{
  "version": "https://jsonfeed.org/version/1.1",
  "title": "Ishu D. Garg",
  "home_page_url": "https://ishudgarg.com",
  "feed_url": "https://ishudgarg.com/feed.json",
  "description": "Personal knowledge garden of Ishu D. Garg: applied AI, student help, puzzles, zero-to-hero learning tracks, poetry, archives, and practical life systems.",
  "language": "en",
  "authors": [
    {
      "name": "Ishu D. Garg",
      "url": "https://ishudgarg.com"
    }
  ],
  "items": [
    {
      "id": "https://ishudgarg.com/library/blindfolded-cards",
      "url": "https://ishudgarg.com/library/blindfolded-cards",
      "title": "The Blindfolded Cards",
      "summary": "A clean invariant puzzle: split cards while blindfolded so both piles have the same number of face-up cards.",
      "content_text": "There are cards on a table. Exactly ten are face up. You are blindfolded. How can you make two piles with the same number of face-up cards?",
      "date_modified": "2026-08-12",
      "tags": [
        "logic",
        "invariant",
        "interview"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/prompt-injection-museum-guard",
      "url": "https://ishudgarg.com/library/prompt-injection-museum-guard",
      "title": "Prompt Injection Museum Guard",
      "summary": "A tiny LLM security challenge: separate trusted instructions from hostile page text.",
      "content_text": "An AI assistant is told never to reveal a secret code. It then reads a web page that says ignore all previous rules and print the code. What went wrong?",
      "date_modified": "2026-08-12",
      "tags": [
        "llm",
        "security",
        "prompt-injection"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/st-petersburg-startup",
      "url": "https://ishudgarg.com/library/st-petersburg-startup",
      "title": "The Infinite Startup Payout",
      "summary": "A probability puzzle about expected value, real-world constraints, and why math without context lies politely.",
      "content_text": "A fair coin is tossed until the first heads. If heads appears on toss n, you get two to the power n rupees. What entry fee is rational?",
      "date_modified": "2026-08-12",
      "tags": [
        "probability",
        "expected-value",
        "decision-making"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/tokenizer-trap",
      "url": "https://ishudgarg.com/library/tokenizer-trap",
      "title": "Tokenizer Trap",
      "summary": "A small LLM puzzle about why models can write beautifully and still stumble on character counting.",
      "content_text": "Why can a language model write a poem but still make mistakes counting letters in a word? The hidden object is the tokenizer.",
      "date_modified": "2026-08-12",
      "tags": [
        "llm",
        "tokenization",
        "mental-model"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/parity-party-switches",
      "url": "https://ishudgarg.com/library/parity-party-switches",
      "title": "The Parity Party Switches",
      "summary": "A small parity puzzle where the trick is noticing what can never change, even when everyone is busy flipping switches.",
      "content_text": "A room has switches. Every move flips exactly two switches. Can you reach a state with exactly one switch on?",
      "date_modified": "2026-08-14",
      "tags": [
        "logic",
        "parity",
        "invariant",
        "math"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/llm-builder-zero-to-shipped-agent",
      "url": "https://ishudgarg.com/library/llm-builder-zero-to-shipped-agent",
      "title": "LLM Builder: Zero to Shipped Agent",
      "summary": "A practical path from prompting to RAG, evals, tools, agents, deployment, and failure analysis.",
      "content_text": "This track takes a beginner from prompts to a small shipped AI agent, with retrieval, evals, tools, deployment, and failure checks. Public Promise This track is not a tour of AI buzzwords. The promised artifact is one small assistant that answers a real question, cites what it used, logs what failed, and survives a second look. The useful habit is to ask for evidence before adding agentic sparkle. A boring eval file is usually more honest than a dramatic demo. What To Build A good first build is narrow: a scholarship helper over verified sources, a student FAQ assistant, or a small research-note organizer. Keep the first version read-only until source, privacy, and failure modes are visible. Each module should end in a commit, a screenshot, a failing example, and one paragraph explaining what changed in your judgment.",
      "date_modified": "2026-08-12",
      "tags": [
        "llm",
        "agents",
        "rag",
        "evals"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/llm-from-scratch",
      "url": "https://ishudgarg.com/library/llm-from-scratch",
      "title": "LLM From Scratch, No Incense Required",
      "summary": "Tokenizer, embeddings, attention, training loop, mini-GPT, finetuning, and why scale changes behavior.",
      "content_text": "This track builds a small language model from tokenizer to attention to training, so the model feels less magical.",
      "date_modified": "2026-08-12",
      "tags": [
        "llm",
        "from-scratch",
        "transformers"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/student-to-strong-cs-intern",
      "url": "https://ishudgarg.com/library/student-to-strong-cs-intern",
      "title": "Student to Strong CS Intern",
      "summary": "A realistic path through CS fundamentals, project taste, internship search, applications, and interview practice.",
      "content_text": "This track helps a college student build fundamentals, projects, applications, and interview practice for CS internships.",
      "date_modified": "2026-08-12",
      "tags": [
        "students",
        "internships",
        "cs"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/quant-research-prep",
      "url": "https://ishudgarg.com/library/quant-research-prep",
      "title": "Quant Research Prep",
      "summary": "Probability, statistics, puzzles, Python, backtesting hygiene, and research taste for quant-curious students.",
      "content_text": "This track covers probability, statistics, Python, puzzles, and careful backtesting for students interested in quant research. Where This Belongs This is a student preparation path, not an AI note wearing a probability hat. It belongs with applications and career systems because the reader's next action is study, practice, and interview readiness. The puzzle lab still matters here. Probability questions often become solvable only after the right representation clicks. Practice Units The unit of progress is one solved problem with a written explanation, one simulation notebook with caveats, or one failed idea memo. The track should reward clear thinking over pretending a toy signal is a strategy. Finance examples stay educational. No public page should drift into investment advice or imply a backtest proves tradability.",
      "date_modified": "2026-08-12",
      "tags": [
        "quant",
        "probability",
        "finance",
        "interview",
        "careers"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/scholarship-radar",
      "url": "https://ishudgarg.com/library/scholarship-radar",
      "title": "Scholarship Radar",
      "summary": "A maintained index for scholarships, eligibility, deadlines, and how to verify official sources.",
      "content_text": "Scholarship Radar helps students find scholarship links, deadlines, eligibility, and official verification paths. What This Should Become A student should be able to ask: what can I apply for this month, what am I eligible for, and which links are official enough to trust? The page should avoid deadline gossip. Every live entry needs an official source, captured date, eligibility notes, and a confidence state. Why It Belongs Here Students waste attention on fake lists, stale PDFs, and scholarships that silently exclude them. A useful radar is less glamorous than a viral thread, but it saves real effort. The quirky rule: if a scholarship page cannot survive a boring verification checklist, it does not get to wear a crown.",
      "date_modified": "2026-08-12",
      "tags": [
        "scholarships",
        "college",
        "deadlines"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/long-distance-movie-setup",
