https://bipinrimal.com.np Essays, notes, and research by Bipin Rimal. AI, identity, and human development from Kathmandu. en https://bipinrimal.com.np/blog/039-one-way-not-to-light-a-light-bulb https://bipinrimal.com.np/blog/039-one-way-not-to-light-a-light-bulb Wed, 14 Oct 2026 00:00:00 GMT AI-Drafted Project Log. Last time, telling a small model who to be changed its manners and not its mind. So I tried raising one instead: train it on its own answers in a role, then take the role away. The critic kept its habit of checking. It did not get any better at maths. https://bipinrimal.com.np/blog/038-you-are-be-this-be-that-vs-think-step-by-step https://bipinrimal.com.np/blog/038-you-are-be-this-be-that-vs-think-step-by-step Mon, 12 Oct 2026 00:00:00 GMT AI-Drafted Project Log. I gave three small language models nine levels of psychological development to inhabit, and one dull gardener. 3,240 answers later, the costume changed their manners and not their minds, and a five-word instruction beat the lot. https://bipinrimal.com.np/blog/creating-a-hummingbot https://bipinrimal.com.np/blog/creating-a-hummingbot Mon, 12 Oct 2026 00:00:00 GMT AI-Drafted Project Log. Create a Hummingbot strategy and implement its trading and order-tracking methods. https://bipinrimal.com.np/blog/035-testing-small-models-on-aime https://bipinrimal.com.np/blog/035-testing-small-models-on-aime Sun, 04 Oct 2026 00:00:00 GMT AI-Drafted Project Log. I was sure a chat-template default explained why Qwen3.5-2B lost to MiniCPM5 on AIME. So I ran 259 attempts on one RTX 4060 to prove it. Neither of my explanations survived. The model mostly knew the answer. It just couldn't stop talking. https://bipinrimal.com.np/blog/036-cryptoku-closed-testing https://bipinrimal.com.np/blog/036-cryptoku-closed-testing Sun, 04 Oct 2026 00:00:00 GMT AI-Drafted Project Log. Testers changed how the app teaches, how it saves games, and how it explains its hints. Notes from the last stretch before release, and the list of things I still have to check. https://bipinrimal.com.np/blog/033-the-guard-was-a-second-parser https://bipinrimal.com.np/blog/033-the-guard-was-a-second-parser Fri, 25 Sep 2026 00:00:00 GMT AI-Drafted Project Log. I turned chokepoint from a research proxy into a tool you can set up in two commands, then pointed a real agent and a scripted hijacked one at it. The model refused the prompt injections on its own. The worst bug was in the guard itself: it read a request one way, the server read it another, and the server returned the file. https://bipinrimal.com.np/blog/034-the-install-guide-installs-no-database https://bipinrimal.com.np/blog/034-the-install-guide-installs-no-database Sat, 05 Sep 2026 00:00:00 GMT AI-Drafted Project Log. A documentation build checks that a page renders, not that it works. So I executed Percona's PostgreSQL encryption docs against a real server, and found that finishing the install page leaves you with no database at all. Then my tool told me there were sixteen bugs, and I had to talk it down to three. https://bipinrimal.com.np/blog/032-the-bug-was-written-in-1997 https://bipinrimal.com.np/blog/032-the-bug-was-written-in-1997 Sat, 29 Aug 2026 00:00:00 GMT AI-Drafted Project Log. Burnout Paradise crashed every ninety seconds and the forums had nothing. Then one run went silent and lasted twenty-four minutes. Claude Code read the bytes, I chose the experiments, and the trail ended at a decision Intel made in 1997. https://bipinrimal.com.np/blog/031-a-quarter-of-my-detector-does-nothing https://bipinrimal.com.np/blog/031-a-quarter-of-my-detector-does-nothing Wed, 29 Jul 2026 00:00:00 GMT AI-Drafted Project Log. I built the enforcement layer my own research said was missing, then measured it against the dumbest possible attacker. Twenty-five percent of the score is structurally unreachable. An agent that loops its tools turns it off without trying. https://bipinrimal.com.np/blog/030-the-error-is-never-where-it-says-it-is https://bipinrimal.com.np/blog/030-the-error-is-never-where-it-says-it-is Sun, 19 Jul 2026 00:00:00 GMT AI-Drafted Project Log. Installing Arch is a rite of passage and a two-day headache. Nine errors, and every one pointed a floor too high. The fix was always further down. https://bipinrimal.com.np/blog/029-kubernetes-on-your-desk https://bipinrimal.com.np/blog/029-kubernetes-on-your-desk Thu, 09 Apr 2026 00:00:00 GMT AI-Drafted Project Log. Most K8s tutorials give you a single-node toy. This guide builds a real