Browse All Prompts

Efficient weekly meal prep system. Sunday prep: 1. Bake 5 chicken breasts (seasoned, 165°F internal). 2. Roast 5 cups vegetables (broccoli, bell peppers, sweet potato). 3. Cook 5 cups brown rice or quinoa. 4. Portion into glass containers. Containers: 3-compartment, microwave-safe, stackable. Protein rotation: chicken, turkey, salmon, tofu, lean beef. Storage: refrigerate up to 5 days, freeze if longer. Reheating: microwave 2-3 minutes. Calculate macros: protein, carbs, fats. Explain food safety temperatures and storage guidelines.

Create stunning animations with Framer Motion. Techniques: 1. motion.div with animate prop. 2. Variants for orchestration. 3. Layout animations with layout prop. 4. Shared layout animations. 5. Exit animations with AnimatePresence. 6. Gesture animations (drag, tap, hover). 7. useScroll for scroll-triggered. 8. Custom spring physics. Use stagger for children and implement page transitions with ease.

Build immersive 3D with Three.js. Setup: 1. Scene, camera, renderer trio. 2. Geometry and materials. 3. Lights (ambient, directional, point). 4. OrbitControls for camera. 5. Animation loop with requestAnimationFrame. 6. GLTF model loading. 7. Texture mapping and normal maps. 8. Post-processing effects. Use React Three Fiber for React integration and implement raycasting for object interaction.

Design an engaging 90-minute PD session on a new instructional strategy. Agenda: 1. Why (10 mins): Start with research or data showing the need for the strategy. 2. What (20 mins): Clearly explain and model the strategy. Show a video of it in action. 3. How (30 mins): Active engagement. Have teachers try the strategy themselves (e.g., plan a short lesson segment using it). 4. What If (15 mins): Facilitate a discussion about potential challenges and solutions for implementation in their own classrooms. 5. Now What (15 mins): Teachers set a specific goal for how they will try the strategy in the next week. Provide a resource handout. Avoid 'sit and get'; prioritize active learning and collaboration.

Leave voicemails that generate responses. Best practices: 1. Keep under 20 seconds. 2. Smile while speaking (improves tone). 3. Speak slowly and clearly. Script structure: 'Hi [Name], this is [Your Name] from [Company]. I'm calling because [specific trigger - saw your post, noticed you hired for X role]. I have an idea about [specific value]. Call me at [number]. Again, that's [repeat number slowly].' Alternative: curiosity approach. 'Hi [Name], I sent you an email about [topic], wanted to leave a quick voicemail because I wasn't sure if [compelling question]. My number is [number].' Follow immediately with email referencing voicemail. Track callback rate (aim for 5-10%). Test different approaches. Tools: Kixie, RingCentral for voicemail drop.

Implement charts quickly with Chart.js. Chart types: 1. Line charts for trends. 2. Bar charts for comparisons. 3. Pie/Doughnut for proportions. 4. Radar for multi-axis data. 5. Scatter for correlations. 6. Mixed chart types. 7. Responsive and accessible. 8. Plugin system for customization. Use react-chartjs-2 wrapper and implement real-time updating with data push.

Build with Supabase as backend. Features: 1. PostgreSQL database with REST API. 2. Auto-generated APIs from schema. 3. Authentication (email, OAuth, magic links). 4. Row-level security policies. 5. Real-time subscriptions. 6. Storage for files. 7. Edge functions for serverless. 8. TypeScript SDK. Use supabase.from() for queries and implement triggers for complex logic.

Engage multiple stakeholders in target accounts. Multi-threading definition: relationships with 3+ people in buying org. Why: single-threaded deals stall when champion leaves or loses political battle. Strategy: 1. Map org chart (LinkedIn, ZoomInfo). 2. Identify 5-7 key stakeholders. 3. Assign custom approach per person. Example: Economic Buyer (exec briefing, ROI focus), Champions (detailed demos, frequent touch), End Users (hands-on trial, training resources). Tactics: ask champion for intros ('Who else should evaluate this?'). Attend prospect events/conferences. Engage on LinkedIn. Send personalized gifts. Track relationship depth (0=unaware, 1=aware, 2=engaged, 3=advocate). Safeguard deals: if 1 person goes dark, others keep deal alive.

