We needed better tax software. So we started building it.

The practice supplied the problem

PrepReturns grew out of Trio Tax. Its starting point was the ordinary complexity of professional preparation: client documents, financial statements, payroll, asset history, shareholder information, and all the handoffs between evidence and a completed return.

The frustrations were familiar categories of professional software work. Information already present in a source could require another round of entry. A correction could be difficult to explain later. Useful connections could depend on limited interfaces. Evaluating software could offer less freedom than actually working in it would require.

These concerns did not begin as a plan to build a broad software company. They began as questions about tools for the practice itself. If the work could be modeled more directly, perhaps preparation could become clearer from beginning to end.

From isolated improvements to a system

The early question was narrow: can we make this part of an S-corporation return better?

Each useful answer exposed a neighboring problem. A calculation engine needed established facts. Establishing facts required source review. Source review needed a way to preserve corrections. A professional workspace needed saved history and recoverability. Review needed workpapers and forms that referred to the same preparation.

Those relationships gradually changed the project’s scale. The value was not only in a better calculation or a cleaner input screen. It was in the continuity between them.

PrepReturns came to represent a complete path: source documents and financial data, reviewed evidence, professional decisions, deterministic tax calculations, workpapers, diagnostics, official forms, and a return package. Each stage retained obligations to the stages before and after it.

Modern tools made a serious attempt possible

Frontier AI-assisted development changed what a small practice could attempt. Models could help research, implement, construct tests, challenge assumptions, inspect interfaces, and iterate on the result. Coordinated roles made it possible to work on bounded problems while keeping integration and product authority visible.

That did not make the software a one-prompt project. The Alpha generations were successive engineering questions: can the tax core stand on its own, can preparation begin with source evidence, can the workspace preserve history, can guidance focus on unresolved decisions, and can the system produce a reviewable official-form package?

The questions became more demanding as capability grew. Owners needed distinct output. Corrected evidence needed a preserved original. Current inputs and historical packages needed clear separation. A PDF had to look right to a human, not merely contain plausible mapping data.

The development page tells that cumulative story through the actual product work. AI accelerated implementation and review; human professional judgment remained central to the project’s scope and acceptance.

S corporations made the idea concrete

S corporations provided a substantial first domain. Their preparation connects business books, adjustments, assets, owners, basis, allocations, supporting information, and corporate forms. An end-to-end product had to represent those connections rather than stop at a calculation.

The current public generation is Alpha 14: human testing and broader operating-business coverage. It builds on earlier work in source review, Guided Preparation, owner and basis support, official forms, saved packages, and professional review.

The synthetic examples show that progression. Juniper establishes the core workflow. Alder adds multiple-owner and basis complexity. Cedarline combines richer conventional business facts, including inventory, rental property, investments, vehicles, and amortization within supported boundaries.

This is a defined return universe, not a promise to prepare every S corporation. The product pages make the current capabilities and limitations visible.

The broader Trio connection follows the same direction: from the practice outward.

Trio Tax is the tax practice where the preparation problems originated. Trio Ledger is the sibling ledger product for very small businesses, focused on straightforward books and useful financial statements. PrepReturns is the professional preparation project.

They share practical concerns about understandable financial information and useful workflows. Their relationship does not imply an implemented integration. Any future connection must preserve the meaning and authority of the information it carries.

The next question is larger

The strategic question eventually changed. If a small tax practice can coordinate enough research, implementation, testing, and review to build this much, what does that mean for the future of professional software?

PrepReturns is exploring that question in working software. Form 1040 comes next, using a model designed for households and individual tax law while carrying forward the preparation principles learned here. Open source is the intended direction, with the project private today.

The opportunity is broader than one product’s feature list. Practitioners can contribute more directly to the tools that shape their work, and software builders can learn from a preparation process grounded in the practice. We welcome conversations about testing, architecture, integrations, and collaboration.