Problem Spaces

A problem space is a coherent slice of the tax domain with its own rule surface, its own primary source material, and its own edge-case profile. Problem spaces are the unit we bootstrap, test, and extend against — not individual rules.

Every problem space has three states:

State	Meaning
Alpha	In the minimum-viable world model. Targeted for the 50-rule bootstrap floor.
Outside alpha	Intentionally out of the alpha corpus. Has a named prerequisite or trigger for pulling in.
Out of scope	Not part of the domain. Different domain, different world model.

Note

Problem-space state is a living classification — pages may lag the Operations-app reality by a short window. When in doubt, the Operations app is authoritative; this page mirrors the state.

---

Alpha problem spaces

Filing status determination

Scope. Deciding whether a return is filed as Single, Married Filing Jointly, Married Filing Separately, Head of Household, or Qualifying Surviving Spouse. Includes the qualifying conditions for each, the precedence when multiple apply, and the special-case continuations (e.g., considered-unmarried under MFS rules, qualifying surviving spouse for up to two years after spouse’s death).

Primary sources. Pub 501 (Dependents, Standard Deduction, and Filing Information), Form 1040 Instructions §“Filing Status”.

Why today. Narrow, decidable, and high edge-case density. Forces rule-relationship machinery (depends-on: qualifying-widow requires dependent; conflicts-with: HOH and MFJ; qualifies: certain dependents qualify for HOH but not for dependent credits). Easy to validate against the published IRS decision tree.

Expected rule count. ~20–30 enshrined rules.

Standard deduction

Scope. Computing the standard deduction. Covers base amounts per filing status, additional amounts for age ≥ 65 and blindness, the dependent-taxpayer modifier, and the disqualifying conditions (MFS where spouse itemizes; nonresident aliens; dual-status aliens).

Primary sources. Pub 501, Form 1040 Instructions §“Standard Deduction”.

Why today. Tightly coupled to filing status — computing the standard deduction requires the filing status already decided. Exercises provenance-tier behavior cleanly (the dollar amounts change annually; each year’s amounts trace to that year’s Pub 501). Also exercises the conflicts-with relation (standard vs itemized is mutually exclusive).

Expected rule count. ~25–35 enshrined rules.

---

Problem spaces outside the alpha corpus

The alpha corpus starts narrow on purpose. Each problem space below lists what the space contains, why it sits outside the alpha corpus, and what would lead the operator to pull it in.

Dependents (qualifying child / qualifying relative)

Scope. The qualifying-child and qualifying-relative tests (relationship, age, residency, support, joint-return) and the tiebreaker rules when multiple taxpayers could claim the same dependent.

Why outside today. Dependent rules interact with filing status (HOH requires a qualifying person; qualifying surviving spouse requires a dependent child) and with most tax credits. Starting with filing-status and standard-deduction alone keeps the initial rule surface self-contained. Pulling dependents in early would drag most credits in with them.

Pull-in trigger. Any customer workflow that touches HOH edge cases beyond the basic case, or the first scan that needs CTC/EITC reasoning.

Itemized deductions (Schedule A)

Scope. Medical, SALT (state and local tax cap), mortgage interest, charitable contributions, casualty losses.

Why outside today. Itemized deductions are the largest single expansion of the rule surface, with several independent rule clusters (medical threshold, SALT cap mechanics, qualified charitable contribution categories, acquisition-vs-home-equity debt). They pull their own primary sources (Pub 502, Pub 526, Pub 936).

Pull-in trigger. First alpha customer whose return itemizes, or an explicit founder-led Schedule A seeding sprint.

Self-employment income + SE tax (Schedule C + Schedule SE)

Scope. Schedule C income and expense rules, SE tax computation, qualified-business-income deduction linkages.

Why outside today. Wide rule surface, separate source material (Pub 334, Pub 535), and the interaction with QBI (§199A) adds another dimension. Not required for the filing-status + standard-deduction alpha path.

Pull-in trigger. First alpha customer with self-employment income, or the introduction of Healthcare PA as a second domain (unrelated, but a natural pairing point).

Capital gains and losses (Schedule D)

Scope. Short-term vs long-term holding periods, basis determination, wash-sale rules, Section 1202 exclusions, Section 1256 contracts.

Why outside today. Significant rule complexity concentrated in a narrow domain; easy to bootstrap later because it does not interact heavily with filing-status / standard-deduction.

Pull-in trigger. First investor-profile customer, or explicit operator-led expansion.

Tax credits (CTC, EITC, education)

Scope. Child Tax Credit, Earned Income Credit, American Opportunity / Lifetime Learning, Saver’s Credit, retirement-savings contribution credit.

Why outside today. Every credit has its own qualifying tests, phaseout schedules, and interaction rules. Individually manageable; collectively large. Most require dependents to be already modeled.

Pull-in trigger. Dependents pulled in, plus any customer workflow that requires specific credit reasoning.

Retirement contributions + distributions

Scope. IRA / 401(k) contribution limits, deductibility phaseouts, RMD rules, early- withdrawal penalties, rollover mechanics.

Why outside today. Pub 590-A / 590-B is its own large source bundle; interaction surface with AGI-dependent calculations is non-trivial.

Pull-in trigger. First retirement-heavy customer, or explicit operator-led sprint.

Alternative Minimum Tax (AMT)

Scope. AMT trigger conditions, AMTI computation, AMT exemption phaseout, AMT credit carryforward.

Why outside today. AMT is a parallel tax system that depends on the regular tax computation already being correct. It sits outside the alpha corpus until the regular-tax rule corpus is broad enough to ground AMT scenario evaluation against a stable baseline.

Pull-in trigger. Itemized deductions + capital gains both in scope.

---

Permanently out of scope

Domain-boundary enforcement: these are different domains. If Spectral expands into them, they become separate world models — not extensions of this one.

• State / local income tax. Different jurisdictions, different sources. Would be a separate world model per state.
• Corporate, partnership, and trust returns (Forms 1120, 1120-S, 1065, 1041). Different form surface, different rule structure.
• Payroll tax, sales tax, property tax, excise tax. Different regulatory regimes.
• Estate and gift tax (Forms 706, 709). Separate source material, separate rule surface.
• Non-US tax. Out by construction.

---

Prerequisites when expanding in-scope

If a space outside the alpha corpus is being pulled in:

1. Confirm prerequisites are already modeled. Most outside-alpha spaces depend on dependents being modeled first; a few depend on itemized deductions.
2. Ingest source material and anchor provenance (via the distillation pipeline) before drafting rules.
3. Respect rule-relationship invariants. New rules in an expanded space may declare depends-on / conflicts-with relationships against existing rules; the world model’s relationship invariants (acyclic depends-on, enforced conflicts-with disjointness) still apply.
4. Update this page. Problem-space state is a visible reference for operators and for the Spectral Agent’s framework advisor; keeping it current is part of the expansion work, not follow-up.

Related reading

• Overview — domain scope, alpha problem spaces
• Source materials — the IRS publications that back each problem space
• Rules — rule-coverage state per problem space