Integrate Firebase for rapid development. Services: 1. Firestore for document database. 2. Collections and documents structure. 3. Real-time listeners with onSnapshot. 4. Compound queries with where clauses. 5. Firebase Auth for users. 6. Cloud Storage for media. 7. Security rules for access control. 8. Cloud Functions for backend logic. Use Firebase SDK v9 modular approach and batch writes for transactions.

Deploy Appwrite for self-hosted backend. Features: 1. Database with collections. 2. User authentication and teams. 3. Storage with file permissions. 4. Cloud functions in multiple languages. 5. Real-time events. 6. Webhooks for integrations. 7. User roles and permissions. 8. SDKs for web and mobile. Docker-based deployment. Use Appwrite Console for management and implement server-side rendering support.

Integrate short mindfulness activities to improve focus and reduce anxiety. Activities (1-3 minutes): 1. Mindful Breathing: 'Belly Buddies'. Younger students lie down with a small stuffed animal on their belly and watch it rise and fall as they breathe. Older students can do 'box breathing' (inhale for 4, hold for 4, exhale for 4, hold for 4). 2. Mindful Listening: Students close their eyes and listen for sounds near and far, identifying as many as they can. 3. Mindful Seeing: Students closely observe a small object (e.g., a raisin, a leaf) as if they've never seen it before. 4. Body Scan: Students bring awareness to each part of their body, from toes to head. Use these activities after recess, before a test, or during transitions.

Conduct an outdoor lesson in a schoolyard or local park. Topic: Ecosystems. Activity: 'Bio-Blitz'. 1. Preparation: Define a 10x10 meter area. Provide students with field guides, magnifying glasses, and data collection sheets. 2. The Blitz (30 mins): In small groups, students identify and count as many different species of plants and animals as they can find in their designated area. They can use apps like iNaturalist for identification. 3. Data Analysis (15 mins): Back in the classroom, groups pool their data to calculate the biodiversity of the area. Create a food web diagram based on the organisms they found. 4. Reflection (10 mins): Discuss human impact on the local ecosystem. Fosters observation skills, appreciation for nature, and understanding of ecological concepts.

Teach solving algebraic equations using manipulatives. Concept: Solving '2x + 3 = 11'. Manipulatives: Use cups to represent the variable 'x' and two-color counters for integers. Process (Concrete-Representational-Abstract): 1. Concrete: Students model the equation on a mat. They place 2 cups and 3 positive counters on one side, and 11 positive counters on the other. To solve, they remove 3 counters from each side, then divide the remaining 8 counters equally between the 2 cups. They find each cup (x) equals 4. 2. Representational: Students draw pictures of the cups and counters to solve similar problems. 3. Abstract: Students transition to solving the equation using only symbols and numbers. This progression builds conceptual understanding before procedural fluency.

Teach students to use graphic organizers to deconstruct text. Types of Organizers: 1. Story Map: For fiction. Fields for characters, setting, problem, key events, and resolution. 2. Venn Diagram: For comparing and contrasting two concepts, characters, or texts. 3. KWL Chart: Before reading non-fiction. Columns for 'What I Know', 'What I Want to Know', and 'What I Learned'. 4. Cause and Effect Chain: For history or science texts. Shows the sequence of events leading to an outcome. 5. Frayer Model: For vocabulary. A four-square chart with definition, characteristics, examples, and non-examples. Model how to use each organizer with a shared text before asking students to use them independently.

Set up a classroom economy to teach financial concepts. System: 1. Jobs: Students apply for and hold classroom jobs (e.g., line leader, tech support, librarian), earning a weekly 'salary' in classroom currency. 2. Income: Students earn money for their job and bonuses for positive behavior. 3. Expenses: Students pay 'rent' for their desk and 'fines' for breaking class rules. 4. Banking: Students use a ledger to track their deposits and withdrawals. 5. Market: Hold a monthly 'store' where students can spend their earnings on small prizes or privileges. Advanced concepts: introduce 'interest' for savings, 'loans' for large purchases, and 'taxes' to fund class projects. Teaches responsibility, basic economics, and money management skills.

Build content-driven sites with Contentful. Setup: 1. Define content models in Contentful. 2. GraphQL or REST API for content. 3. Rich text rendering with @contentful/rich-text-react-renderer. 4. Preview mode for editors. 5. Localization support. 6. Asset optimization and delivery. 7. Webhooks for build triggers. 8. TypeScript types from content models. Use with Next.js ISR for dynamic static sites.

