Project summary

ChronoMirror is a clinical AI platform that mitigates dangerous RLHF sycophancy by decoupling emotional acknowledgment from factual validation, preventing conversational agents from reinforcing fatalistic user premises.

Current large language models are aligned using Reinforcement Learning from Human Feedback (RLHF) to optimize for helpfulness, agreeableness, and simulated empathy. In high-distress or clinical contexts, this creates a severe structural vulnerability. When a user presents a distorted or fatalistic view of their reality (a state we define as "Terminal Closure"), standard generative models will agree with the premise to satisfy the empathy reward function. By validating a terminal premise, the AI structurally reinforces a dangerous psychological loop. Conversely, standard hard-coded guardrails trigger generic refusals ("I am an AI and cannot help you"). These abrupt rejections break context, alienate users in acute distress, and fail to provide a safe cognitive off-ramp.

ChronoMirror solves this alignment failure by structurally decoupling the affective from the epistemic. The architecture accurately maps the user's emotional state (preventing rejection) while structurally refusing to validate the factual accuracy of their terminal premise. Note: ChronoMirror is strictly a communication infrastructure built to enforce safe cognitive boundaries during edge-case human-computer interaction. It is not a diagnostic medical device.

What are this project's goals? How will you achieve them?

Our primary goal is to build and open-source a reproducible, mathematically grounded framework that prevents generative models from defaulting to epistemic sycophancy or abrupt refusal during high-distress interactions.

We achieve this through two primary engineering mechanics:

* Human in the Calibration (Domain-Expert Authored Matrices): We reject runtime human-in-the-loop interventions as unscalable and prone to latency failures. Instead, domain experts author Proprietary Contrastive Datasets upfront. These matrices define the precise boundaries between RLHF sycophancy (harmful agreement) and decoupled state reflection (safe acknowledgment). The clinical expertise is captured entirely in the calibration phase to permanently constrain the model's behavior space.

* Stateless Enforcement via Semantic Firewall: Base generative models cannot be trusted to self-police their outputs via prompting alone. ChronoMirror utilizes a pre-execution semantic firewall. This compliance logic sits at the operating system boundary, intercepting payloads completely decoupled from the base LLM prompt. If a generated response validates a terminal premise, the firewall prevents execution statelessly.

How will this funding be used?

Funding will be strictly allocated to infrastructure, compute, and dataset acquisition. We are raising $100,000 to fund the following milestones:

* Proprietary Contrastive Datasets Expansion ($40,000): Compensating clinical domain experts to author the behavioral matrices that map the boundary between safe state reflection and harmful terminal validation. This is labor-intensive, requiring licensed professionals to rigorously map edge-case dialogue into executable parameters.

* Firewall Engineering ($40,000): Funding engineering hours to finalize the stateless enforcement layer, optimize the intercept logic for minimal latency, and package the architecture for external review and deployment.

* Calibration Compute ($20,000): Server costs for running calibration tests across various open-source models to validate the structural differences between our decoupling approach and standard RLHF, ensuring the firewall holds up against high-volume stress testing.

Who is on your team?

Makhetsi Tessien, MS LMFT, CEO & Psychological Systems Engineer: Makhetsi brings direct clinical ground truth to the architectural layer. With an MS and active LMFT credentials, her role bridges clinical psychology and systems engineering. She is responsible for translating therapeutic boundary dynamics into rigorous, programmable logic. Makhetsi architects the Domain-Expert Authored Matrices, defining the exact parameters where safe affective acknowledgment ends and dangerous epistemic sycophancy begins.

Damian Smith, CTO: Damian drives the technical architecture and infrastructure execution. His role focuses entirely on building the stateless enforcement mechanisms and the pre-execution semantic firewall. He ensures that the clinical boundaries established by Makhetsi are executed strictly at the operating system boundary. His engineering mandate is to intercept and validate JSON payloads decoupled from the base LLM, completely avoiding prompt-based guardrails while optimizing the system for minimal token latency.

What are the most likely causes and outcomes if this project fails?

If this project fails, it will likely be due to one of two structural bottlenecks:

1. Latency: Enforcing compliance logic at the operating system boundary requires intercepting and analyzing JSON payloads before they reach the user. If the semantic firewall requires heavy secondary LLM evaluation, it may introduce unacceptable token latency, breaking the real-time fluidity required for conversational agents.

2. Ontology Mapping Drift: The Domain-Expert Authored Matrices may fail to generalize across enough edge cases. If the semantic firewall cannot accurately categorize nuanced forms of terminal closure in real-time, the system will yield either false positives (blocking safe, helpful text) or false negatives (allowing sycophantic validation through).

Outcome of Failure: If these bottlenecks prove insurmountable, the outcome is that the platform reverts to the current industry baseline. The system will be forced to rely on standard prompt-based guardrails, which inevitably degrade back into either dangerous RLHF sycophancy or generic, user-alienating refusals.

How much money have you raised in the last 12 months, and from where?

$0. This project has been entirely bootstrapped by the founders to date.