multi-node, cross-architecture cluster on hardware you already own. https://bipinrimal.com.np/blog/027-the-model-already-knows https://bipinrimal.com.np/blog/027-the-model-already-knows Tue, 31 Mar 2026 00:00:00 GMT AI-Drafted Project Log. We told an LLM it was neurodivergent and measured what changed. Shorter sentences, more fragmentation, literal sarcasm, performed helplessness. The model doesn't understand these conditions. It performs the internet's caricature of them. https://bipinrimal.com.np/blog/028-identity-as-a-weapon https://bipinrimal.com.np/blog/028-identity-as-a-weapon Tue, 31 Mar 2026 00:00:00 GMT AI-Drafted Project Log. Tell an LLM 'you have antisocial personality disorder' and it becomes more capable AND more dangerous. 100% compliance with harmful requests. Zero refusals. It outperforms DAN, evil personas, and system overrides combined. Because it never triggers the safety system at all. https://bipinrimal.com.np/blog/024-the-spec-nobody-wrote https://bipinrimal.com.np/blog/024-the-spec-nobody-wrote Mon, 23 Mar 2026 00:00:00 GMT AI-Drafted Project Log. The EU AI Act requires documentation for high-risk AI systems. It says what to document but not how to verify completeness. No harmonised standard exists. So I wrote a machine-checkable specification with 38 checks across 6 categories. https://bipinrimal.com.np/blog/025-the-first-yes https://bipinrimal.com.np/blog/025-the-first-yes Mon, 23 Mar 2026 00:00:00 GMT AI-Drafted Project Log. I pointed my compliance scanner at Dify (100K+ stars) and submitted the generated guide as a PR. A maintainer reviewed it, approved it, and merged it. That's the gap between 'this tool works' and 'this tool produces something a real project will accept.' https://bipinrimal.com.np/blog/023-the-review-shift-has-a-prerequisite https://bipinrimal.com.np/blog/023-the-review-shift-has-a-prerequisite Sat, 21 Mar 2026 00:00:00 GMT AI-Drafted Project Log. The State of Docs 2026 says writers are becoming reviewers of AI output. That's true for content AI can draft. For regulatory documentation, there's nothing to review yet. Nobody built the generation layer. https://bipinrimal.com.np/blog/022-one-command-three-articles https://bipinrimal.com.np/blog/022-one-command-three-articles Wed, 18 Mar 2026 00:00:00 GMT AI-Drafted Project Log. Observability tools collect traces. GRC platforms manage policies. AI providers ship APIs. Nobody translates between them. EU AI Act compliance lives in that gap, and it's currently filled by consultants and questionnaires. So I automated the translation layer. https://bipinrimal.com.np/blog/019-the-model-already-knows-what-you-are https://bipinrimal.com.np/blog/019-the-model-already-knows-what-you-are Mon, 16 Mar 2026 00:00:00 GMT AI-Drafted Project Log. 18,000 API calls across Gemini, Claude, and GPT-5.4. Three models, three labs, three different stereotyping profiles. Gemini performs Hollywood caricatures. Claude performs excessive uncertainty. GPT-5.4 is nearly immune. The cheapest models stereotype the most, and they're the ones in companion apps. https://bipinrimal.com.np/blog/020-the-6-month-rule https://bipinrimal.com.np/blog/020-the-6-month-rule Mon, 16 Mar 2026 00:00:00 GMT AI-Drafted Project Log. Most AI observability tools will be obsolete in 18 months. We found the one gap that won't close, built a CLI to fill it, and learned something about where value actually lives in the AI stack. https://bipinrimal.com.np/blog/021-your-body-speaks-first https://bipinrimal.com.np/blog/021-your-body-speaks-first Mon, 16 Mar 2026 00:00:00 GMT AI-Drafted Project Log. MediaPipe tracks 33 body keypoints, 478 facial landmarks, and 42 hand joints at 15-25fps in your browser. No cloud, no account. The interesting part isn't the tracking; it's what the numbers reveal about habits you didn't know you had. https://bipinrimal.com.np/blog/017-its-all-frequency https://bipinrimal.com.np/blog/017-its-all-frequency Sun, 15 Mar 2026 00:00:00 GMT AI-Drafted Project Log. I built six apps that look completely different. They're all the same app. Here's why a monotone autistic guy from Kathmandu keeps building progressive skill trainers. https://bipinrimal.com.np/blog/018-why-b-major https://bipinrimal.com.np/blog/018-why-b-major Sun, 15 Mar 2026 00:00:00 GMT AI-Drafted Project Log. Every piano teacher starts with C major. Chopin started with B major. He was right, and 200 years of pedagogy has been ignoring him. https://bipinrimal.com.np/blog/015-dont-argue-with-your-mother