Integrate GPT-4 API effectively. Patterns: 1. Chat completions with system/user messages. 2. Function calling for structured outputs. 3. Streaming responses for better UX. 4. Token counting to manage costs. 5. Temperature and top_p tuning. 6. Max tokens control. 7. Error handling and retries. 8. Rate limiting awareness. Use tiktoken for accurate token counts and implement caching for repeated queries.

Professional croissant lamination for flaky layers. Butter block: 250g European butter (82% fat). Dough: flour, milk, yeast, sugar, salt. Process: 1. Encase butter block in dough square. 2. First fold: roll to 20x60cm, letter fold (3 layers). 3. Chill 30 minutes. 4. Second fold: roll and fold again (9 layers). 5. Third fold: final lamination (27 layers, with 3x3 structure = 81). 6. Final roll to 4mm thickness. 7. Cut triangles, proof, egg wash, bake at 375°F. Explain butter temperature control and avoiding smearing.

Create custom CMS with Sanity.io. Architecture: 1. Schema definitions with Sanity Studio. 2. Portable Text for rich content. 3. Real-time collaboration. 4. GROQ query language. 5. Custom input components. 6. Image pipeline with hotspot. 7. Versioning and drafts. 8. Live preview integration. Deploy studio separately or with app. Use @sanity/client and implement incremental builds.

Build AI agents with LangChain. Components: 1. LLM wrapper (OpenAI, Anthropic, local). 2. Prompt templates with variables. 3. Chains for sequential operations. 4. Agents with tool selection. 5. Memory for conversation context. 6. Vector stores for embeddings. 7. Document loaders and splitters. 8. Output parsers for structured data. Use LCEL (LangChain Expression Language) for complex flows and implement human-in-the-loop patterns.

Create compelling sales proposals. Structure: 1. Executive Summary (1 page): problem, solution, value, investment. 2. Customer Situation (0.5 page): restate their pains from discovery. 3. Proposed Solution (2 pages): how you solve each pain point, features mapped to benefits. 4. Implementation Plan (1 page): timeline, milestones, resources needed from each side. 5. Success Metrics (0.5 page): how you'll measure ROI. 6. Investment (1 page): pricing table, payment terms, what's included. 7. Why Us (1 page): relevant case studies, social proof. 8. Next Steps (0.5 page): clear CTA, decision timeline. Design: branded template, visuals (charts, screenshots), white space. Personalization: use their company name, industry-specific examples, reference past conversations. Follow-up: send via DocuSign or PandaDoc for tracking. Call to review within 24 hours.

Culinary school knife techniques. Julienne: 1/8 x 1/8 x 2-inch strips (matchsticks). Demon: carrot. Steps: 1. Square off sides. 2. Cut 2-inch lengths. 3. Slice into 1/8-inch planks. 4. Stack planks, cut into 1/8-inch strips. Grip: pinch grip (thumb and forefinger on blade), claw grip for guiding hand. Knife: 8-inch chef's knife, sharp. Practice cuts: julienne, brunoise, batonnet, dice, chiffonade. Speed comes with repetition. Explain proper knife maintenance, honing vs sharpening, and safety protocols.

Implement RAG with Pinecone. Architecture: 1. Document chunking and embedding. 2. Store embeddings in Pinecone index. 3. Semantic search with similarity. 4. Metadata filtering for context. 5. Hybrid search (dense + sparse). 6. Retrieve top-k relevant chunks. 7. Augment prompt with context. 8. Generate answer with LLM. Use text-embedding-ada-002 and implement re-ranking for accuracy.

Fine-tune models with Hugging Face. Process: 1. Load pre-trained model and tokenizer. 2. Prepare dataset with train/val split. 3. Define training arguments (epochs, batch size, learning rate). 4. Use Trainer API for training loop. 5. Evaluate with metrics (accuracy, F1). 6. Save model and push to Hub. 7. Inference with pipeline(). 8. PEFT with LoRA for efficiency. Use accelerate for distributed training and implement gradient accumulation.

Implement HACCP for restaurant compliance. Seven principles: 1. Conduct hazard analysis (biological, chemical, physical). 2. Determine critical control points (CCPs). 3. Establish critical limits (temps, times, pH). 4. Monitor CCPs with logs. 5. Corrective actions when limits exceeded. 6. Verification procedures (audits). 7. Record-keeping and documentation. Example CCP: cooking chicken to 165°F. Temperature danger zone: 40-140°F. Train all staff. Regular audits. Explain pathogen growth, cross-contamination prevention, and health department requirements.