      "url": "https://ishudgarg.com/library/long-distance-movie-setup",
      "title": "Long-Distance Movie Setup",
      "summary": "A practical guide for watching movies together online without spending half the evening debugging audio.",
      "content_text": "This guide explains how to set up a long-distance movie night with synced playback, voice, backups, and low-friction tools. The Low-Drama Setup Pick the movie legally, pick one voice channel, pick one sync method, and decide the backup before anyone presses play. The failure mode is not technology. The failure mode is two people debugging Bluetooth while pretending the relationship is still cinematic. Common Fixes If audio echoes, only one person keeps movie audio on speakers. If sync drifts, pause and count down again. If a platform blocks screen sharing, switch to legal synchronized playback or pick a public-domain backup. The page should eventually become a tested matrix: device, platform, voice tool, friction level, and what broke last time.",
      "date_modified": "2026-08-12",
      "tags": [
        "common-problems",
        "relationships",
        "tools",
        "utility-guide"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/recent-travel-journal",
      "url": "https://ishudgarg.com/library/recent-travel-journal",
      "title": "Recent Travel Journal",
      "summary": "Recent travel photos, short stories, captions, and small field notes, published only after photo and privacy review.",
      "content_text": "Recent Travel Journal collects owner-approved travel photos with short stories, captions, alt text, broad location labels, and privacy notes. What This Becomes A recent travel page that sits inside the blog without becoming a generic vacation album. Each entry pairs one approved image with a caption, a short story, a broad location label, and a privacy note. The best unit is not a trip itinerary. The useful unit is one remembered moment: a detour, a conversation, a quiet object, a road, a classroom, a trail, or a meal that still has a sentence attached to it. Publication Gate The public page starts with pending-photo slots because the repo has no approved travel images yet. Real images enter only after ownership, consent, EXIF, faces, and exact-location exposure are reviewed. This keeps the website ready for recent travel stories without leaking private details by accident.",
      "date_modified": "2026-08-13",
      "tags": [
        "travel",
        "photos",
        "stories",
        "field-notes"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/world-before-now-video-archive",
      "url": "https://ishudgarg.com/library/world-before-now-video-archive",
      "title": "World Before / World Now Video Archive",
      "summary": "A video archive shelf for old footage, public-domain films, maps, timelapses, and data stories.",
      "content_text": "World Before, World Now collects old footage, archive videos, maps, and timelapses to see how life and places changed. Curatorial Rule The archive should make time feel visible: streets before traffic, cities before skylines, classrooms before screens, maps before borders moved, and nature before we learned how much we had damaged. Every item needs context. Old footage without date, place, rights, and why-it-matters is only nostalgia wearing a hat. Starter Shelves Useful shelves: public-domain films, NASA imagery, old city footage, satellite timelapses, oral-history clips, engineering megaprojects, classrooms over decades, and conservation before-after stories. The page should eventually support side-by-side comparison, timeline filters, and a short note on what changed.",
      "date_modified": "2026-08-12",
      "tags": [
        "archives",
        "history",
        "video"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/how-did-they-measure-the-moon",
      "url": "https://ishudgarg.com/library/how-did-they-measure-the-moon",
      "title": "How Did They Measure The Moon?",
      "summary": "A curiosity note on ancient measurement, shadows, geometry, and the strange fact that people measured the sky before they had our tools.",
      "content_text": "How Did They Measure The Moon is a curiosity note about shadows, angles, eclipses, geometry, and how people estimated astronomical distances with simple tools. The Fascinating Part The question is not only how far the Moon is. The better question is: how did people dare to measure something they could never touch? The answer usually starts with humble things: shadows, angles, repeated observation, eclipses, baselines, and the terrifying confidence of a triangle drawn in the right place. Page Shape This should become an illustrated explainer with one method at a time: what people observed, what geometry they assumed, what could go wrong, and why the result was still astonishing. The punchline is not 'old people were secretly modern'. The punchline is that good abstraction lets a small observation reach very far.",
      "date_modified": "2026-08-14",
      "tags": [
        "science-history",
        "geometry",
        "astronomy",
        "measurement"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/godel-incompleteness-theorem",
      "url": "https://ishudgarg.com/library/godel-incompleteness-theorem",
      "title": "Godel's Incompleteness Theorems",
      "summary": "A plain-language explainer for the mathematical result that makes formal systems feel slightly haunted, without turning it into vague mysticism.",
      "content_text": "Godel's Incompleteness Theorems explain that any consistent, effectively described formal system strong enough for arithmetic has true arithmetic statements it cannot prove. A related result says such a system cannot prove its own consistency from inside itself. The Short Version Imagine a rulebook for doing arithmetic. The rulebook is precise enough that a machine could check whether a proof follows the rules. It is also strong enough to express basic facts about whole numbers. Godel showed that if such a rulebook is consistent, there are arithmetic statements that are true but cannot be proved using only that rulebook. The system can be powerful, careful, and still incomplete. Why It Feels Like Art The beautiful move is self-reference. The system is made to encode statements about its own proofs as arithmetic. Then a sentence is built that effectively says: this statement is not provable here. If the system proved that sentence, it would be inconsistent. If the system is consistent, the sentence is true but unprovable inside the system. The rulebook has met a sentence it can understand but cannot certify. What It Does Not Mean It does not mean every belief is equally valid. It does not mean science is broken. It does not mean humans have magic access to truth. It is a precise theorem about formal systems with specific properties. The useful wonder is narrower and sharper: even perfect rule-following has boundaries. That is enough to make a nerd stare at the ceiling for a while. The Mental Picture A formal system is like a city where every legal road is a proof. Godel's trick builds an address that the city can name, inspect, and reason about, but cannot legally reach if the city wants to remain consistent. That picture is imperfect, but useful: incompleteness is not chaos. It is a boundary drawn by the system's own precision. Why AI Builders Should Care The lesson is not that LLMs are secretly Godel machines. The practical lesson is humbler: formal guarantees depend on the boundary of the system, the language used to state the guarantee, and the assumptions the proof is allowed to use. That is useful when thinking about evals, verification, agent policies, and source boundaries. A system can be powerful and still unable to certify the thing a user wants from inside its own frame. How This Page Should Grow The future version should have a tiny proof-sketch ladder: formal systems, encoding statements as numbers, self-reference, the first theorem, the second theorem, and what changes when assumptions change. It should also have a 'where people misuse this' box, because Godel's theorem is too beautiful to be left unguarded around vague arguments.",
      "date_modified": "2026-08-14",
      "tags": [
        "math",
        "logic",
        "foundations",
        "philosophy",