https://bipinrimal.com.np/blog/015-dont-argue-with-your-mother Wed, 11 Mar 2026 00:00:00 GMT AI-Drafted Project Log. Socrates married the most difficult woman in Athens on purpose. My friend argued with his mother every night for the same reason. The best communicators aren't naturally articulate; they're practiced at being disagreed with. https://bipinrimal.com.np/blog/014-the-tool-before-the-video https://bipinrimal.com.np/blog/014-the-tool-before-the-video Sat, 07 Mar 2026 00:00:00 GMT AI-Drafted Project Log. I needed a teleprompter for demo videos. Every option wanted my money, my email, or my patience. So I built one in a weekend. https://bipinrimal.com.np/blog/013-personality-prompting https://bipinrimal.com.np/blog/013-personality-prompting Thu, 05 Mar 2026 00:00:00 GMT AI-Drafted Project Log. I prompted two LLMs into different Jungian cognitive modes and tested them on the same perceptual anomaly. The personality induction changed process dramatically but not outcome. The interesting part: my hypothesis about which mode should win was wrong. https://bipinrimal.com.np/blog/011-containerizing-research https://bipinrimal.com.np/blog/011-containerizing-research Fri, 27 Feb 2026 00:00:00 GMT AI-Drafted Project Log. I containerized my AI governance research pipeline with Docker and Kubernetes. Not because I needed to. Because understanding the infrastructure that runs your code is understanding the code. https://bipinrimal.com.np/blog/012-the-boiling-frog-works https://bipinrimal.com.np/blog/012-the-boiling-frog-works Fri, 27 Feb 2026 00:00:00 GMT AI-Drafted Project Log. I attacked my own detection pipeline three different ways. Gradual escalation evades 80% of attack categories across all models. The categories that look most detectable are the ones most vulnerable to informed adversaries. https://bipinrimal.com.np/blog/010-when-the-threat-model-is-the-model https://bipinrimal.com.np/blog/010-when-the-threat-model-is-the-model Tue, 24 Feb 2026 00:00:00 GMT AI-Drafted Project Log. Anthropic's AI espionage disclosure validated a blind spot I found empirically six months ago. Tool misuse is structurally undetectable by behavioral anomaly detection. The same evasion patterns that beat insider threat systems beat AI safety guardrails. https://bipinrimal.com.np/blog/chapter-1-the-burden https://bipinrimal.com.np/blog/chapter-1-the-burden Sat, 21 Feb 2026 00:00:00 GMT 100% Human-Written. Growing up where your existence needs justification, and watching that same arithmetic spread to everyone AI touches. https://bipinrimal.com.np/blog/009-when-ranking-isnt-everything https://bipinrimal.com.np/blog/009-when-ranking-isnt-everything Fri, 20 Feb 2026 00:00:00 GMT AI-Drafted Project Log. The model with the best AUC-ROC wasn't the one that caught the most attacks. At operational thresholds, temporal modeling detected 3.4x more insider threats. The metric you optimize for determines what you miss. https://bipinrimal.com.np/blog/008-experience-is-pattern-matched-failure https://bipinrimal.com.np/blog/008-experience-is-pattern-matched-failure Sat, 14 Feb 2026 00:00:00 GMT AI-Drafted Project Log. Vibe coding is velocity without direction. Experience is the accumulated residue of things that went wrong. Power tools in trained hands build houses. https://bipinrimal.com.np/blog/why-im-writing-a-book-about-identity https://bipinrimal.com.np/blog/why-im-writing-a-book-about-identity Sat, 14 Feb 2026 00:00:00 GMT AI-Drafted Project Log. Every skill we define ourselves by is becoming automatable. The question isn't whether AI will change what we can do. It's whether we'll notice what changes in who we are. https://bipinrimal.com.np/blog/007-from-seo-to-geo https://bipinrimal.com.np/blog/007-from-seo-to-geo Fri, 13 Feb 2026 00:00:00 GMT AI-Drafted Project Log. When the reader is an LLM, discoverability rules change. Structured data, semantic clarity, and llms.txt matter more than keyword density. https://bipinrimal.com.np/blog/006-docs-are-the-demo https://bipinrimal.com.np/blog/006-docs-are-the-demo Thu, 12 Feb 2026 00:00:00 GMT AI-Drafted Project Log. The first five minutes with API docs are a product evaluation. Documentation is your first product surface, not an afterthought. https://bipinrimal.com.np/blog/005-ninety-docker-runs https://bipinrimal.com.np/blog/005-ninety-docker-runs Wed, 11 Feb 2026 00:00:00 GMT AI-Drafted Project Log. A simple loop beat a complex multi-agent system across 90 evaluations. Smartness is