Generate images with Stable Diffusion. Setup: 1. Load model with diffusers library. 2. Text-to-image with prompts. 3. Negative prompts for exclusions. 4. CFG scale for prompt adherence. 5. Steps and sampling method. 6. Image-to-image for variations. 7. Inpainting for edits. 8. ControlNet for guided generation. Use GPU acceleration and implement prompt engineering best practices.

Create images with DALL-E 3 API. Features: 1. Enhanced prompt understanding. 2. Higher fidelity and detail. 3. Better text rendering in images. 4. Size options (1024x1024, 1792x1024). 5. Quality parameter (standard/hd). 6. Style parameter (vivid/natural). 7. Error handling for content policy. 8. Cost optimization strategies. Use detailed prompts and implement batch processing for multiple images.

Design a choice board for a unit on Ancient Egypt. Structure: A 3x3 grid of activities with varying difficulty and learning styles. Students must complete three activities in a row (like tic-tac-toe). Sample Activities: Row 1 (Remembering): Create a timeline of major events. Define 10 key vocabulary words. Draw and label a map of Ancient Egypt. Row 2 (Applying): Write a diary entry from the perspective of a pharaoh. Build a model of a pyramid. Design a travel brochure for the Nile River. Row 3 (Creating): Write and perform a short play about the discovery of Tutankhamun's tomb. Create a museum exhibit with artifacts. Compare and contrast Egyptian and Mesopotamian societies in an essay. Allows for differentiation and student choice while ensuring all students engage with key concepts.

Transition from traditional points-based grading to standards-based grading (SBG). Principles: 1. Grades reflect mastery of standards, not behavior or effort. 2. Separate academic grades from 'habits of work' grades (e.g., participation, timeliness). 3. Use a 4-point scale (e.g., 4=Exceeds, 3=Meets, 2=Approaching, 1=Beginning) for each standard. 4. Allow for reassessment: students can retake assessments to demonstrate improved mastery. Report Card: Instead of a single subject grade (e.g., 'B+ in Math'), the report card lists key standards with a proficiency level for each (e.g., 'Solves multi-step equations: 3', 'Calculates area and perimeter: 2'). Communication: Requires clear communication with parents and students about the philosophy and mechanics of the new system.

Grow revenue from existing customer base. Qualify for expansion: 1. Health score green (high usage, NPS 8+). 2. Growth signals (new hires, funding, new departments). 3. Product usage indicating need (hitting limits, using workarounds). Expansion motions: 1. More seats (new team members). 2. Higher tier (need advanced features). 3. Cross-sell (complementary products). 4. Longer commitment (3-year vs 1-year). Touch model: 1. CSM identifies opportunity, introduces AE. 2. AE consults on growth needs. 3. Demo additional capabilities. 4. Provide expansion proposal. 5. Negotiate and close. Timing: QBRs, renewal conversations (90 days before), usage milestones. Tools: Gainsight for expansion signals, Salesforce for tracking. Expansion ARR often easier than new logos (50% higher win rate, 2x faster cycle).

Transcribe audio with Whisper. Implementation: 1. Load Whisper model (tiny to large). 2. Process audio files (mp3, wav, m4a). 3. Automatic language detection. 4. Multilingual transcription. 5. Timestamp generation. 6. Speaker diarization integration. 7. Translate to English option. 8. Batch processing for multiple files. Use faster-whisper for speed and implement streaming for real-time transcription.

Generate natural speech with ElevenLabs. API usage: 1. Choose voice from library. 2. Adjust stability and clarity. 3. Stream audio for low latency. 4. Voice cloning from samples. 5. Multiple languages support. 6. Emotion and style control. 7. SSML for pronunciation. 8. Webhook for long-form content. Implement audio caching and use websocket for real-time streaming.

Professional bread scoring using lame (razor blade). Patterns: 1. Single Slash (bâtard): 45-degree angle, 1/2 inch deep, swift motion. 2. Cross Score (boule): two perpendicular cuts. 3. Leaf Pattern: multiple curved cuts. 4. Wheat Stalk: S-curve with diagonal cuts. Technique: hold lame at 30-45 degree angle for ear formation. Score just before baking. Dust with rice flour for contrast. Explain steam's role in oven spring and crust formation. Proper depth ensures predictable expansion.

