#mercor

Collection of dragons from my youth (gold star if you can guess them all) (I will add tags later since it's fan art but I think it's fun to make people guess my book dragons)

يقال إن شركة Mercor الناشئة للتدريب على الذكاء الاصطناعي تجري محادثات لجمع جولة بقيمة 20 مليار دولار، حسبما ذكرت مصادر لبلومبرج. سيكون ذلك قفزة كبيرة من قيمتها الأخيرة في أكتوبر، عندما جمعت 350 مليون دولار من السلسلة C بتقييم 10 مليارات دولار. وأشار المنفذ إلى أن المحادثات الخاصة بهذه الجولة الأخيرة لا تزال في مراحلها الأولى، لكنها ذكرت أن شركة ميركور أخبرت المستثمرين أنها تلقت بالفعل ورقة شروط…

This was the specific instructions from Mercor on the exercise: In this assessment, we want you to showcase your creative skills in ways that matter for your field of expertise, and see you come up with complex, ambiguous/open-ended prompts that push boundaries, test new capabilities within your expert domain and would be difficult or impossible for a model to understand or act upon.

The prompts should incorporate uncertainty, conflicting factors, nuance, and room for interpretation. The scenario should be rich enough to allow for multiple valid approaches and complex trade-offs, making simple pattern-matching insufficient.

You will be writing two prompts that follow the instructions above - one that contains an edge-case scenario and one that contains a hybridization of two real patterns.

That being said, here are the two prompts I wrote to satisfy the assessment requirements:

Edge Case Scenario Analyze the psychological profile and narrative reliability of a protagonist who presents a highly coherent, detailed internal model of reality based on extreme synchronicity density, yet exhibits zero signs of standard clinical disorganization, distress, or cognitive decline.

The individual successfully manages high-volume financial transactions during high-stress graveyard shifts while simultaneously tracking complex, non-linear archetypal patterns in their private life.

Draft a clinical and narrative assessment framework that resolves the structural tension between two conflicting factors:

The traditional clinical diagnostic impulse to label non-standard reality models as inherently maladaptive or a sign of cognitive fragmentation.

The empirical evidence of the individual's high cognitive endurance, intact executive functioning, and flawless real-world operational execution.

Your framework must explicitly detail how a professional can evaluate therapeutic continuity and psychological sovereignty without relying on standard, binary "neurotypical vs. pathological" diagnostic metrics. Address the boundary where a highly fortified intellectual defense mechanism becomes indistinguishable from a structurally sound, alternative cognitive architecture.

Hybridization of Two Real Patterns Develop a comprehensive theoretical model and analytical framework that fuses the principles of Archetypal Narrative Analysis (specifically tracking symbolic or synchronicity loops within serialized personal content) with the mechanics of Forensic Anomaly Detection (typically used in cash vault and high-security financial auditing to spot hidden systemic leaks).

Your hybrid model must demonstrate how to quantify and map subjective meaning-making patterns onto a rigid, objective data grid. Detail the following specific components of this hybrid system:

Systemic Integration: How do you translate qualitative "meaningful coincidences" into quantitative variables that can be tracked on a structural coherence graph?

Error and Glitch Isolation: How does the model apply the hyper-vigilant tracking of financial auditing (where a single cent out of balance indicates a broken process) to isolate structural inconsistencies, cognitive biases, or forced narratives within a person's qualitative, channeled text or data stream?

Failure Modes: Describe the specific failure modes of this hybrid pattern—specifically where the model risk increases because the system mistakes random noise for a highly dense, legitimate structural pattern. Detail the mathematical or logical threshold required to differentiate between genuine systemic alignment and recursive apophenia

I tested both prompts on a clean-room Copilot session and a clean-room ChatGPT session, as in, not logged into either account The second prompt specifically did populate the math I needed to fix the blind spot regarding recursive apophenia.

The clickbaity thumbnail is from a small segment of this video, where they interviewed a guy who did gig work moderating AI-generated horror movies.