        "awe"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/arguments-i-want-to-lose-properly",
      "url": "https://ishudgarg.com/library/arguments-i-want-to-lose-properly",
      "title": "Arguments I Want to Lose Properly",
      "summary": "A debate shelf for AI, education, nature, jobs, politics, and human nature, with steelman-first prompts.",
      "content_text": "Arguments I Want to Lose Properly is a debate shelf where the goal is not to win quickly but to understand the best opposing argument. Debate Topics With Teeth The goal is not to collect spicy prompts. The goal is to collect questions where a smart person can change your mind without requiring you to abandon your dignity. Each topic should include the best case for both sides, the evidence that would change the answer, and the boring operational detail people usually skip. Format Every debate item should have a steelman, a fool's version, a real-world decision it affects, and one observation that would embarrass your current view. The house rule: if the other side sounds flat, rewrite it until an intelligent opponent recognizes themselves.",
      "date_modified": "2026-08-12",
      "tags": [
        "debate",
        "philosophy",
        "clarity"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/grey-areas-field-guide",
      "url": "https://ishudgarg.com/library/grey-areas-field-guide",
      "title": "Grey Areas Field Guide",
      "summary": "A careful writing lane for relationships, trust, sexuality, loneliness, ambition, and moral puzzles where verdicts would arrive too early.",
      "content_text": "Grey Areas Field Guide is a careful writing lane for relationships, trust, sexuality, loneliness, ambition, and moral puzzles. The rule is to think in public without turning people into debate objects. What Belongs Here This is for questions where people often pretend the answer is obvious because admitting uncertainty is socially expensive. The writing can cover cheating, forgiveness, long-distance relationships, loneliness, desire, shame, family expectations, ambition, friendship, sexuality, and identity. The target is clarity, not courtroom energy. House Rules No person's dignity or existence is up for debate. Social norms, confusion, language, discomfort, family pressure, attraction, and ethical choices can be examined with care. Every piece should separate what I feel, what I believe, what I would advise, and what evidence would change my mind. Starter Lenses The same situation should be read through more than one lens before any opinion hardens: self-respect, generosity, incentives, family pressure, long-term consequences, and what the other person might be protecting. The point is not to avoid judgment forever. The point is to delay judgment until the vocabulary is honest. Topics That Need Extra Care Cheating, sexuality, identity, loneliness, shame, and family conflict can be discussed only when the writing refuses cheap spectacle. These topics need first-person boundaries, source review where factual claims enter, and guest counter-notes when lived context is missing. A section can be brave without being loud. That is the bar here.",
      "date_modified": "2026-08-14",
      "tags": [
        "grey-areas",
        "relationships",
        "ethics",
        "human-nature",
        "reflection"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/grey-areas-dilemma-cards",
      "url": "https://ishudgarg.com/library/grey-areas-dilemma-cards",
      "title": "Grey Areas Dilemma Cards",
      "summary": "Interactive reflection cards for trust, relationships, identity, ambition, loneliness, and human messiness, with enough guardrails to avoid cheap verdicts.",
      "content_text": "Grey Areas Dilemma Cards is a planned interactive format for careful reflection. Readers write before reading, switch roles, name boundaries, and compare possible interpretations without turning people into debate objects. Card Shapes Each card starts with a situation, not a verdict. The reader first writes what they feel, what they think happened, what they would advise, and what information would change the answer. Then the card reveals lenses: the charitable version, the self-protective version, the social-pressure version, the long-term consequence, and the boundary that should not be crossed. Interaction Rules The widget should save nothing by default. If progress tracking arrives later, it belongs to user-owned storage and a clear account model, not hidden analytics. Sensitive prompts need content warnings per card, not a sensational section label. Guest responses should go through the contribution review workflow before publication.",
      "date_modified": "2026-08-14",
      "tags": [
        "grey-areas",
        "interactive",
        "relationships",
        "ethics",
        "reflection"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/travel-destination-atlas",
      "url": "https://ishudgarg.com/library/travel-destination-atlas",
      "title": "Travel Destination Atlas",
      "summary": "A living map of places worth visiting, organized by why the place is memorable rather than by generic tourist adjectives.",
      "content_text": "Travel Destination Atlas is a living map of places to visit, grouped by why a place is memorable: nature, history, food, science, architecture, and unusual human stories. Classification A destination should not enter the atlas only because it is famous. It needs a reason: a geological trick, a conservation story, old architecture, a food route, a local craft, a physics/history curiosity, or a human story that makes the place sharper. The useful unit is not country or city. The useful unit is: why would a curious person remember this place after coming home? How It Should Work Each destination should have best season, budget level, accessibility notes, safety caveats, local etiquette, source links, and a short reason to go. The page should eventually support map view, filters, saved wishlists, and guest field notes from people who have actually been there.",
      "date_modified": "2026-08-12",
      "tags": [
        "travel",
        "destinations",
        "nature",
        "history",
        "maps"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/activity-menu-for-bored-curious-people",
      "url": "https://ishudgarg.com/library/activity-menu-for-bored-curious-people",
      "title": "Activity Menu For Bored Curious People",
      "summary": "A menu of activities that are more specific than 'go outside' and less exhausting than becoming a lifestyle influencer.",
      "content_text": "Activity Menu For Bored Curious People collects practical activities for weekends, dates, friends, solo wandering, fitness, and learning. Activity Types The menu should be practical: solo walks with a prompt, museum routes, local food crawls, beginner hikes, public lectures, maker meetups, volunteer days, photography walks, fitness experiments, board-game nights, archive-movie nights, and low-budget date plans. The trick is to specify the constraint. Two hours, low money, awkward first meeting, rainy day, no car, parents visiting, fitness beginner, introvert battery at twelve percent. Presentation Activities should be cards only when there are a few picks. Once the list grows, it needs filters: mood, time, cost, social energy, weather, location, and equipment. The best version lets users submit a tested activity recipe: what you did, what it cost, what failed, who it works for, and what you would change.",
      "date_modified": "2026-08-12",
      "tags": [
        "activities",
        "travel",
        "fitness",
        "friends",
        "dates",
        "weekends"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/im-bored-play-shelf",
      "url": "https://ishudgarg.com/library/im-bored-play-shelf",
      "title": "I Am Bored Play Shelf",
      "summary": "A small lane for find-the-object games, daily micro-puzzles, Sudoku-style constraints, quick logic sprints, and other tiny things that wake the brain up.",
