resolution test: not yet met Tracked instruments taxing AI output (EU + member states) = 0 (2026-11), test gte 1

The euro area's high debt level, projected at 87.7% of GDP in 2025, combined with the rising old-age dependency ratio, creates a fiscal environment where finance ministries are compelled to find new tax bases. The EU27's old-age expenditure already accounts for 10.6% of GDP, and the dependency ratio continues to rise at approximately 0.463 points per year. This growing fiscal pressure mirrors the conditions that led France, Italy, and Spain to implement digital services taxes when an EU-level proposal stalled. Similarly, AI output presents a concentrated and mostly foreign-incident base, and the existing AI Act provides the necessary supervisory framework for such a levy.

Currently, the resolution test for the forecast is not met, as no AI-output tax instrument has reached proposed status. However, all premises supporting the mechanism are holding. The euro area's debt remains high, the old-age dependency ratio is increasing, and member states have a history of taxing digital bases unilaterally. The EU also has AI-specific supervisory machinery in place. The main obstacle is the absence of a proposed AI-output tax instrument, and euro-area deficits have returned to the 3% limit, slightly weakening the fiscal motive. A change in these conditions, particularly the introduction of a proposed AI-output tax, would alter the forecast.

holding The euro area carries a debt level high enough that no member state has room to cut taxes elsewhere
Euro area (20) general government gross debt (% GDP) = 87.7 (2025), test gte 85
0.77
holding The EU27 old-age dependency ratio is still rising, so the age-related claim on public money keeps growing
EU27 old-age dependency ratio trend 0.463 per period over 6 points (2020 to 2025), test rising 0
1.00
holding EU27 old-age expenditure is already a double-digit share of GDP
EU27 old-age public expenditure (% GDP, COFOG GF1002) = 10.6 (2024), test gte 10
0.80
holding Member states have already shown they will tax a digital base unilaterally when the EU-level instrument stalls
3 instrument(s) matching {"base":["digital_services_revenue"],"min_status":"in_force"}; furthest ladder 1.00; count+ladder = 3; France: taxe sur les services numeriques (3% digital services tax), Loi n. 2019- [in_force]
0.75
holding The EU already has AI-specific supervisory machinery in force that a levy could attach to
5 instrument(s) matching {"base":["ai_systems"],"min_status":"in_force"}; furthest ladder 1.00; count+ladder = 5; Regulation (EU) 2026/1744 of the European Parliament and of the Council of 8 Jul [in_force]
1.00
broken At least one instrument whose base is AI output exists at proposed status or beyond
Tracked instruments taxing AI output (EU + member states) = 0 (2026-11), test gte 1
0.00
weakening Euro-area deficits have returned to the 3% limit, which would relieve the motive for a new base (counter-argument, scored against the thesis)
Euro area (20) government net lending/borrowing (% GDP) = -3 (2025), test gte -3
0.50
Falsifier. policy_ai_tax_count_eu is still 0 at the horizon, with no AI-output instrument found by the EUR-Lex probes or added to the curated national list. Secondary falsifier: euro-area debt drops below 85% of GDP and the old-age dependency ratio stops rising, removing the fiscal motive the mechanism rests on.
Resolution rule. True if, on or before 2027-12-31, policy_instruments holds at least one row with base in (ai_output, ai_compute) at status proposed or beyond and evidence_tier MEDIUM or better. False otherwise. jobs/thesis/check.php flags the horizon; a human confirms the instrument really taxes AI output rather than merely mentioning AI, then runs the resolve path.
italian_residencyeu_residencyeur_revenue_bookingrented_model_dependencyeu_customer_base
last checked 2026-11-18 15:51:26 · narrative 2026-11-16 15:36:17