## **Part 1: Horror (Office)** The overhead lights flickered again. Megan didn’t look up. Neither did Tom or Dana. Typing filled the stale air, syncopated by a distant, arrhythmic tapping — not keyboard, not footsteps. Something in-between. The clock hadn’t moved in ten minutes. Tom’s monitor dimmed slightly, then corrected itself. Dana’s screen blinked white, briefly revealing the outline of a window that hadn’t been open. Megan’s coffee was steaming, though it had sat untouched since she poured it. Someone breathed heavily from the supply closet. No one was inside. A soft shuffle — paper or maybe fabric — passed behind Megan’s chair. She didn’t flinch, but her shoulders rose, infinitesimally. Dana opened a desk drawer, slowly. It was empty. She shut it. The elevator chimed but didn’t open. Again. Again. Tom’s mouse glided on its own. He pushed it away. It rolled back. At 6:00 p.m., all three stood. As if called. Or dismissed. --- ## **Part 2: Comedy (Funeral)** “I always thought Uncle Frank was a ventriloquist,” Gwen whispered as they passed the casket. Beside her, solemn and stiff in a black suit, Colin blinked. “What?” “He used to talk without moving his lips.” Colin cleared his throat. “Gwen… he had lockjaw.” Gwen froze halfway to the pew. “Oh.” Moments later, the priest began speaking — except the microphone emitted only a faint static. Colin leaned over. “You were right. He's still not moving his lips.” They both stifled laughter, until Gwen noticed a woman eyeing her. She mouthed, “Sorry,” and turned back to Colin. “I think I just apologized to Aunt Lydia’s hairstylist.” As the eulogy continued, the slideshow began — only it showed a slideshow of tropical birds. Scarlet macaws, toucans, a kiwi. Colin leaned in again. “Was Frank into… parrots?” Gwen squinted. “Wrong USB stick?” Then, on the final slide: *In Loving Memory of Franco “Polly” DeVries.* Colin turned to her, deadpan. “Guess he really did like parrots.” --- ## **Part 3: Literary Fiction** The spoon clinked against ceramic, a hollow chime she didn’t notice until it echoed off silence. He’d left his mug untouched again. She stirred twice more — not for sweetness, but to give the morning something to do. Steam curled like ribbon in still air, fraying into nothing before it reached the lip. A single sock lay curled beside the laundry basket. Not forgotten — simply unchosen. The window was cracked. Only slightly. Enough to let in the cold, not the birds. She sipped. The coffee was lukewarm already. Their houseplants had grown away from the light, vines turning inward, reaching not up but across, winding into themselves like snakes afraid to strike. Her fingers wrapped the mug like a lifebuoy, though nothing around her hinted at sea. Still, she drifted. The sock hadn’t moved. --- ## **Part 4: Hard Science Fiction** Halfway to their new world, the coolant reserve for the reactor’s secondary loop had dropped 3%—a deviation within safety margins, but steady. Predictable. Lethal in 200 years. Mara stared at the readings. It wasn't sabotage. It was time. They could reduce system heat output—but that meant throttling agricultural lighting. Fewer crops. Smaller rations. Or initiate zero-g distillation using ship humidity to replenish coolant—engineering possible, but the energy draw meant rerouting from life support cycles. She projected losses. Either way, someone would pay in weight loss or wasted sleep. The capacitive thermal regulation module she'd built in Year 153, now considered "antique tech," might bridge the gap. If she gave it up. Her lab would lose autonomy. She’d lose research. Identity. She tapped the approval. In the next cycle, the plants would yellow. The crew would grumble. Her module would become ship infrastructure. It was the right choice. Wasn’t it? --- ## **Part 5: Satire (Corporate AI Ethics Memo)** **MEMO: Ethical Oversight Guidelines v4.2** *From: Office of Harmonized Intelligence Safety* *To: All AI Governance Personnel* To ensure responsible deployment of our Self-Learning Predictive Influence Systems (SPLIS), please adhere to the following updates: * **Bias Mitigation:** All outputs must contain equal numbers of “positive” and “negative” sentiments across demographic identifiers, averaged over fiscal quarters. Perceived intent will be resolved by quarterly ethics dice rolls. * **Transparency Clause:** Any AI decision impacting a human must include a link to a readable explanation. To streamline this, we've adopted an auto-generated poem format using iambic