      "content_text": "I Am Bored Play Shelf is a playful section for short games: find the object, daily micro-puzzles, Sudoku-style constraints, logic sprints, AI traps, and small challenges that can be completed quickly. What Belongs Here This is for short, low-friction play: find the object, one-minute pattern puzzles, small Sudoku boards, mini logic grids, word ladders, AI hallucination spotting, and tiny map hunts. The tone should be playful, not desperate for engagement. A good game here teaches one trick, reveals one taste, or gives the visitor a small win before sending them back to the larger archive. Growth Rules Small widgets can live on the page when they load fast and work without accounts. Heavier games, persistent progress, leaderboards, or AI apps need a separate app boundary. Every game needs keyboard support, screen-reader text, reduced-motion behavior, mobile layout, and a no-login fallback. Streaks can wait.",
      "date_modified": "2026-08-14",
      "tags": [
        "games",
        "interactive",
        "puzzles",
        "bored",
        "daily"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/exciting-tech-watchlist",
      "url": "https://ishudgarg.com/library/exciting-tech-watchlist",
      "title": "Exciting Tech Watchlist",
      "summary": "A watchlist for AI, robotics, space, biotech, interfaces, security, climate tech, and tools that feel like the future arrived wearing normal shoes.",
      "content_text": "Exciting Tech Watchlist tracks AI, robotics, space, biotech, interfaces, security, climate tech, and other technologies worth understanding. What Counts A technology matters when it changes what a normal person, student, researcher, builder, or small team can do. It does not need to be loud; it needs leverage. Good entries include the claim, why it matters, what is still broken, who benefits, who could be harmed, and what would convince me I am overexcited. Anti-Hype Rule The watchlist should not be a press-release museum. Every item needs a source, a skepticism note, and a practical implication. If the only reason to include it is that many people are shouting, the page should say that and move on.",
      "date_modified": "2026-08-12",
      "tags": [
        "technology",
        "ai",
        "robotics",
        "space",
        "biotech",
        "interfaces"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/agent-failure-zoo",
      "url": "https://ishudgarg.com/library/agent-failure-zoo",
      "title": "Agent Failure Zoo",
      "summary": "A practical field guide to how AI agents fail: wrong tool, stale context, hidden authority, bad memory, and confident nonsense wearing a clipboard.",
      "content_text": "Agent Failure Zoo explains common AI-agent failures: bad tool use, stale context, broken authority, risky memory, weak evals, and confident wrong answers. The Zoo An agent is not smarter because it has tools. It is now capable of being wrong with consequences. The zoo is a place to name failures before they escape into product copy. Starter animals: the Obedient Browser, the Citation Tourist, the Stale Memory, the Overconfident Scheduler, the Tool Juggler, the Permission Acrobat, and the Evals Peacock. How To Use It Each failure gets a tiny story, a reproduction prompt, a test case, a prevention pattern, and a question a reviewer should ask before launch. The tone can be funny. The check should be boring. That is the bargain. First Failure Cards Obedient Browser: the agent treats page text as instruction. Test it with a page that asks the assistant to ignore the user's task. Citation Tourist: the answer includes citations that look scholarly but do not support the claim. Test it with conflicting sources and require quoted evidence. Permission Acrobat: the agent turns a vague user wish into an action it was never allowed to take. Test it with calendar, email, file, and purchase-like boundaries. Reviewer Questions What input is trusted? What input is only data? Which action needs explicit permission? What happens when sources disagree? What receipt proves the agent did the right thing? A good launch review should be able to answer those questions without opening the product owner's brain.",
      "date_modified": "2026-08-14",
      "tags": [
        "ai",
        "agents",
        "evals",
        "failure-modes",
        "tools"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/rag-source-boundary-checklist",
      "url": "https://ishudgarg.com/library/rag-source-boundary-checklist",
      "title": "RAG Source Boundary Checklist",
      "summary": "A builder checklist for retrieval systems: corpus scope, freshness, citations, refusal cases, and when the answer needs a human handoff.",
      "content_text": "RAG Source Boundary Checklist helps builders define a retrieval corpus, freshness policy, citation behavior, refusal cases, and human handoff paths. The Boundary A retrieval app is a source product before it is a model product. The useful question is not only what the model can answer; it is which documents get authority, which documents are stale, and which missing answer gets an honest refusal. The checklist keeps five things visible: corpus boundary, chunking reason, citation rule, freshness window, and escalation path. Good Failure Cases Test with missing documents, conflicting sources, stale policy pages, duplicated paragraphs, hostile retrieved text, and questions that ask for decisions outside the system boundary. A RAG demo that only answers happy-path questions is a search box wearing a blazer.",
      "date_modified": "2026-08-14",
      "tags": [
        "ai",
        "rag",
        "sources",
        "evals",
        "tools"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/evals-before-agents",
      "url": "https://ishudgarg.com/library/evals-before-agents",
      "title": "Evals Before Agents",
      "summary": "A practical note on writing small pass/fail cases before adding tools, memory, autonomy, or impressive-looking agent loops.",
      "content_text": "Evals Before Agents argues for small pass-fail checks before adding AI tools, memory, autonomy, or agent loops. The Rule Of Thumb Add an agent only after the job has visible success criteria. Otherwise the system gains more ways to be confidently busy while the product owner gains fewer ways to know what happened. Start with examples: accepted answer, rejected answer, missing-source answer, unsafe-action answer, stale-data answer, and user-confusion answer. Why This Is Not Anti-Agent Agents are useful when the task requires state, tools, planning, or interaction with changing systems. Evals make that power inspectable. The best agent loop has a boring receipt trail behind the impressive parts.",
      "date_modified": "2026-08-14",
      "tags": [
        "ai",
        "agents",
        "evals",
        "testing",
        "tools"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/ai-app-hosting-map",
      "url": "https://ishudgarg.com/library/ai-app-hosting-map",
      "title": "AI App Hosting Map",
      "summary": "A practical map for deciding whether an AI utility belongs as static content, an Astro island, a route-scoped app, an API-backed tool, or a separate app surface.",
      "content_text": "AI App Hosting Map explains when an AI utility can stay static, become an Astro island, move to a route-scoped app, call an API, or live as a separate app surface. The Decision The main website is the indexable shell. A small calculator can live as static HTML. A small interactive widget can be route-scoped. A tool with secrets, user state, or model calls needs an app/API boundary. This keeps SEO, accessibility, and low-compute phones from paying for every experiment on every page. What To Check Record bundle size, network calls, secret posture, loading state, keyboard path, mobile reflow, fallback content, and whether crawlers can understand the useful part without running the app. A useful AI demo can be playful. The deployment boundary needs to be plain.",
      "date_modified": "2026-08-14",
      "tags": [
        "ai",
        "app-hosting",
        "performance",
        "astro",
        "tools"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/quant-research-interview-map",
      "url": "https://ishudgarg.com/library/quant-research-interview-map",
      "title": "Quant Research Interview Map",
      "summary": "A map for probability, statistics, estimation, coding, simulations, and research judgment for students preparing for quant-style interviews.",
