CostEngine Decision Integrity Agent

What the company appears to be: A self-reported decision integrity agent for fleet replacement decisions, built as a hackathon project. The system uses a deterministic engine to govern economic calculations and GPT-5.6 as a constrained orchestration layer for user challenges and read-only audits.

What changed: The project was extended during a Build Week hackathon to add an evidence-governed challenge workflow that allows users to question comparator or evidence basis in fleet decisions, while maintaining deterministic control over all economic outcomes.

The single most important open question: Is there any evidence of real-world use, traction or revenue beyond the author's own write-up? The description states no such evidence exists.

The description states that CostEngine Decision Integrity Agent is a system for fleet replacement decisions. It implements two read-only integrity tools:

• audit_comparator_integrity — checks comparator coherence and same-baseline gates.
• audit_evidence_integrity — checks evidence completeness and public provenance labels.

It uses deterministic gates (I16 and I31) to block flawed comparisons, and a workflow that creates a proposed regression seal requiring expert approval. The system is described as being isolated to the web layer with a feature-flagged Python orchestration module implementing a constrained Responses API tool contract.

The system is said to be "deliberately malformed" in its historical fixture design to prevent confusion with current production output, and it fails closed on unexpected inputs or runtime failures.

Evidence: The author's own write-up.

Inference: This appears to be a constrained AI agent for auditing fleet decision-making workflows, not a general-purpose AI tool. It is built around deterministic systems that retain authority over economic calculations.

The description states the system was inspired by the author’s experience at the National Renewable Energy Laboratory and aims to productize physics-based vehicle energy models into an "evidence-governed system for fleet decisions that must be explainable, auditable, and defensible."

It positions itself as a tool for high-consequence economic decisions where correctness and auditability are paramount. The author claims it prevents the same defect from returning by creating durable regression evidence.

Evidence: Self-reported in the write-up.

Inference: The positioning is focused on safety-critical decision environments, particularly in fleet management, with an emphasis on explainability and defensibility over automation.

The description states that the system is for "fleet replacement decisions" and was inspired by work in vehicle energy modeling. It is described as being built to address a real failure mode involving electric bus comparisons against mismatched baselines.

Evidence: The author's own write-up.

Inference: The target customer appears to be organizations involved in fleet management, particularly those in public transit or heavy-duty transportation sectors, where decisions carry significant financial and regulatory weight.

Not evidenced. The description does not contain any information about pricing, monetization, or business model.

The system is built with:

• Python
• JavaScript
• HTML/CSS
• OpenAI GPT-5.6 (used in Codex)
• Analytics and API components
• Deterministic engine for calculations
• Read-only integrity tools
• Feature-flagged orchestration module
• Responses API tool contract

It uses deterministic gates I16 and I31 to block flawed comparisons, and a workflow that requires expert approval before regression seals are added to the permanent test suite.

Evidence: The author's own write-up.

Inference: The system is built with a strong emphasis on control and safety, using deterministic systems for core calculations while allowing GPT-5.6 to act as an orchestration layer only.

Not evidenced. There is no mention of customers, revenue, usage metrics, or any signs of product-market fit beyond the author’s own account.

Not evidenced. No information is provided about competitors or market positioning.

• The system is described as a hackathon extension with no live GPT-5.6 call in the demonstration.
• No evidence of real-world deployment, traction, or adoption.
• The author states that all numbers and gate outcomes come from a deterministic engine — this may limit scalability or flexibility.
• The system requires expert approval for regression seals, which could slow decision-making.

Evidence: Self-reported in the write-up.

1. What is the actual use case or problem you are solving beyond the hackathon?
2. Are there any real-world deployments or pilot programs with fleet managers?
3. How does this system integrate with existing fleet management platforms or procurement systems?
4. What is the long-term vision for the product beyond the current hackathon extension?
5. Is there a plan to monetize or scale this solution beyond the author’s own use case?