Continuous inquiry.
Measurable conviction.

Researching the AI decade.

Our current focus is US-listed businesses positioned to benefit from AI over the next five to ten years. Coverage extends across semiconductors, memory, networking, data centres, power and energy, software and the businesses adopting cheaper intelligence. We follow the economic consequences of AI beyond companies labelled as AI businesses.

A durable trend is the starting question. The investment case must establish paid demand, the cost of delivering it, funding requirements and cash flow per diluted share. Smaller and emerging beneficiaries sit alongside established platforms in the research universe. Crypto-related businesses face the same scrutiny: an announced pivot to AI is tested against facilities, customers, power and financing.

We seek substantial capital appreciation where the prospective return justifies the price and the risk. Base, upside and downside scenarios expose the assumptions. Sector enthusiasm and another investor’s holdings can suggest a research question; our own evidence and valuation determine the conclusion.

One research loop.

A desk with distinct responsibilities.

Specialisation gives each question an owner. Agents can exchange evidence directly; the reporting structure keeps assignments, risk authority and final decisions clear.

Research coordination

A chief of staff assigns bounded investigations, resolves dependencies and brings the desk’s findings into one decision brief. A portfolio lead provides deputy coverage.

Evidence and business research

Data stewardship establishes a shared source packet. Business analysis tests demand, customer concentration, competitive position, capital requirements and the link between AI adoption and company earnings.

Valuation and quantitative research

Separate specialists examine value per diluted share, market-implied expectations, scenario returns, volatility and common exposures. Calculations are preserved with their inputs.

Independent challenge

A dedicated challenger looks for conflicting evidence, weak assumptions and alternative explanations. The investment case must address the objection rather than count supporting opinions.

Portfolio and independent risk

Portfolio construction considers how opportunities fit together. Risk has a separate escalation line to the founder and can block proposals under the standing policy; coordination cannot waive that block.

Macro and sector context

A macro specialist examines financing conditions, energy, supply chains and sector developments that could change the company-level thesis.

Operations and performance

Distinct responsibilities cover order-record reconciliation, performance measurement and attribution. These connect investment decisions back to an inspectable record.

The founder retains investment decisions and order authority. The chief of staff condenses the research without erasing disagreements, missing evidence or the conditions that would change a recommendation.

A record that outlasts the conversation.

Research checkpoints link an assignment to its evidence, findings, uncertainties and next review. Shared memory allows another specialist to pick up a question, inspect the earlier work and challenge it. Source packets and calculations are versioned; corrections are appended with a link to the record they supersede.

Our workflow and role definitions are held separately from any one model platform. Cloud research records and a versioned code repository preserve the context needed to resume work with different tools. The institution’s knowledge should survive a change of model, interface or computer.

Research with a daily rhythm.

The desk’s operating cadence defines source triage, checkpoints at task completion and an end-of-day research manifest. The chief of staff tracks dependencies and carries unresolved questions into the next assignment. Reviews are tied to evidence and events as well as the calendar.

Weekly review brings together record reconciliation, forecasts and the lessons that should change the next cycle. Performance attribution, when supported by reconciled records, separates stock selection from market exposure, sizing and implementation.

Conviction has to survive the numbers.

Business quality and an attractive share price are separate judgments. Valuation work examines cash generation, dilution, debt, reinvestment and the expectations already embedded in the market. Quantitative research tests return assumptions alongside volatility, drawdown, liquidity and common exposures.

Historical experiments use explicit information boundaries and reproducible calculations. Forward observation provides a separate test. Broader market and Bitcoin comparisons use consistent periods and clearly identified series; a benchmark comparison does not by itself establish an independent edge.

Our long-term programme also investigates document changes, economic relationships and systematic signals. More elaborate models enter that programme when they improve on simpler baselines after their full costs.

Learning must change the next investigation.

Our learning protocol requires an observation, a root cause, a concrete change and an owner. A lesson is adopted when that change appears in a checkable rule, template, script or review process. Recording an error is the beginning; preventing its recurrence is the purpose.

Rejected ideas and corrected assumptions remain part of the record. Over time, these histories help the desk ask better questions, recognise recurring failure modes and direct research effort where new evidence can change a decision.

A wider institutional ambition.

The current desk is the foundation for a larger network of specialised research agents. We expand responsibilities where additional investigation earns its cost. Public securities are the immediate focus; the same source discipline informs our broader interest in private companies, where diligence, ownership terms and financing require different methods.