https://orionfold.com/story/ Building in public: short, honest notes on what we shipped, what broke, and what we learned along the way. en-us Orionfold LLC Wed, 22 Jul 2026 00:00:00 GMT https://orionfold.com/story/limitless-without-the-pill/ https://orionfold.com/story/limitless-without-the-pill/ Wed, 22 Jul 2026 00:00:00 GMT Going AI-native has felt like the closest real-world version of that fantasy... without the pill, the blackouts, or the Russian loan sharks. Orionfold Flow AI-native work Building in public https://orionfold.com/story/keep-the-data-see-the-margin/ https://orionfold.com/story/keep-the-data-see-the-margin/ Sun, 05 Jul 2026 00:00:00 GMT The AI you rent to run your client work can see your client work. So I built the opposite. Orionfold Relay is a free engine you run yourself, every result waits for your approval, and it shows you the real margin on each client. It launches on Product Hunt today. Orionfold Relay Building in public Sovereign AI Product Hunt https://orionfold.com/story/the-fix-that-changed-the-leaderboard/ https://orionfold.com/story/the-fix-that-changed-the-leaderboard/ Sat, 27 Jun 2026 00:00:00 GMT I found two honesty bugs in my own product and fixed both the same day. Then I reran the three-way governance bench on the fixed code. A four-billion-parameter model on my laptop scored highest, the famous frontier model came last, and the cost column finally told the truth. Here is the whole receipt, bugs and all. Proof Local AI Cost Trust https://orionfold.com/story/a-4b-model-scored-18-out-of-21-on-my-laptop/ https://orionfold.com/story/a-4b-model-scored-18-out-of-21-on-my-laptop/ Wed, 24 Jun 2026 00:00:00 GMT A published governance benchmark, run on a four-billion-parameter model, on an Apple laptop, for nothing. It refused all nine trick questions, matched the big-iron baseline answer for answer, and missed the same three. Here is the whole receipt. Proof Local AI DGX Spark Cost https://orionfold.com/story/same-input-same-receipt/ https://orionfold.com/story/same-input-same-receipt/ Wed, 24 Jun 2026 00:00:00 GMT A benchmark you cannot rerun is a rumor. Every Proof Receipt carries a config hash, twelve characters that reproduce the whole run to the byte. Here is what that small string actually buys you, and why I built the product around it. Proof Trust Local AI https://orionfold.com/story/the-famous-model-lost/ https://orionfold.com/story/the-famous-model-lost/ Wed, 24 Jun 2026 00:00:00 GMT I put a four-billion-parameter model that runs on my laptop in a head-to-head with Claude Opus and a frontier GLM, on a governance task, scored by a machine. The small local one won, for nothing, and faster. Here is the whole receipt, including the part where the scorer was too harsh. Proof Local AI Cost Trust https://orionfold.com/story/the-lab-that-shipped-itself/ https://orionfold.com/story/the-lab-that-shipped-itself/ Sun, 14 Jun 2026 00:00:00 GMT The paid edition of my AI lab launched this weekend, and the lab shipped itself. My plan said ten weeks. A team of AI agents I conduct did it in a day. Here is how, and what you get. Arena Field Edition Building in public Multi-agent DGX Spark https://orionfold.com/story/building-in-public/ https://orionfold.com/story/building-in-public/ Sun, 07 Jun 2026 00:00:00 GMT Every story here, every number and honest miss, I put out in public on purpose. This is why I build in the open, and how you can help shape what comes next. Building in public Sponsorship AI Native https://orionfold.com/story/the-run-i-almost-wasted/ https://orionfold.com/story/the-run-i-almost-wasted/ Sun, 07 Jun 2026 00:00:00 GMT I built Kepler, a space-math model, on one desktop. Three cheap checks told me the big training run would change nothing. I ran it anyway, and proving them right was the win. Models Fine-tuning DGX Spark https://orionfold.com/story/a-book-you-can-read-and-run/ https://orionfold.com/story/a-book-you-can-read-and-run/ Sat, 06 Jun 2026 00:00:00 GMT I write books you can run, not just read. The code works and every page is online. Here is why open, reproducible books beat theory behind a paywall. Books Offerings AI Native https://orionfold.com/story/the-glue-tax/ https://orionfold.com/story/the-glue-tax/ Fri, 05 Jun 2026 00:00:00 GMT Every AI project, I kept rewriting the same boring plumbing. I named it the glue tax. So I built fieldkit, an open kit of 12 tested parts, to stop paying it. fieldkit Software Building in public https://orionfold.com/story/four-things-you-actually-control/ https://orionfold.com/story/four-things-you-actually-control/ Thu, 04 Jun 2026 00:00:00 GMT The AI divide feels out of your hands. It is not. There are four things you can own instead of rent. Playbooks, software, models, and a device of your own. AI Native Offerings Local AI https://orionfold.com/story/why-most-teams-get-nothing-from-ai/ https://orionfold.com/story/why-most-teams-get-nothing-from-ai/ Wed, 03 Jun 2026 00:00:00 GMT Companies poured 30 to 40 billion dollars into AI and 95% got nothing back. The reason is not cost or tech. It is learning. That gap is why Orionfold exists. AI Native Conviction Building in public https://orionfold.com/story/why-i-folded-orionfold/ https://orionfold.com/story/why-i-folded-orionfold/ Tue, 02 Jun 2026 00:00:00 GMT After nearly nine years at Amazon, I left to build Orionfold solo. The twenty-five-year story behind that choice, and what the name means. By Manav Sehgal. Founder AI Native Building in public https://orionfold.com/story/the-cockpit-for-my-models/ https://orionfold.com/story/the-cockpit-for-my-models/ Fri, 29 May 2026 00:00:00 GMT I had a shelf of models on one desktop and no way to drive them. In fifteen hours I built a cockpit to run, compare, and score them, all on that desk. Tools Local AI Agentic coding https://orionfold.com/story/teaching-a-small-model-my-field/ https://orionfold.com/story/teaching-a-small-model-my-field/ Thu, 23 Apr 2026 00:00:00 GMT I taught a 3B model my own field with 231 question-answer pairs. It stopped refusing and started answering, in my voice. Small and tuned beat big and general. Fine-tuning Models Custom models https://orionfold.com/story/my-first-model-on-a-desktop/ https://orionfold.com/story/my-first-model-on-a-desktop/ Wed, 22 Apr 2026 00:00:00 GMT I ran my first model on a small computer on my desk. 52 milliseconds to the first word, no cloud, no per-use bill. It felt like a local function, not a service. DGX Spark Local AI Personal devices https://orionfold.com/story/access-first-models-second/ https://orionfold.com/story/access-first-models-second/ Tue, 21 Apr 2026 00:00:00 GMT On day one with my desktop AI machine I did not pick a model. I set up how I reach it. Models change every six months. Good access lasts for years. Local AI AI Native Pragmatism https://orionfold.com/story/a-spec-at-breakfast-an-app-by-lunch/ https://orionfold.com/story/a-spec-at-breakfast-an-app-by-lunch/ Thu, 04 Dec 2025 00:00:00 GMT With Kiro I turned three sentences into a 24,000-line app, about four hours of my time. The spec did the heavy lifting, and speed became a skill. Spec-driven Vibe coding Building in public https://orionfold.com/story/one-agent-three-faces/ https://orionfold.com/story/one-agent-three-faces/ Mon, 01 Sep 2025 00:00:00 GMT The same AI agent should meet a developer in the terminal, a manager in a dashboard, and a phone user by voice. One brain, many doors. Agents Personalized for one AI Native https://orionfold.com/story/keeping-my-data-in-the-room/ https://orionfold.com/story/keeping-my-data-in-the-room/ Mon, 25 Aug 2025 00:00:00 GMT I built an analyst AI that never sends my work to the cloud. Privacy is not a setting you flip at the end. It is a choice you make in the design. Privacy Local AI AI Native https://orionfold.com/story/vibe-coding-is-not-passive/ https://orionfold.com/story/vibe-coding-is-not-passive/ Fri, 22 Aug 2025 00:00:00 GMT After my AI-built weekend I thought coding got easy. Andrew Ng named what I felt. Vibe coding is not passive, it is engineering done fast. Vibe coding Agents AI for everyone https://orionfold.com/story/i-built-a-real-product-in-a-weekend/ https://orionfold.com/story/i-built-a-real-product-in-a-weekend/ Tue, 28 Jan 2025 00:00:00 GMT Over one holiday weekend I built Web Memo, a working browser tool, about 2,400 lines of code. AI wrote most of it. My sense of my own ceiling moved. Vibe coding Building in public https://orionfold.com/story/frontier-reasoning-at-a-tenth-of-the-cost/ https://orionfold.com/story/frontier-reasoning-at-a-tenth-of-the-cost/ Tue, 21 Jan 2025 00:00:00 GMT In early 2025 a model called DeepSeek did frontier-level reasoning for about 95% less. And it was small enough to run on my own machine. Open models Cost Reduce the gap https://orionfold.com/story/the-year-the-gap-closed/ https://orionfold.com/story/the-year-the-gap-closed/ Tue, 12 Nov 2024 00:00:00 GMT In late 2024 I watched the AI tool stack collapse into the model itself. The first sign that one person could build what used to take a team. Open models AI Native Cost