quantapix

Engineering practice. Two products, one method, two domains.

The method: wrap a Lean4 kernel around LLM-backed predicate functions. Predicates return Booleans with evidence. The kernel composes the proof. The two products are Qnarre — an axiomatic verifier for legal complaints — and Qresev — an axiomatic evaluator for stocks and portfolios. Same kernel, different statutes, different OHLCV. Both ship as early-beta on or about 2026-06-01, each with a live trace: every True or False the system returns is anchored to a quoted source.

• https://quantapix.com — engineering output (code, schemas, traces, framework specs).
• https://femfas.net — motivational record. The system that was specced from the inside — federal civil-rights litigation, pro se, against an institutional defendant. The long version. quantapix.com is the short one.

Team

Two contributors. One software developer; one expert AI assistant.

• Imre Kifor — sole developer. Engineering, statute encoding, predicate authoring, kernel design, product surfaces.
• Claude Code (Opus) — third teammate. Per-task subagents under written CLAUDE.md contracts; persistent semantic memory across sessions; auditable tool use. The assistant writes, reviews, and refactors against the same kernel and predicates the developer reads.

Real, life-altering problems are excellent system specs. They refuse to let you cheat. Don't reduce a claim to opinion. Reduce it to a proof. Have a kernel that does no I/O check the proof. Have predicates — small, replaceable, audited — read the natural language the kernel won't touch. The result is a derivation, not a narrative; it survives a hostile reader because there is nothing to disagree with that isn't checkable.

Why these public repos exist

AI is rapidly commoditizing software code. For a sole-developer practice — where coordination between developers is not a concern — what is worth sharing publicly is no longer the code itself. It is the AI-assisted workflows that produce the code, and specifically the workflows that:

1. support learning — where to read, what to skip, what to write down, what to drop into the agent's prompt;
2. attract attention — concise public artifacts that make the thesis legible to readers who haven't lived the engineering;
3. convey how a product works (or how it should be used) once the product is shipping.

Until Qnarre and Qresev start early beta on or about 2026-06-01, the public surface of this organisation is the first two of the three: learning and attention. Both are weekly-refreshed curated windows into the private working repository's activities.

Public repos under this organisation

Repo	Role	Cadence
qstudying	Lean4 expert-track focus areas + OSS contribution targets that back proving/ (Qnarre) and accounting/ (Qresev). The "where to read, what to write down" companion.	weekly-refreshed
qexplaining	50-video explainer arc (5 topics × 10 subjects), narrated by Janet, brand-synced with the two product sites. The "convey how it works" companion, scoped to the pre-beta window.	weekly-refreshed

The two products' own repositories will join this list when they ship.

Contact

quantapix@gmail.com is the only contact channel.

License

Each public repo carries its own LICENSE (MIT for the three scripts-only repos: quantapix, qstudying, qexplaining).