I study how sustainability and technological change affect company economics, financing and long-term value. HHFinAi is my independent research portfolio: investment cases, source-linked financial models and analyst-controlled workflows.
Research focus: financial materiality · capital allocation · climate and nature · credit and sovereign finance · responsible AI · stewardship
Updated 21 October 2026. No installation is needed to read the findings. Each case distinguishes evidence, assumptions, analysis and unresolved questions. The latest research uses public primary sources and original calculations; background reports guide analytical thinking, not conclusions. What changed and what remains unresolved.
| Investment question | Finding or analytical contribution | Read the work | Status |
|---|---|---|---|
| AI power and storage: what must faster, flexible capacity earn? | Grid/gas delivery economics are extended with a battery dispatch/resilience hurdle. Backup reserve and augmentation costs prevent free flexibility assumptions. | AI infrastructure · New storage case and chart | 16–21 October 2026; public context plus illustrative economics; site underwriting NEEDS_DATA |
| Climate and staples: how does input pressure reach profit? | A three-issuer monthly model separates procurement, hedging, pricing and volume, and exposes a COGS comparability trap. | Memo, chart and model | 21 October 2026; disclosed cost structures; hypothetical exposure; no risk ranking |
| Water and waste: what is the operating signal? | Xylem order concentration, Veralto segment growth and WM pricing/volume are evaluated separately, with prospective hypotheses. | Public operating monitor | 21 October 2026; Q2 evidence; valuation NEEDS_DATA; Q3 outcomes pending |
| Brazil agriculture: who captures productivity, and what happens to land? | CONAB forecasts are decomposed into area/yield effects; price-cost sensitivity challenges profit capture and land-sparing assumptions. | Nature case and model | 21 October 2026; agency forecasts plus hypothetical farm economics; no habitat outcome claimed |
| Ageing: how much fiscal pressure is policy-dependent? | Three-country demographic/pension evidence and an original policy identity expose the importance of employment and benefit indexation. | Sovereign case and model | 21 October 2026; published projections plus policy experiments; no sovereign spread forecast |
| EIB EuGB 2037: is a credible label enough? | Cash flows, carry, curve risk and trading costs test the premium to conventional debt. | Investment memo · Evidence and model | Historical case; 16 October 2025 cutoff; WATCH; signed terms and matched executable quotes needed |
Further studies and portfolio applications: Heidelberg Brevik memo and forward valuation; Microsoft water-resilient infrastructure; responsible-AI engagement and milestones; integrated carbon/nature/stewardship portfolio. Each retains its original scope, evidence date and unresolved gates; proposed engagement is not a recorded issuer interaction.
Investment judgment: read the conclusion, the counterfactual and the strongest contrary case. Ask which evidence would change the decision.
Financial discipline: inspect the model inputs, units, cash-flow boundaries, sensitivities and break-even conditions. Reported facts, management expectations and analyst assumptions are not interchangeable.
Research execution: follow the sources and run the deterministic calculations. The evaluation record distinguishes arithmetic and metadata checks from the observed source-extraction comparison. Inspect the frozen questions, prompts, raw responses, labels and errors before interpreting the results. Human adjudication, productivity gains and investment alpha are not established by this comparison.
Prospective discipline: the 21 October forecast register freezes three narrow operating hypotheses, subjective probabilities and scoring rules before the selected Q3 releases. Outcomes are pending. It is separate from extraction evaluation and investment performance.
Host integration: the recorded filesystem-host session exercises real research-mode submissions, stale-revision rejection, downstream invalidation and export. The reproducibility helper replays archived artifacts; it does not generate new research or certify human approval.
Evidence → economic mechanism → financial model → investment judgment → stewardship or monitoring question.
The original HHFinAi work is the implementation: counterfactuals, financial bridges, evidence boundaries, review controls and research-to-engagement handoffs. Third-party frameworks are attributed rather than presented as original inventions. AI assistance is disclosed; publication is not a claim that every source or assumption has received independent human validation.
Research standards and AI-use boundaries · Research development and open questions
Financial materiality · Bond diligence · Stewardship · Climate research
The wider suite covers nature, transition plans, carbon accounting, sustainable mandates and social/impact finance. These are supporting workflows, not evidence of completed research in every domain. Browse the full capability catalogue.
AI supports research; people remain accountable for evidence, assumptions and decisions. Verify primary sources, disclose material AI assistance, protect confidential information and retain uncertainty. These workflows do not authorize autonomous trading or certify sustainability or regulatory compliance.
Read the responsible AI policy — practical standards for evidence, disclosure, privacy, human oversight and corrections.
Independent research and educational work, not investment advice, regulatory certification, an audited performance record or an autonomous trading service. The local engines organize and validate research; a human or separately authorized AI host retrieves evidence and performs substantive analysis. Cases and demonstrations carry their own dates and limitations. Material unknowns remain visible.