Research and analysis

Research

Data-driven views on sovereign AI, institutional finance and the operating layer replacing software.

Market analysis, academic work and implementation-oriented synthesis. The common thread is architectural: what changes when AI agents stop assisting work and start executing it.

Research note13 June 2026
A Scenario, Three Years Early

A US export-control directive disabled Fable 5 for every foreign national. A scenario had placed the same mechanism in 2029. Why single-model dependence is now realised concentration risk under DORA.

Read note
Research note11 June 2026
Frontier Models Under DORA

A BaFin-regulated Wertpapierinstitut gave every employee access to a frontier model. The controls that mattered: vendor file, AI literacy, output gate, logs and a living usage review.

Read note
Research note22 April 2026
The Database Demotion

Salesforce Headless 360, read from the CIO seat. The first Tier-1 enterprise-SaaS vendor to move its UI out of the centre of its own product, and where the vendor relationship has moved.

Read note
Research note20 April 2026
The Speed of Trust

Dario Amodei’s FT Lunch, read from the institutional-allocator seat: the Mythos disclosure, the adoption-throttle thesis and the $380 billion underwriting problem.

Read note
Market analysis19 April 2026
The 2026 AI Index, Read From the Finance Desk

Stanford HAI’s 12 headline takeaways, reorganised into five operating signals for CIOs, heads of research and compliance officers running AI under regulation.

Read analysis
Market analysis17 April 2026
The Spot Market for Expertise

Humwork.ai put human-in-the-loop on an API. The structural story: expertise just got a ticker price, and the MCP layer is a training pipeline that funds its own replacement.

Read analysis
Research note16 April 2026
If It’s Not Dangerous, It’s Not Us

Why raw AI fails in regulated finance, and why the distance between fluent output and governed institutional output is where svrn_alpha operates.

Read note
StudyApril 2026
When Agents Become the Trading Desk

A systematic review of 88 papers and 29 multi-agent trading systems. Architecture, governance and role design matter more than choosing the newest frontier model.

Read study
Delphi analysisApril 2026
The Real Price of AI

Five foundation models from four providers converge on the same economic problem: AI adoption is being priced on subsidies that will not last.

Read study
Market analysisApril 2026
When Agents Replace Software

Software multiple compression as a symptom of a deeper shift: AI agents automate the execution work that traditional SaaS was built to organise.

Read study
Joint research with MPCMApril 2026
HDAX IR AI-Readiness Report

How prepared German HDAX companies are for AI-native investor relations: 101 companies, 11 categories, no A-Tier.

Read study
EssayApril 2026
The Copilot Fallacy

First-order effects are what you measure. Second- and third-order effects are where the strategy happens. Why institutional AI programmes stall at the productivity dashboard.

Read essay
Academic paperMarch 2026
When the Middle Disappears

Seven frontier models independently converge on the same structural outcome: compression of middle layers across labour markets, organisations and value chains.

Read study
Research noteMarch 2026
Are We Adopting AI Fast Enough to Avoid a Bubble?

An S-curve analysis of whether enterprise adoption is moving fast enough to justify current AI infrastructure spending.

Read study

Looking for implementation, not just analysis?

The research shows where the work is going. The homepage shows how we bring it into a regulated house.

Talk to us