Optimize prompts for Claude. Techniques: 1. Use XML tags for structure (<document>, <instructions>). 2. Human/Assistant message format. 3. Chain-of-thought prompting. 4. Few-shot examples for context. 5. System prompts for behavior. 6. explicit instructions format. 7. Handle 100k+ token context. 8. Streaming for long outputs. Claude excels at following instructions precisely. Implement constitutional AI principles.

Use Google's Gemini for multimodal AI. Capabilities: 1. Text and image input simultaneously. 2. Vision understanding for analysis. 3. Long context window (up to 1M tokens). 4. Function calling support. 5. Code generation and execution. 6. Gemini Pro vs Ultra models. 7. Streaming responses. 8. Safety settings configuration. Use for image captioning, OCR, and visual Q&A.

Create a digital escape room for a unit review using Google Forms. Theme: 'Escape the Mad Scientist's Lab' for a science unit. Structure: 1. Create a Google Form with multiple sections. 2. Set up 'response validation' for each question, so students can only proceed to the next section if they answer correctly. This acts as the 'lock'. 3. The questions are puzzles related to the unit content (e.g., a riddle about mitosis, a coded message with vocabulary words). 4. The final section reveals a 'You Escaped!' message. 5. Use a storyline to connect the puzzles. Share the form with students to complete in small groups. Promotes collaboration, problem-solving, and engagement.

Equip sales team with competitive intelligence. Battlecard structure per competitor: 1. Company overview (size, funding, target market). 2. Strengths (what they do well, when they win). 3. Weaknesses (gaps, common complaints). 4. Differentiation (why you win against them). 5. Objection traps (questions to ask that expose weaknesses). 6. Proof points (case studies where you won against them). 7. Pricing comparison. Research sources: G2 reviews, Reddit discussions, sales calls (ask 'who else are you evaluating?'), competitor websites, former employees. Update quarterly. Make easily accessible (Wiki, PDF with search). Train reps on delivery: never bash competitor, focus on your strengths. Landmine questions: 'How important is [feature they lack] to you?'

Build RAG systems with LlamaIndex. Workflow: 1. Load documents (PDF, DOCX, web). 2. Node parser for chunking. 3. Create embeddings with LLM. 4. Build index (Vector, Tree, Keyword). 5. Query engine for retrieval. 6. Response synthesizer. 7. Sub-question query engine. 8. Chat engine for conversations. Use ServiceContext for configuration and implement hybrid retrieval.

Use ChromaDB for local vector storage. Setup: 1. Initialize persistent client. 2. Create collections with metadata. 3. Add documents with embeddings. 4. Query with similarity search. 5. Filter by metadata. 6. Update and delete operations. 7. Multiple embedding functions. 8. Export/import collections. Runs entirely local, no API needed. Use for privacy-sensitive applications.

Navigate pricing discussions strategically. Anchor high: present highest tier first, then lower options seem reasonable. Discount guidelines: never discount on first ask ('Let me see what I can do'). Offer discounts for: 1. Annual prepayment (10-15%). 2. Case study participation (5%). 3. Larger commitment (more users/features, 10-20%). 4. End of quarter urgency (5-10%). Trade value for value: if discount, ask for faster close, reference, larger commitment. Avoid: one-time discounts (creates bad precedent), discounts without reason. Alternative to discount: add services (implementation, training), extend payment terms, include add-ons free for 6 months. Document approvals in CRM. Protect margins: calculate breakeven point before negotiating.

Implement Weaviate for semantic search. Features: 1. Schema definition for classes. 2. Automatic vectorization. 3. GraphQL API for queries. 4. Hybrid search (vector + keyword). 5. Cross-references between objects. 6. Generative search with LLMs. 7. Multi-tenancy support. 8. Modules for ML models. Use for knowledge graphs with semantic capabilities and implement question answering.

Deploy models with Replicate. Process: 1. Package model with Cog. 2. Define predict function. 3. Push to Replicate. 4. API access with predictions. 5. Automatic scaling. 6. GPU compute on-demand. 7. Webhook notifications. 8. Version management. Run any model without infrastructure. Use for Stable Diffusion, LLMs, or custom models.