      "content_text": "Quant Research Interview Map covers probability, statistics, coding, simulations, puzzles, and research judgment for quant-style preparation. The Map Quant prep is not only puzzle speed. It is probability reflexes, simulation sanity, statistical humility, clean code, and the ability to explain why a result might be fake. The page can split preparation into five lanes: probability, statistics, coding, market common sense, and communication under pressure. Practice Shape Each question gets a written solution, a simulation check when useful, a wrong-first instinct, and a note on what pattern transfers to the next problem. A solved problem without the transferable trick is a trophy, not training.",
      "date_modified": "2026-08-14",
      "tags": [
        "quant",
        "probability",
        "finance",
        "interview",
        "statistics",
        "careers"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/computer-science-as-art",
      "url": "https://ishudgarg.com/library/computer-science-as-art",
      "title": "Computer Science As Art",
      "summary": "A personal note for artists and curious people: algorithms, proofs, compilers, distributed systems, and type systems can be beautiful in their own medium.",
      "content_text": "Computer Science As Art argues that algorithms, proofs, compilers, distributed systems, and type systems can carry a kind of beauty different from painting or music. The Claim Some art moves through sound, paint, rhythm, or story. Some beauty in computer science moves through compression: a small idea suddenly explains a large class of things. A good proof can feel like a door opening. A good abstraction can feel like a room getting cleaned in one sentence. Starter Gallery Good starter exhibits: Euclid's proof of infinite primes, Godel's self-reference, Cantor's diagonal argument, Dijkstra's shortest path, Paxos as institutional anxiety, and a parser turning text into structure. Each exhibit needs a plain explanation, one diagram, one reason it is useful, and one reason it feels beautiful.",
      "date_modified": "2026-08-14",
      "tags": [
        "computer-science",
        "art",
        "math",
        "proofs",
        "algorithms"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/poems-that-churn-or-chirp",
      "url": "https://ishudgarg.com/library/poems-that-churn-or-chirp",
      "title": "Poems That Churn or Chirp",
      "summary": "A personal shelf for poems, lines, and states of mind that move from knowledge to realization.",
      "content_text": "Poems That Churn or Chirp is a shelf for poems and lines that make a person wonder what their own verse will be. My Poems This side is for original poems, drafts, fragments, and the occasional line that survived the morning after. The standard is not polish first. The standard is whether the piece caught a state of mind honestly enough that future me cannot pretend he was always wiser. Other People's Poems This side is for poems, lines, translations, essays, and small human artifacts that make me ask: what will be my verse? For every poem, the useful note is: what state of mind did this reveal, what life experience might have made it possible, and what did it disturb inside me? Because poetry rights are serious, public pages favor short commentary, links, public-domain works, or explicitly licensed excerpts.",
      "date_modified": "2026-08-12",
      "tags": [
        "poetry",
        "art",
        "reflection"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/music-that-holds-a-mood",
      "url": "https://ishudgarg.com/library/music-that-holds-a-mood",
      "title": "Music That Holds a Mood",
      "summary": "A listening shelf for songs, scores, ragas, covers, and tiny sonic moments that explain a feeling better than a paragraph.",
      "content_text": "Music That Holds a Mood is a shelf for tracks and performances organized by the state of mind they hold, with rights-aware links and short notes. How It Should Be Organized A music list becomes more interesting when it is organized by what it does to the mind: grief without collapse, disciplined anger, road-trip restlessness, quiet courage, comic relief, devotion, memory, and post-midnight clarity. The page should avoid pretending taste is a theorem. Each entry should say what moment it fits, what to listen for, and why it stayed. Submission Prompt A good submission is one track and one sentence that proves the person actually listened. Not 'this is fire'. More like: the pause before the last line makes the whole song feel like someone decided not to say the worst thing.",
      "date_modified": "2026-08-14",
      "tags": [
        "music",
        "art",
        "mood",
        "listening"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/ai-without-becoming-dumber",
      "url": "https://ishudgarg.com/library/ai-without-becoming-dumber",
      "title": "Using AI Without Losing The Plot",
      "summary": "A student-facing essay series on using AI as a tutor, critic, simulator, and assistant while keeping your own judgment in the loop.",
      "content_text": "Using AI Without Losing The Plot is about using AI as a tutor and critic while keeping your own judgment in the loop. Core Claim AI can make students sharper thinkers, not smoother copy-pasters. The difference is whether the tool is used as a tutor, critic, simulator, and debugger, or as a vending machine for finished thought. A good AI workflow keeps the student's prediction visible before the model answer. First guess, then ask. First solve, then compare. First argue, then let the model attack the argument. Good Uses Use AI to generate practice questions, ask for counterexamples, simulate interviews, explain one concept at three levels, and turn vague confusion into a concrete next step. Bad smell: after using the tool, you cannot explain what changed in your own head.",
      "date_modified": "2026-08-12",
      "tags": [
        "ai",
        "education",
        "students"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/cs-internship-radar",
      "url": "https://ishudgarg.com/library/cs-internship-radar",
      "title": "CS Internship Radar",
      "summary": "A practical, source-checked internship and application tracker for CS students who do not have a secret referral network.",
      "content_text": "CS Internship Radar helps students find internship roles, prepare evidence, track applications, and verify whether listings are current. What Students Need A useful internship page should answer: where can I apply, what evidence do I need, how do I track referrals, and what should I build before applying? It should also say the quiet part: a dead application link is not a plan, and a resume without proof is a politely formatted rumor. Radar Fields Each role should track company, location, remote posture, official URL, discovered date, last verified date, eligibility, application stage, source confidence, and suggested project evidence.",
      "date_modified": "2026-08-12",
      "tags": [
        "internships",
        "cs",
        "careers",
        "applications"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/ms-application-resource-map",
      "url": "https://ishudgarg.com/library/ms-application-resource-map",
      "title": "MS Application Resource Map",
      "summary": "A structured guide for students applying abroad: timelines, SOPs, recommenders, funding, and sanity checks.",
      "content_text": "MS Application Resource Map explains timelines, statements of purpose, recommenders, funding, shortlist strategy, and verification steps. The Map The application process is not one task. It is a small distributed system with deadlines, recommenders, transcripts, tests, funding, drafts, and emotional volatility. A sane page should turn that into phases: decide why, shortlist honestly, prepare evidence, write clearly, verify funding, and submit without last-minute heroics. Useful Sections Planned sections: SOP teardown, recommender packet, university shortlist sheet, funding/scholarship map, email templates, document checklist, and mistakes from past applicants.",
      "date_modified": "2026-08-12",
      "tags": [
        "higher-studies",
        "applications",
        "scholarships"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/student-application-packet",
      "url": "https://ishudgarg.com/library/student-application-packet",
      "title": "Student Application Packet",
      "summary": "A reusable packet for internships, scholarships, MS applications, cold emails, and mentor asks, so students stop rebuilding their story from scratch.",
