The Dispatch is editorial, not schedule-bound. We publish when there is an argument worth publishing.
The Dispatch is editorial, not schedule-bound. We publish when there is an argument worth publishing.
Issues
15 issues
Bear cases
Corrections and retractionsEach issue states the condition under which its argument fails, and the date that condition resolves. Outcomes are recorded here, whichever way they go.
If an operator contract is awarded and at least three hubs are running named cohorts by the end of 2027, the design worked and the funding concern here was a misreading of a first tranche. If no operator has been contracted and no hub has opened by then, NAAI joins the National AI Fund as a second AI vehicle without an operational footprint — and the constraint was never money.
If Pakistan’s technology exports keep growing at double digits through FY2026-27 and either a named foreign cloud or compute commitment lands or a published division of labour between the AI institutions appears, the thirteen-month build was a compounding foundation. If exports flatten and no coordination document is published by the end of 2027, this issue over-read velocity as direction.
If blended-finance mobilisation triples to approximately USD 45 billion a year by 2028 and shifts materially toward AI and development transactions with published inclusion criteria, the thesis strengthens; if it does not, capital work moves to the centre of the programme.
If Statnett or the Ministry of Energy publishes an allocation methodology with public-benefit criteria before the end of 2027, the governance gap named here is closed.
If the national AI fund disburses to a published pipeline and training numbers are independently verified before the end of 2028, the drift reading in this issue was wrong.
If the next adoption dataset shows the North–South gap narrowing for two consecutive years, diffusion is self-correcting and literacy policy is not the binding constraint.
If Norges Bank Investment Management publishes the screening methodology, including model scope and human-review thresholds, before the end of 2027, the transparency gap this issue rests on closes.
If a follow-up federal review before the end of 2028 finds AI-specific clauses in a majority of sampled contracts, the diagnosis was a transition artefact rather than a structural gap.
If, by the end of 2028, no national procurement authority in the EU or EEA has published AI-specific contract clauses covering literacy, evaluation and exit, then the tender is not where these decisions are being made.
If no AI cooperation instrument between a Nordic state and a large Global South state reaches signature by the end of 2028, the climate analogue does not transfer.
If a state demonstrably holds all three of control, capability and openness through 2028 without trading one away, the trilemma is not a trilemma.
If Norwegian grid capacity reserved for data centres converts into domestic model training and serving capacity by 2029, rather than into export-oriented colocation, the mismatch closes on its own and no industrial policy was needed.
If a European public buyer adopts inclusion conditions in an AI tender and the resulting deployment shows no measurable difference in who the system works for, then procurement is not the lever this argues it is.
If the Nordic states publish a joint AI governance position by the end of 2027 that survives contact with a divergent national procurement decision, the "brand over craft" reading is too pessimistic and this argument weakens.
If, by the end of 2027, no European supervisory authority has taken an enforcement action or issued formal guidance that turns on Article 4, then literacy was a reporting line rather than a policy instrument and this argument fails.
Past issues will appear here as they're published. Subscribe above to receive each new dispatch directly.