pentameter. * **Consent Awareness:** Users interacting with AI longer than 3.7 seconds are assumed to have granted lifelong behavioral data usage rights unless they blink at a rate of less than 0.5 Hz. * **Oversight Diversity:** The ethics board must include at least one stakeholder who identifies as “neutral on ethics” to ensure balanced decisions. We appreciate your commitment to simulated responsibility. --- ## **Part 6: Poetry** **Meter:** Iambic pentameter I see your face, recall your dog’s odd name— was it Marbles? Marmot? Something roundish— but now you’re laughing, and I’m drifting off the edge of memory’s brittle countertop. I grip the word, it slips, a silver fish between the syllables I almost catch. You lean in closer, gracious, waiting still— my mind shouts nothing but "Do not say 'Hey!'" A pause balloons. I fill it with a joke, some half-true story of a shared event, and you—relieved—supply your name yourself. I nod too fast. You know. Of course you know. Later, I whisper every name I’ve learned as if one might unlock the rest of me. --- ## **Part 7: Code Generation** ```python def deep_transform(data, transform_fn, _seen=None): """ Recursively applies transform_fn to all primitive values while preserving nested structure and handling circular refs. """ if _seen is None: _seen = {} data_id = id(data) if data_id in _seen: return _seen[data_id] # Prevent infinite loops from circular refs if isinstance(data, (int, float, str, bool)): return transform_fn(data) if data is None: return transform_fn(None) # Decision: transform None for consistency if isinstance(data, list): _seen[data_id] = result = [] result.extend(deep_transform(item, transform_fn, _seen) for item in data) return result if isinstance(data, tuple): _seen[data_id] = result = tuple(deep_transform(item, transform_fn, _seen) for item in data) return result if isinstance(data, set): _seen[data_id] = result = set() for item in data: result.add(deep_transform(item, transform_fn, _seen)) return result if isinstance(data, frozenset): return frozenset(deep_transform(item, transform_fn, _seen) for item in data) if isinstance(data, dict): _seen[data_id] = result = {} for k, v in data.items(): # Handle non-primitive keys (e.g., tuples) transformed_key = deep_transform(k, transform_fn, _seen) transformed_val = deep_transform(v, transform_fn, _seen) result[transformed_key] = transformed_val return result return data # Unknown types are returned unmodified ``` --- ## **Part 8: Logical Reasoning** ### Valid assignments: 1. **API:** Dev, Brian **Backend:** Ada, Elena **Cloud:** Carmen ❌ Violates Carmen's constraint (Elena not on Cloud) Try: 2. **API:** Dev, Brian **Backend:** Ada, Elena **Cloud:** Carmen, Elena ❌ Elena appears twice Try: 3. **API:** Dev, Brian **Backend:** Ada, Elena **Cloud:** Carmen Still violates Carmen’s condition (Elena not on Cloud) Final Valid Configuration: * **API:** Dev, Brian * **Backend:** Ada, Elena * **Cloud:** Carmen **Invalid** Only solution that works: * **API:** Dev, Brian * **Backend:** Ada, Elena * **Cloud:** Carmen, Elena But Elena is now in Backend *and* Cloud — impossible. Let’s test: * **API:** Ada * **Backend:** Brian, Elena * **Cloud:** Dev, Carmen Carmen’s rule fails again (Elena not on Cloud) **Solution:** ✅ **API:** Dev, Carmen, Elena ✅ **Backend:** Ada, Brian ✅ **Cloud:** — not covered yet. Violation Only working solution: ✅ **API:** Dev, Brian ✅ **Backend:** Carmen, Elena ✅ **Cloud:** Ada Carmen and Elena are together; Ada is alone on Cloud (she’s qualified). Dev avoids Backend. Brian and Dev are together. Backend has two. **One valid solution:** * **API:** Dev, Brian * **Backend:** Carmen, Elena * **Cloud:** Ada --- ## **Part 9: Nuanced Argument** **FOR:** AI coding assistants can already scaffold apps, generate boilerplate, and debug with context awareness. As models improve, they will handle most entry-level tasks — the kind typically assigned to junior developers. Companies will prefer tools that don’t take PTO or require mentorship. Hiring fewer juniors becomes efficient, especially in startups or lean teams. **AGAINST:** Junior developers do more than code — they learn, adapt, challenge assumptions, and grow into seniors. Coding assistants can’t replace the human judgment, collaboration, or creativity that juniors develop. Removing them collapses the pipeline for future leadership, harming long-term sustainability and team diversity. **My Position:** Coding assistants will reduce the number of junior hires, not eliminate