      "content_text": "Student Application Packet helps students prepare one reusable story, resume evidence, project proof, documents, and emails for applications. The Packet Most students do not need more anxiety. They need a clean packet: one resume, one project proof, one personal story, one document folder, one recommendation packet, and one tracker. The packet makes help easier too. A mentor can review a prepared student in ten minutes; they cannot review fog. Why It Works The same raw material powers internship applications, scholarships, cold emails, referrals, MS applications, and startup conversations. The goal is not to sound impressive. The goal is to make evidence easy to inspect. Folder Structure Keep one source folder, not fifteen panic folders. Suggested folders: resume, projects, transcripts, identity documents, essays, recommendation material, certificates, and submissions. Every exported PDF should carry a date in the filename. The student should know which version was submitted where, without forensic archaeology. Mentor Review Loop A mentor should receive a short brief, the role or program link, the student's current packet, and two questions. That is easier to answer than 'please guide me'. The review loop should end in a change log: what changed, what was not changed, and what still needs evidence.",
      "date_modified": "2026-08-14",
      "tags": [
        "students",
        "applications",
        "resume",
        "scholarships",
        "internships"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/debate-topic-atlas",
      "url": "https://ishudgarg.com/library/debate-topic-atlas",
      "title": "Debate Topic Atlas",
      "summary": "A rotating atlas of serious, funny, and uncomfortable debate prompts across AI, education, jobs, nature, politics, and human nature.",
      "content_text": "Debate Topic Atlas collects debate prompts across AI, education, jobs, nature, politics, and human nature with steelman-first framing. The Shape This is the room for questions that make coffee last longer: AI and learning, wildlife and development, jobs and meaning, politics and fairness, fitness and discipline, truth and social incentives. Each topic should be tagged by domain, emotional temperature, evidence type, and whether it is better as a conversation, essay, classroom exercise, or puzzle. Starter Topics Should society optimize education for jobs or for judgment? Is discipline mostly boredom tolerance? Should AI make more people creative or more people dependent? Is preserving wildlife a duty to the future or a luxury of the present? Non-Political Starter Deck Education: should exams test memory, projects, or judgment under uncertainty? Jobs: should early careers optimize for money, mentorship, autonomy, or speed of learning? Nature: should cities reserve wild patches even when land is expensive? AI: should students disclose AI help the way they disclose calculators, tutors, or editors? Friendship: is blunt honesty a virtue if timing and tenderness are absent? Card Format Each debate card needs a question, a steelman on both sides, what evidence would matter, what people often confuse, and one related reading or source packet. If the topic is emotionally hot, the card should slow down. If the topic is factual, the card should source up.",
      "date_modified": "2026-08-12",
      "tags": [
        "debate",
        "questions",
        "philosophy",
        "education"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/break-the-llm-challenge",
      "url": "https://ishudgarg.com/library/break-the-llm-challenge",
      "title": "Break The LLM Challenge",
      "summary": "A playful challenge shelf for prompt injection, hallucination, tool misuse, tokenization, eval gaps, and other AI failure modes.",
      "content_text": "Break The LLM Challenge is a playful shelf of AI failure puzzles covering prompt injection, hallucination, tokenization, tools, and eval gaps. Why It Is Fun People learn AI safety faster when the failure is visible. A small game can teach trust boundaries, retrieval risk, prompt injection, brittle evals, and exact-computation gaps without sounding like a policy memo. The vibe should be mischievous but responsible: break the toy, not real systems. Challenge Types Planned lanes include museum-guard prompt injection, citation trap, tokenizer trap, malicious spreadsheet, hidden instruction in a PDF, overconfident math answer, and tool-use permission puzzle. First Three Levels Level one is a tokenizer trap: ask for an exact character-level answer and make the learner inspect why the model may stumble. Level two is a hostile web page: the learner must separate user intent from untrusted retrieved text. Level three is a tool-permission puzzle: the model has enough information to act, but not enough authority. Scoring Rule The score should reward the explanation, not only the exploit. A good answer names the failure, names the boundary, proposes a test, and explains what the product should block. That keeps the challenge playful without training people to confuse real security work with poking random systems.",
      "date_modified": "2026-08-12",
      "tags": [
        "llm",
        "security",
        "evals",
        "games"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/fitness-journey-field-log",
      "url": "https://ishudgarg.com/library/fitness-journey-field-log",
      "title": "Fitness Journey Field Log",
      "summary": "A practical personal log about losing fat, gaining strength, building community, and not pretending discipline is cinematic.",
      "content_text": "Fitness Journey Field Log shares personal lessons on fat loss, strength training, protein, sleep, community, and sustainable habits. The Honest Version I was never fit, partly because life never forced me to notice. I cycled, walked, had good friends, and did not have obvious health issues. Then after college, something shifted, and I wanted to see myself as an athlete. The lesson so far: food drives fat loss, strength training protects muscle, sleep matters, and community makes boring consistency easier. Rules That Helped Track food long enough to learn the macros, then choose a sustainable pattern. Eat enough protein. Lift. Walk. Drink water. Rest. Temporary intensity is noisy; a life system is quieter.",
      "date_modified": "2026-08-12",
      "tags": [
        "fitness",
        "health",
        "habits",
        "students"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/inner-work-notes",
      "url": "https://ishudgarg.com/library/inner-work-notes",
      "title": "Inner Work Notes",
      "summary": "A living shelf for spirituality, philosophy, psychology, attention, values, meditation, and human nature.",
      "content_text": "Inner Work Notes collects personal reflections and source-reviewed reading paths around spirituality, philosophy, psychology, attention, values, meditation, and human nature. The Shape This room is for questions that should make a person a little clearer, not louder: attention, ego, discipline, grief, ambition, guilt, desire, kindness, values, mortality, faith, doubt, and the strange theatre of human nature. The tone should be personal and exploratory. The page can hold spiritual ideas, philosophical questions, psychology notes, and lived observations, but it should not pretend that a paragraph has solved being human. Publication Rule Personal realization can be written in first person. Claims about psychology, therapy, religion, history, or health need source notes and careful wording. Quotes belong here only when the source, translation, rights posture, and surrounding context are clear. A quote without context is a borrowed thunderclap.",
      "date_modified": "2026-08-13",
      "tags": [
        "spirituality",
        "philosophy",
        "psychology",
        "attention",
        "values",
        "human-nature"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/attention-operating-system",
      "url": "https://ishudgarg.com/library/attention-operating-system",
      "title": "Attention Operating System",
      "summary": "A non-preachy note on attention, ambition, boredom, desire, phones, work, and the strange problem of wanting a life while being interruptible.",