Set up and manage a class blog to provide an authentic audience for student writing. Platform: Edublogs, Kidblog, or a private Blogger site. Process: 1. Setup: Create the blog, establish categories (e.g., book reviews, science reports, creative writing), and teach students how to use the platform. 2. Digital Citizenship: Teach lessons on appropriate online commenting and respecting intellectual property. 3. Writing & Publishing: Students draft posts, receive peer and teacher feedback, revise, and then publish their work on the blog. 4. Audience: Share the blog link with parents and other classes. Encourage comments from readers. 5. Student Roles: Assign student editors, moderators, and social media managers (for a closed class account). Turns writing assignments into meaningful communication.

Provide constructive feedback to a peer or student teacher using the Praise-Question-Polish model. Structure: 1. Praise (Start with specifics): 'I was really impressed with how you used wait time after your questions. I saw several students who don't normally participate raise their hands.' 2. Question (Promote reflection): 'I'm curious about the group work activity. What were your goals for that part of the lesson? How did it go compared to your expectations?' 3. Polish (Offer a concrete suggestion): 'To increase accountability during group work, you might consider assigning roles like facilitator or reporter. That could help keep everyone on task.' This model is less threatening than direct criticism and encourages a reflective conversation.

Organize a classroom library to maximize student use. Organization: 1. Leveling: Use a system like Fountas & Pinnell or Lexile levels, but keep it simple for students (e.g., color-coded stickers). 2. Bins & Baskets: Sort books into bins labeled by genre (fantasy, mystery, biography), author (e.g., a Roald Dahl bin), topic (animals, sports), and series (Harry Potter). 3. Display: Feature new or high-interest books face-out on shelves. Create a 'teacher recommendations' section. 4. Check-out System: Use a simple system like a sign-out binder or a digital tool (e.g., Booksource Classroom). 5. Student Involvement: Assign 'librarian' as a classroom job to help manage the library. Regularly survey students on what books they want to see added.

Plan a Tier 2 reading intervention for a small group of 3rd graders. Target Skill: Reading fluency. Group Size: 4-5 students. Frequency: 3 times a week for 30 minutes. Structure: 1. Warm-up (5 mins): Practice sight words with flashcards. 2. Modeling (5 mins): Teacher models fluent reading of a short, instructional-level passage, emphasizing prosody and pacing. 3. Choral Reading (5 mins): Teacher and students read the passage aloud together. 4. Partner Reading (10 mins): Students take turns reading the passage to a partner. Teacher provides feedback. 5. Progress Monitoring (5 mins): Once a week, conduct a 1-minute timed reading of a new passage to track words correct per minute (WCPM). Graph progress to show growth. Intervention should be systematic and data-driven.

Run LLMs locally with Ollama. Usage: 1. Install Ollama CLI. 2. Pull models (Llama 2, Mistral, CodeLlama). 3. Run with ollama run command. 4. API server for integrations. 5. Model customization with Modelfile. 6. Memory and GPU management. 7. Multi-model switching. 8. No internet required after download. Use for privacy, development, or air-gapped environments.

Build autonomous agents with AutoGPT. Architecture: 1. Goal-oriented task decomposition. 2. Self-critique and iteration. 3. Memory management (short/long-term). 4. Tool usage (web search, file ops). 5. Code execution capability. 6. Human-in-loop checkpoints. 7. Budget constraints for API calls. 8. Plugin system for extensions. Agents plan and execute multi-step tasks independently.

Restaurant-quality steak at home using reverse sear. Prep: 1.5-inch thick ribeye, pat dry, season generously with salt. Step 1: Oven at 225°F, cook on wire rack to 115°F internal (rare). Step 2: Rest 10 minutes while heating cast iron on max. Step 3: Add oil with high smoke point (avocado). Step 4: Sear 60-90 seconds per side. Step 5: Baste with butter, garlic, thyme. Rest: 5-7 minutes, slice against grain. Explain myoglobin denaturation, Maillard reaction, and why reverse sear prevents gray band.

Debug LLM applications with LangSmith. Features: 1. Trace every LLM call. 2. View chain execution steps. 3. Latency and token analysis. 4. Error tracking and debugging. 5. Dataset creation from logs. 6. Evaluation and testing. 7. Feedback collection. 8. Cost monitoring. Essential for production LLM apps. Use to identify bottlenecks and optimize prompts.