Field noteNov 17, 20265 min read1 source

World 2030 models the US economy through 2030 across a baseline and two 2028 shock scenarios

In one minute

World 2030 models the US economy through 2030 across a baseline and two 2028 shock scenarios, using nine interacting groups and six-month turns. Its projections put pressure on software seat pricing and junior hiring, while making cryptographic verification a pivotal condition in one branch. These are scenario outputs, not observed economic outcomes.

What changed: A new 2030 simulation forecasts that AI will cut junior desk hiring about 40%, erode per-seat software pricing, and keep physical-inflation pressures high across three economic scenarios.

What happened

World 2030 describes itself as a macroeconomic simulation running from late 2026 through late 2030. It organizes the US economy into nine groups: AI labs, large employers, banks and insurers, utilities, startups, universities, workers, governments, and the Federal Reserve. The groups move in six-month turns. Each starts from the same Q4 2026 snapshot, which the site characterizes as including more than $250 billion a year in cloud capital spending, an 18 GW firm-power supply against 19.4 GW of demand, a 3.75% Fed rate, roughly 2.9% inflation, and 82% of legacy corporate IT unready for autonomous agents.

The simulation has three branches. The baseline has no added shock. In the Automated Assurance branch, a single early-2028 shock has insurers, regulators, and courts accept machine-checkable cryptographic execution proofs, which the site says cut verification and migration costs by 80%. In the Credit Crunch branch, private-credit funds gate redemptions, spreads rise by 250 basis points, and Congress passes a No-Bailout Act. The site says that, apart from those two specified shocks, later events are not scripted. Groups act simultaneously without seeing one another’s current-turn choices, and a referee resolves collisions using constraints that include power construction lead times, a 95% physical/5% consumer-digital CPI split, a separate B2B seat-software index, and a GDP-growth identity based on productivity and hours worked.

Across all branches, the model projects consumer digital prices falling about 4% to 5% annually by 2030, while physical services keep the larger CPI basket near 3% annual inflation. It projects legacy seat-software index values of 52.6 in the baseline, 32.0 in Automated Assurance, and 40.4 in Credit Crunch, with late 2026 set at 100.0. The site also projects US entry-level desk hiring at 61.4%, 58.0%, and 63.6% of its 2026 level, respectively. In its account, human-sign-off requirements do not preserve those jobs: employers freeze campus recruiting and move review work offshore or, in the assurance branch, to automated provers.

The branches differ most sharply on corporate IT and equity outcomes. Legacy corporate IT readiness reaches a projected 30.2% in the baseline, 60.2% with Automated Assurance, and 36.4% in Credit Crunch. Tech equities end at 77.8, 109.2, and 84.8 on a 2026 index of 100. The site attributes the assurance branch’s higher endpoint to proofs lowering migration costs, while the credit-crunch branch clears troubled debt sooner through Chapter 11 auctions. For power, projected demand and firm supply converge by 2030 in all three branches, at roughly 32–41 GW.

What the archive adds

New Runtime’s earlier coverage, “ChatGPT’s expanding world of work”, described OpenAI’s quarterly data on ChatGPT use across work activities and countries. That is related context on AI and work, but World 2030 is a new event: it is a forward-looking macroeconomic simulation rather than a usage dataset.

Trend signal

This does not yet fit a pattern New Runtime tracks. World 2030 is a scenario study about interactions among economic actors, not evidence of a durable product boundary or operating model. Its emphasis on cryptographic proof as a condition for broader autonomous IT deployment may warrant future comparison with developments in verification, but this simulation alone does not establish a pattern.

Why now

The simulation starts in late 2026 and treats early 2028 as the point where either of its two modeled shocks occurs. That framing makes the work a near-term scenario exercise: it asks how current pressures on power, software contracts, rates, and recruiting could interact over the next four years, rather than extrapolating a single trend in isolation.

Connections to other events

New Runtime’s earlier coverage of ChatGPT usage across work activities offers a useful contrast: observed usage data can describe where people use a product, while World 2030 projects how employers, workers, utilities, lenders, and policymakers might respond to interacting economic pressures. The two address different questions and should not be treated as corroborating evidence for the simulation’s forecasts.

Why it matters

For teams building with AI, the simulation’s most useful contribution is its explicit dependency structure. Lower inference costs do not automatically reduce the cost of living or solve power constraints; cheaper digital services coexist in the model with expensive physical services and substantial infrastructure spending. Likewise, automating code review is not enough to make legacy systems agent-ready if insurers, regulators, and lenders still require costly manual verification.

That distinction can improve planning. A software buyer can stress-test renewal assumptions separately from AI adoption, while an engineering leader can treat verification and migration capacity as constraints rather than assuming that more capable models translate directly into deployed automation. The model also highlights labor-market consequences: the projected hiring decline appears in every branch, including the branch where automated proofs improve corporate IT readiness. These are not forecasts to budget as certainties, but they are coherent questions to put into scenario planning.

Limits and uncertainty

World 2030’s figures are outputs of a constructed simulation, not independently verified forecasts or observed outcomes. The page does not establish that the modeled starting conditions will persist, that the groups’ assumed decisions will match real behavior, or that the specified shocks will occur. Its results also depend on explicit accounting rules and a narrow set of three branches; other shocks or responses could change the path. The page presents precise endpoints, but precision within a simulation does not establish predictive accuracy. No independent validation of the model’s assumptions or results is described.

Reader takeaway

Read World 2030 as a structured set of conditional scenarios, not as a prediction that software valuations or hiring will land at its endpoint numbers. For builders, the durable prompt is to test how verification costs, legacy-system readiness, power access, and seat-based contracts interact in their own plans—and to keep the assumptions visible when those plans change.

Evidence / sources

  1. 1world-2030.pages.dev
All Field Notes

Next reads

Read next

Open the graph
  1. 01topicModelsExplore the models topic hub.
  2. 02archiveField NotesOpen the latest editorial analysis.
  3. 03sourcesSourcesSee the original sources this archive cites.
Who read this page?Machine requests, hidden until opened

Loading the privacy-safe route aggregate...

Open the JSON contract