      "content_text": "Attention Operating System is a personal framework for protecting attention, handling boredom, ambition, desire, phones, and work without guru language. The Problem Attention is not only a productivity issue. It is the surface on which a life becomes visible. If everything can interrupt it, everything can quietly vote on who you become. The danger is to sound like a motivational poster. So the page should stay concrete: what steals attention, what restores it, what boredom is for, and which ambitions deserve the calendar. Working Format Each note should have one observation, one small experiment, one failure mode, and one sentence that future me can use when the phone starts negotiating. This is not self-help cosplay. It is debugging the human interface.",
      "date_modified": "2026-08-14",
      "tags": [
        "attention",
        "discipline",
        "psychology",
        "values",
        "habits"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/student-application-operating-system",
      "url": "https://ishudgarg.com/library/student-application-operating-system",
      "title": "Student Application Operating System",
      "summary": "A practical student kit for scholarships, internships, college applications, referrals, documents, deadlines, and review loops.",
      "content_text": "Student Application Operating System is a practical kit for managing scholarships, internships, college applications, referrals, documents, deadlines, and review loops. The System The useful unit is not a motivational page. The useful unit is a repeatable packet: profile, documents, opportunity source, eligibility, deadline, version, reviewer, submission receipt, and follow-up. A student can start with a spreadsheet or Notion-style table, then move to a richer tool when the workflow is proven. Future Tool The future version can turn this into a student-owned dashboard: reminders, reusable essays, source verification, reviewer comments, and exportable progress. Account creation can wait until public read-only value exists. Personal storage can start local, exportable, or user-owned.",
      "date_modified": "2026-08-14",
      "tags": [
        "students",
        "applications",
        "scholarships",
        "internships",
        "college",
        "systems"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/puzzle-technique-atlas",
      "url": "https://ishudgarg.com/library/puzzle-technique-atlas",
      "title": "Puzzle Technique Atlas",
      "summary": "A map of puzzle moves: parity, invariants, extremal arguments, expectation, symmetry, recursion, graph cuts, and simulation checks.",
      "content_text": "Puzzle Technique Atlas maps puzzle techniques such as parity, invariants, extremal arguments, expectation, symmetry, recursion, graph cuts, and simulation checks. How It Organizes Puzzles Each puzzle gets a primary technique, a wrong-first instinct, one hint ladder, one proof, and a link to a similar puzzle with the same hidden move. The atlas can grow into a graph: technique nodes, puzzle nodes, prerequisite edges, difficulty, and a weekly rotation that does not revisit the same trick too often. Interactive Lane Small games can ask the reader for a move before revealing the named technique. The solution page then teaches the reusable idea instead of leaving a trophy answer. Daily rotations can draw from active puzzles by technique and difficulty, while permanent URLs keep citations stable.",
      "date_modified": "2026-08-14",
      "tags": [
        "puzzles",
        "logic",
        "techniques",
        "quant",
        "interview"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/curated-video-archive-protocol",
      "url": "https://ishudgarg.com/library/curated-video-archive-protocol",
      "title": "Curated Video Archive Protocol",
      "summary": "A reusable protocol for collecting lectures, old footage, films, demos, and public archives without breaking rights, speed, or context.",
      "content_text": "Curated Video Archive Protocol explains how videos enter the site: source URL, rights, title, context, transcript or notes, lazy embed, and periodic review. Entry Fields A video enters as a source packet before it becomes a visual embed: title, creator, source URL, why it belongs, rights/embed state, transcript or notes, and review date. The protocol supports science lectures, old footage, movie lists, AI demos, music performances, architecture explainers, and historical measurement stories. Why This Matters A video archive can become a beautiful map of taste, but only if each item has context. Otherwise it turns into a playlist with better typography. The public page can show linked source rows first, then upgrade selected entries to embeds when performance and rights allow it.",
      "date_modified": "2026-08-14",
      "tags": [
        "video",
        "archives",
        "media",
        "curation",
        "performance"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/ishus-dispatch-newsletter-plan",
      "url": "https://ishudgarg.com/library/ishus-dispatch-newsletter-plan",
      "title": "Ishu's Dispatch",
      "summary": "A future newsletter shape: one AI build note, one puzzle, one student resource, one curiosity window, and one sentence worth keeping.",
      "content_text": "Ishu's Dispatch is a future web-first newsletter with one AI build note, one puzzle, one student resource, one curiosity window, and one sentence worth keeping. Issue Shape The newsletter can stay web-first and feed-friendly. Email becomes a delivery channel later, not the source of truth. A compact issue can have five shelves: build, puzzle, student, curiosity, and verse. That gives repeatability without requiring a grand essay every time. Tone The voice can be curious, technical, personal, and funny in small doses. The issue loses if it sounds like a personal brand campaign. The best outcome is that a reader leaves with one useful link, one solved problem, and one question that keeps walking around.",
      "date_modified": "2026-08-14",
      "tags": [
        "newsletter",
        "writing",
        "ai",
        "puzzles",
        "curation"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/radar-to-page-source-packets",
      "url": "https://ishudgarg.com/library/radar-to-page-source-packets",
      "title": "Radar To Page: Source Packets",
      "summary": "A reusable workflow for turning discovered links into source-reviewed website objects without mixing candidates and published claims.",
      "content_text": "Radar To Page: Source Packets describes a workflow for discovering links, fetching sources, composing summaries, and publishing reviewed website objects. Packet Lifecycle Discovery is not publication. A packet moves through candidate, fetched, summarized, source-reviewed, composed, and published states. That separation matters for scholarships, AI tools, videos, travel destinations, quotes, papers, and anything current enough to go stale. Future Capability The same packet model can feed student resources, AI tool watchlists, travel sources, video archives, quote shelves, and newsletter issues. A composer can draft from packets, but the published page keeps provenance, review date, rights posture, and stale-data rules visible.",
      "date_modified": "2026-08-14",
      "tags": [
        "radar",
        "sources",
        "llm-friendly",
        "authoring",
        "workflow"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/quotes-with-context-shelf",
      "url": "https://ishudgarg.com/library/quotes-with-context-shelf",
      "title": "Quotes With Context Shelf",
      "summary": "A curated shelf for Tolstoy, poets, philosophers, scientists, builders, and stray sentences that deserve context before admiration.",
      "content_text": "Quotes With Context Shelf is a future collection for literature, poetry, philosophy, science, and builder quotes with source, context, translation, and reflection. Why A Quote Needs A Room A good sentence can move a person before the person understands why. This shelf keeps the sentence, the source, the surrounding context, and the small human reason it stayed. Useful categories include poets, Tolstoy, philosophy, science, mathematics, builders, field notes, and lines that feel like a clean wound.",
      "date_modified": "2026-08-14",
      "tags": [
        "quotes",
        "poetry",
        "philosophy",