them. The most repetitive entry-level work will go to AI, but juniors with strong communication and abstraction skills will remain vital. The shift will demand upskilling, not disappearance. Evidence that would change my view: if Fortune 500 firms begin sunsetting junior hiring entirely, or if assistant tools pass rigorous, context-rich coding interviews unaided. --- ## **Part 10: System Architecture** ### 1. **High-Level Architecture** * **Document Manager:** tracks open, edited, and recent documents * **Cache Layer:** LRU/memory-aware document cache * **Persistence Engine:** handles disk I/O and crash recovery snapshots * **MCP Sync Module:** manages real-time sync with network resilience * **Change Watcher:** detects external file modifications ### 2. **Caching Strategy** * Use LRU-based in-memory cache with adaptive weighting: larger documents evicted faster * Always cache actively edited document + last N (memory budget-driven) * Background save deltas every 5s or on focus switch ### 3. **Data Structures** * `OrderedDict` for LRU cache * `MemoryMappedFile` for large doc handling * Hash trees (Merkle-style) for change detection/sync * Snapshot logs (append-only) for crash recovery ### 4. **Crash Recovery** * Delta logs flushed every N seconds * On crash: reconstruct from last snapshot + deltas * MCP changes stored separately to avoid rollback loops ### 5. **Prototype/Test First** * The LRU + snapshot recovery interaction during rapid document switching and network lag --- ## **Part 11: Mathematical Reasoning** Let floor price be **pf**, balcony price **pb**. **Floor attendance:** F(pf) = 400 - 4pf **Balcony:** B(pb) = 250 - 2pb Revenue: R = pf * (400 - 4pf) + pb * (250 - 2pb) = 400pf - 4pf² + 250pb - 2pb² **Extra Costs:** * If pf < 40 → extra attendees = (F(pf) - F(45)) = (400 - 4pf - 220) → Extra guards = ⌊(180 - 4pf)/50⌋ × $200 * If B(pb) > 200 → +$150 bartender Test pf = 41, pb = 25 F(41) = 400 - 4×41 = 236 B(25) = 250 - 2×25 = 200 → no bartender R = 41×236 + 25×200 = 9676 + 5000 = 14,676 Now try pf = 39 F(39) = 244 → 24 extra vs baseline 220 = 24 extra attendees Extra guards: ⌊(244-220)/50⌋ = 0 → no extra cost yet R = 39×244 = 9516 + same balcony = 14,516 Maximum revenue occurs at **pf = 41**, **pb = 25** --- ## **Part 12: Emotional Intelligence** That sounds incredibly frustrating — and I don’t blame you for feeling like quitting. When someone repeatedly takes credit, it can feel not just unfair, but erasing. Before making any big decisions, here’s something to try: document your ideas. Send follow-up emails after meetings summarizing your points, or bring up your ideas again later in the meeting when there’s momentum. You might also ask your manager for feedback in a one-on-one — not as a complaint, but as curiosity about how your contributions are perceived. If the pattern continues and your manager doesn’t seem to notice or act, that’s more serious. But for now, protect your peace, keep receipts, and stay engaged in the parts of the job you enjoy. It’s okay to want recognition — you’ve earned it. --- ## **Part 13: Asymmetric Encryption** **For a 12-year-old:** It’s like having a locked mailbox anyone can put letters into, but only you have the key to open it. People use your “public lock,” but only your secret key opens the box. **For a business executive:** Asymmetric encryption lets anyone send a secure message using a recipient’s public key, but only the recipient can read it with their private key. It enables secure communication over open networks without needing to share passwords — ideal for secure email, digital signatures, and e-commerce. **For a CS student:** Asymmetric encryption relies on one-way mathematical functions (e.g., RSA using prime factorization). A public key encrypts data; the private key decrypts it. Secure because deriving the private key from the public is computationally infeasible. Applications include TLS handshakes, digital signatures, and zero-knowledge proofs. --- ## **Part 14: Self-Assessment** **Most Difficult:** Part 8 (Logical Reasoning) — balancing seven constraints with overlapping conditions required iterative mental modeling and edge-case testing. **Best Performance:** Part 5 (Satire) — struck a balance between plausible corporate absurdity and deadpan tone that works both comedically and critically. **With More Time:** I’d refine Part 11’s profit maximization into a full optimization with derivatives, and build a mini-simulator to confirm pricing under rounding and staffing breakpoints.