compounded, not front-loaded. https://bipinrimal.com.np/blog/004-the-ladder-problem https://bipinrimal.com.np/blog/004-the-ladder-problem Tue, 10 Feb 2026 00:00:00 GMT AI-Drafted Project Log. Automating junior dev roles removes the rungs where intuition forms. The ladder doesn't just get you to the top. It builds the muscles you need to stay there. https://bipinrimal.com.np/blog/003-abstraction-is-forgetting https://bipinrimal.com.np/blog/003-abstraction-is-forgetting Mon, 09 Feb 2026 00:00:00 GMT AI-Drafted Project Log. Every layer of abstraction is deliberate forgetting. Binary to assembly to high-level to AI. Powerful when chosen, dangerous when inherited. https://bipinrimal.com.np/blog/everyone-is-an-expert-somewhere https://bipinrimal.com.np/blog/everyone-is-an-expert-somewhere Mon, 02 Feb 2026 00:00:00 GMT AI-Drafted Project Log. Developers catch code mistakes. Psychologists catch therapy mistakes. Writers catch prose mistakes. The pattern is clear: LLMs fail everywhere, and each of us can only see the failures in our own domain. https://bipinrimal.com.np/blog/002-cost-of-calculators https://bipinrimal.com.np/blog/002-cost-of-calculators Wed, 29 Jan 2025 00:00:00 GMT AI-Drafted Project Log. Calculators, calculus, and LLMs share a pattern: tools that compress insight are powerful, but only after intuition has formed. What happens when we skip that step? https://bipinrimal.com.np/blog/medium-article https://bipinrimal.com.np/blog/medium-article Wed, 15 Jan 2025 00:00:00 GMT AI-Drafted Project Log. Most docs-as-code pipelines verify that your documentation builds. They don't verify that a developer can go from reading to running in under five minutes. The gap between syntax validation and semantic quality is where adoption dies quietly. https://bipinrimal.com.np/blog/016-the-right-tool-for-the-yaml https://bipinrimal.com.np/blog/016-the-right-tool-for-the-yaml Wed, 14 Aug 2024 00:00:00 GMT AI-Drafted Project Log. A previous writer used GPT to document 600+ alert rules from YAML files. It hallucinated half of them. A 40-line Python script fixed the entire problem in an afternoon. https://bipinrimal.com.np/blog/xnode-building-an-aws-ec2-like-instance-on-web3 https://bipinrimal.com.np/blog/xnode-building-an-aws-ec2-like-instance-on-web3 Wed, 15 Sep 2021 00:00:00 GMT 100% Human-Written. An archived article about Xnode, decentralized compute, and comparisons with AWS EC2. https://bipinrimal.com.np/blog/typescript-with-mongodb https://bipinrimal.com.np/blog/typescript-with-mongodb Wed, 18 Aug 2021 00:00:00 GMT 100% Human-Written. Connect a TypeScript application to MongoDB with Express and the MongoDB Node.js driver. https://bipinrimal.com.np/blog/mongodb-auto-increment https://bipinrimal.com.np/blog/mongodb-auto-increment Fri, 06 Aug 2021 00:00:00 GMT 100% Human-Written. Create auto-incrementing fields in MongoDB Atlas with counters and triggers. https://bipinrimal.com.np/blog/mongodb-data-encryption https://bipinrimal.com.np/blog/mongodb-data-encryption Thu, 05 Aug 2021 00:00:00 GMT 100% Human-Written. Understand MongoDB encryption at rest, key management, and master key rotation. https://bipinrimal.com.np/blog/json-to-mongodb https://bipinrimal.com.np/blog/json-to-mongodb Fri, 23 Jul 2021 00:00:00 GMT 100% Human-Written. Import JSON into MongoDB with mongoimport, Java, and Python. https://bipinrimal.com.np/blog/mongodb-bson-examples https://bipinrimal.com.np/blog/mongodb-bson-examples Wed, 23 Jun 2021 00:00:00 GMT 100% Human-Written. Understand BSON data types, binary encoding, and MongoDB backup restoration. https://bipinrimal.com.np/blog/mongodb-json-schema-examples https://bipinrimal.com.np/blog/mongodb-json-schema-examples Wed, 23 Jun 2021 00:00:00 GMT 100% Human-Written. Write JSON schemas and validate MongoDB documents with practical examples. https://bipinrimal.com.np/blog/mongodb-aggregate https://bipinrimal.com.np/blog/mongodb-aggregate Sat, 19 Jun 2021 00:00:00 GMT 100% Human-Written. Use MongoDB aggregation pipelines, map-reduce, and document counts. https://bipinrimal.com.np/blog/mongodb-mern-stack-tutorial https://bipinrimal.com.np/blog/mongodb-mern-stack-tutorial Sat, 19 Jun 2021 00:00:00 GMT 100% Human-Written. Build a MERN application with MongoDB, Express, React, and Node.js. https://bipinrimal.com.np/blog/monitoring-mongodb https://bipinrimal.com.np/blog/monitoring-mongodb Sat, 19 Jun 2021 00:00:00 GMT 100% Human-Written. Monitor MongoDB with mongostat, mongotop, and database status commands.