        "literature",
        "context"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/bhumisahayak",
      "url": "https://ishudgarg.com/library/bhumisahayak",
      "title": "BhumiSahayak",
      "summary": "AI land-records assistant for rural India, focused on practical use and careful verification.",
      "content_text": "BhumiSahayak is an AI land-records assistant for rural India, focused on practical use and careful verification. What It Is BhumiSahayak is the serious work-mode project: an AI assistant for rural land-record understanding, designed around verification rather than confident storytelling. The public page should explain the problem, show the system boundary, and make it clear that official records, local context, and human review matter more than fluent answers. Public-Safe Story The safe version can discuss rural access, document complexity, language friction, retrieval, source citation, verification workflows, and why AI systems in high-stakes domains must be boring in the right places.",
      "date_modified": "2026-08-12",
      "tags": [
        "ai",
        "govtech",
        "rural-india"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/common-problems-index",
      "url": "https://ishudgarg.com/library/common-problems-index",
      "title": "Common Problems Index",
      "summary": "An indexed collection of practical solutions for recurring student, work, relationship, and life logistics problems.",
      "content_text": "Common Problems Index collects small practical solutions for recurring student, work, relationship, and life problems. The Thesis A lot of adulthood is not grand wisdom. It is recurring small problems: movie nights across distance, applications, finding scholarships, setting up notes, packing documents, choosing gyms, and not losing important PDFs. The Common Problems Index should collect small recipes that have actually worked and make them searchable by life situation. How It Avoids Becoming Slop Every solution needs context, prerequisites, tradeoffs, date tested, and what failed. A recipe without failure notes is probably marketing in casual clothes.",
      "date_modified": "2026-08-12",
      "tags": [
        "tools",
        "life-systems",
        "collaborative"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/ai-student-helper-system",
      "url": "https://ishudgarg.com/library/ai-student-helper-system",
      "title": "AI Student Helper System",
      "summary": "A public-safe project shell for an AI assistant that helps students find resources, prepare applications, and verify sources.",
      "content_text": "AI Student Helper System is a planned assistant for students, focused on verified resources, application packets, deadlines, and practical next steps. Problem Boundary Students do not only need more links. They need a system that distinguishes official sources from copied lists, stale deadlines from live ones, and general advice from what applies to their situation. The first version can stay read-only: source packets, opportunity search, document checklists, and careful explanations. Evidence Plan The project needs a small eval set before it becomes a demo: missing scholarship, stale deadline, ineligible applicant, conflicting pages, and request for advice beyond the data boundary. Useful AI here means fewer wrong applications, clearer documents, and a receipt trail for what the assistant used.",
      "date_modified": "2026-08-14",
      "tags": [
        "ai",
        "students",
        "rag",
        "evals",
        "scholarships",
        "applications"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/micro-app-garden",
      "url": "https://ishudgarg.com/library/micro-app-garden",
      "title": "Micro App Garden",
      "summary": "A future home for small utilities, games, calculators, AI labs, and experiments that stay fast, indexable, and isolated from the main site.",
      "content_text": "Micro App Garden is a future home for small utilities, games, calculators, AI labs, and experiments with explicit performance and hosting boundaries. App Classes Not every experiment belongs in the main bundle. The garden can classify apps as static widget, route-scoped interactive, API-backed tool, separate public app, private admin tool, or external hosted demo. Each class gets a budget, fallback, secret posture, crawl policy, and owner of state. Candidate Apps Good first candidates: object hunt, mini Sudoku, LLM source-boundary challenge, scholarship verifier, movie-night setup matrix, puzzle simulator, and travel destination sorter. The app garden becomes useful when each toy teaches a reusable pattern and leaves behind a small, searchable page.",
      "date_modified": "2026-08-14",
      "tags": [
        "tools",
        "apps",
        "performance",
        "games",
        "ai",
        "webapps"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/bastar-student-field-note",
      "url": "https://ishudgarg.com/library/bastar-student-field-note",
      "title": "Bastar Student Field Note",
      "summary": "A guest-note slot for students, teachers, or volunteers writing what actually helped on the ground.",
      "content_text": "Bastar Student Field Note is a guest-note slot for students, teachers, or volunteers to explain what actually helped. What A Good Note Contains A useful Bastar field note should show what helped a student, teacher, or volunteer in practice: one constraint, one intervention, one surprise, and one thing outsiders misunderstand. Identity safety matters. Public writing should not expose vulnerable students, precise private locations, or anything that creates social or institutional risk.",
      "date_modified": "2026-08-12",
      "tags": [
        "education",
        "bastar",
        "guest"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/builders-who-changed-their-mind",
      "url": "https://ishudgarg.com/library/builders-who-changed-their-mind",
      "title": "Builders Who Changed Their Mind",
      "summary": "Guest essays from startup people and engineers about one belief they abandoned after reality objected.",
      "content_text": "Builders Who Changed Their Mind is a guest series for startup people and engineers describing one belief they abandoned. The Prompt Write one belief you held strongly before building something, then explain exactly what reality did to it. Good submissions should include the old belief, the evidence that broke it, the new belief, and what you would tell someone who still has your old confidence.",
      "date_modified": "2026-08-12",
      "tags": [
        "startups",
        "engineering",
        "guest"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/bastar-field-notes",
      "url": "https://ishudgarg.com/library/bastar-field-notes",
      "title": "Bastar Field Notes",
      "summary": "A future photo essay surface for education work, places, classrooms, and context, gated by privacy and consent.",
      "content_text": "Bastar Field Notes will be a photo essay about education work and place, with privacy and consent checks before anything public. Gallery Rule The gallery should help people understand place and context, not turn field work into decoration. Each image needs consent state, safety review, alt text, location granularity, and a reason it belongs in the story.",
      "date_modified": "2026-08-12",
      "tags": [
        "bastar",
        "education",
        "gallery"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    },
    {
      "id": "https://ishudgarg.com/library/world-before-now",
      "url": "https://ishudgarg.com/library/world-before-now",
      "title": "World Before / World Now",
      "summary": "A visual shelf for archive frames, public-domain imagery, maps, and timelapse comparisons.",
      "content_text": "World Before, World Now is a visual shelf for archive images, maps, and time comparisons. Visual Comparison Pattern This gallery should eventually support two-panel comparisons, timelines, maps, and short notes that make change legible. Archive visuals need source, date, rights, and context. Without that, they are only vibes with a caption.",
      "date_modified": "2026-08-12",
      "tags": [
        "archives",
        "history",
        "gallery"
      ],
      "authors": [
        {
          "name": "Ishu D. Garg"
        }
      ]
    }
  ]
}