VenueForge

Research / paper trading

Help and plain-language guide

How to use this pitch: Use Previous and Next, or use Left and Right while the presentation or page body has focus. Home and End move to the first or last slide. On a short slide, Up, Down, Page Up, Page Down, and Space move between slides; on a tall slide those keys keep their normal page-scrolling behavior. Reading layout shows all ten slides and leaves scrolling keys native. Shortcuts are not captured from this help text, form controls, or other interactive controls. Full screen is optional. This presentation describes a proposed research product; it does not promise trading returns or live orders.

Ten topics in plain language

  1. Problem — Fast markets. Fragmented truth. One protocol can be implemented in software and hardware, but meaning can drift between those places. Research and paper trading only; never live orders.
  2. Solution — One specification. Deterministic evidence. FinancialSpec defines meaning once; CanonicalIR is its normalized internal form; reference behavior is checked against evidence. Target generation is planned later.
  3. Proposed business model — Start with proof. Proposed paid validation pilots and replay qualification engagements come before later target-qualification licensing. Design partners would help validate the work.
  4. Technology — Semantics before speed. The implemented reference book, features, and risk are checked for correctness. C++23, eBPF, P4, and SystemVerilog target qualification is planned; hardware performance is unproven.
  5. Proposed go-to-market — Win one protocol first. The proposed path is technical design partners, a synthetic replay pilot, authorized data, and paper research. It is a proof-led pilot, not a promise of trading returns.
  6. Competition — Compete on the evidence chain. Manual implementation, one-shot code generation, and benchmark-only tooling are workflow categories. VenueForge proposes connecting meaning, decisions, and provenance.
  7. Proposed team — Keep ownership and review separate. Architecture and protocol, C++ and accelerators, and independent security and release review are proposed roles, not an operating team.
  8. Proposed milestones — Proof gates. Reference correctness, cross-target equivalence, hardware qualification, paper staging, and a release audit are proposed gates. Forecasts follow validated pilots.
  9. Current status — Reference risk, reviewed and tested. The stated snapshot is 18 scoped tests, 2,189 differential decisions, and Daybreak PASS. Full verification cites 826 Python tests plus C++ suites; VF-022 is complete, production release is not ready, and live trading is disabled. Source revision: 902840d.
  10. Summary and ask — Qualified systems. The proposed ask is design partners, protocol expertise, and qualification hardware to prove target equivalence and a reproducible paper-only release.

Plain-language glossary

FinancialSpec
VenueForge’s financial protocol specification: the written rules for what a system means.
CanonicalIR / IR
A normalized internal representation passed between tools so they can be compared consistently.
C++23
A version of the C++ programming language; say “C plus plus twenty-three.”
eBPF
A constrained program format that can run in the Linux kernel; say “E-B-P-F.”
P4
A language for describing programmable packet-processing pipelines; say “P-four.”
SystemVerilog
A hardware-description and verification language.
Differential decision
One decision compared across implementations to look for disagreement.
Target equivalence
Evidence that a generated or optimized target behaves like the reference for tested inputs.
Daybreak
The independent review gate used here; “Daybreak PASS” means that review reported no blocking finding for its scope.
VF-022
An internal work-item identifier, not a market, performance, or production claim.
Provenance
The record of where an input or result came from and which version produced it.

Scope reminder: The browser workbench uses historical ECB reference observations for a bounded local simulation. There are no real orders, no live quotes, no trading returns claim, and no claim that this MVP is the FinancialSpec compiler or a qualified trading service.

Problem

Fast markets. Fragmented truth.

One protocol. Multiple implementations. Too many places for meaning to drift.

Research and paper trading — never live orders.

Spec Software Hardware

One meaning, three places to lose it.

01 / 10