My 2026 work spans second brains and AI chiefs of staff, a live operating platform for UKOPRA, financial AI, migration tooling and research into how AI can reflect a leader’s priorities. Some projects are live, others are being developed or tested.
Second brains and personal agentsActive development
WatsonOS · CHIE · Foundation
An AI chief of staff with memory and context.
Three connected projects behind my second-brain specialism. WatsonOS provides the agent orchestration and accountability. CHIE is a desktop knowledge engine that connects experiences, sources and working context. Foundation keeps decision principles and shared knowledge separate from individual business contexts.
The work spans agent memory, knowledge graphs, source references and explicit human confirmation. It is an evolving system, with implemented components and further integration in progress.
AI research and engineeringResearch prototype
Third Layer
Explore how AI can reflect a leader’s priorities.
A research and engineering project exploring a values-modulator architecture around an existing language model: learning from structured feedback, retrieving relevant institutional context and deciding what belongs in the model’s working context.
The work includes retrieval and context-assembly components with audit and evaluation infrastructure. Whether the approach measurably improves alignment remains a research question.
A complete operating platformIn production
UKOPRA
The systems around a national racing organisation.
A competitor portal, public website and administration platform for the United Kingdom Offshore Powerboat Racing Association. The work brings boat registrations, racing licences, event entries, payments and results into one operating environment.
It also covers officials, safety-related checks, audit records and on-site check-in. A practical example of connecting the people, rules and workflows around an organisation’s real work.
Institutional memory and operating designClient engagement
European venture capital firm
Keep knowledge connected as investment work develops.
Work on an agent operating environment, institutional memory and the architecture for connecting knowledge and delivery across a European VC firm.
The engagement includes shared agent working practices and requirements for future portfolio and investment tooling. Portfolio-product work is at the requirements and design stage.
Financial AI, full-stack deliveryArchitecture & implementation
Tobin AI
From private model to a working interface.
As the first AI engineer, I designed a hybrid architecture combining a private language model with deterministic logic and built the full-stack implementation, including the user interface.
The engagement ran from March to August 2026. Work included developing and testing Tobin 1.0 around UK corporate financial governance documents, financial models and CRM sales data.
Accounting practice migrationPilot toolkit
IRISv2
Make a complex move easier to manage.
An AI-assisted toolkit for moving UK accounting practices from IRIS Practice Management to modern platforms. It brings data discovery, classification, mapping, reconciliation and review into a managed migration workflow.
The build includes the migration engine, operator interface, approval gates, audit views and a local pilot setup. The pilot brings migration steps and review decisions into one place.
Consumer product, native iOSProduct development
GrowthJar
Make patience something a child can see.
A native iOS app that helps children learn about earning, saving and growth through a visual pocket-money system, with a separate parent experience for jobs, approvals and records.
The work connects product design, child-friendly interactions and a practical parent workflow. Growth is modelled within the family’s pocket-money system; it is not an investment platform.
Human control and data recoveryIn development
airgap
Keep recovery in human hands.
A backup and recovery product being developed around a clear boundary: backups of business tools should remain separate from the credentials available to AI agents, with restoration controlled by a person.
The target sources include Google Workspace, Notion and Airtable. Current development focuses on access boundaries, recovery workflows and the operator experience.