# The Factory That Dreams: 39 AI Agents, No Framework - Rushabh Doshi, Machinecraft AI Engineer · 2026-07-11 Rushabh Doshi, CEO of Machinecraft, explains how his 100-person factory built Ira, a 36-agent AI OS running its go-to-market and memory without a framework. The system uses 36 specialist agents, each with one job, orchestrated by Athena, handling outbound emails, quoting, leads, and replies from one Cursor tab. Built for $30,000 versus $230,000 agency quote, it runs on a few thousand monthly with no training. Doshi says the secret is well-organized memory, vectors, graphs, and biological-inspired architecture including senses, digestion, memory layers, a nightly dream cycle, and immune system for fact-checking. The system is grounded in a SOUL.md constitution based on Jain philosophy, and ForkMyBrain.org helps others build their own forkable brain by mapping their business from the inside. ## Questions this episode answers ### What is Ira's nightly 'dream cycle' and how does it improve Machinecraft's AI system? Every night, Ira runs a sleep cycle where it replays the day's interactions, consolidates useful information, hunts for contradictions, forgets stale data, and turns the day's work into reusable skills, according to Rushabh Doshi. By morning, a dream report shows what was consolidated, let go, or figured out, making the system "smarter overnight." [7:19](https://aie.addtry.com/2b45ebaa-a825-4db3-898e-06b2abbf149e?t=439000) ### What philosophical principles guide Machinecraft's AI agents, and how do they act as guardrails? Rushabh Doshi explains the agents follow a 'soul file' based on Jain philosophy: no single source has the whole truth, so cross-check before speaking; never state things absolutely, cite the document and date; do your own job; report the truth even when ugly; and nobody works alone. These principles serve as the agents' conscience and operational guardrails. [8:06](https://aie.addtry.com/2b45ebaa-a825-4db3-898e-06b2abbf149e?t=486000) ### How much did Machinecraft spend to build Ira compared to traditional agency quotes? An agency quoted $230,000 to build the system, but Rushabh Doshi says they built it for about $30,000, with monthly running costs of a couple thousand dollars. He emphasizes there was no training bill—the real expense was teaching the company to remember itself. [8:45](https://aie.addtry.com/2b45ebaa-a825-4db3-898e-06b2abbf149e?t=525000) ### Why did Rushabh Doshi choose a multi-agent architecture with 36 agents for Ira instead of a single large language model? Rushabh Doshi argues one prompt that does everything ends up doing nothing well. Instead, Ira uses 36 specialist agents—like Athena (orchestrator), Prometheus (sales), Plutus (pricing), Hephaestus (machine specs), Vera (fact-checking), and Memon (corrections)—each with one job. They hold meetings and argue to produce coherent answers. [4:07](https://aie.addtry.com/2b45ebaa-a825-4db3-898e-06b2abbf149e?t=247000) ## Key moments - **[0:00] The Problem** - [0:58] For three generations, Machinecraft’s entire company knowledge lived in exactly three human brains, risking catastrophic forgetting with every departure. - **[1:33] The Idea** - **[2:31] Data Diet** - [2:50] Machinecraft built its AI brain without training any model, relying on vector embeddings and graph relationships from internal documents. - **[3:28] Agent Design** - **[4:54] Nine Jobs** - [4:54] Ira operates Machinecraft’s entire front business with nine daily jobs: outbound emails, account briefs, quotations, and lead revival. - **[6:44] Layered Memory** - **[7:19] Dream Cycle** - [7:19] Ira’s nightly dream cycle consolidates learning, hunts contradictions, forgets stale data, and produces a morning report, making it smarter overnight. - **[8:06] Soul File** - **[8:45] BrainOS** - [8:45] Machinecraft built its AI system for $30,000, compared to a $230,000 agency quote, with monthly costs of a couple thousand dollars. ## Speakers - **Rushabh Doshi** (guest) ## Topics Agent Memory, Multi-Agent Orchestration, Agent Commerce ## Mentioned Google (company), Machinecraft (company), BrainOS (product), Cursor (product), Fork My Brain (product), Ira (product) ## Transcript ### The Problem **Rushabh Doshi** [0:01] Okay, I'm going to tell you a story about a factory that taught itself how to remember. Hi, I'm Rushabh. I run Machinecraft, a 100-people factory in India. No data science team, no ML budget, none of that. And somehow we ended up building a 36 AI agent that runs our entire go-to-market. I think that's still a little ridiculous. Let me show you how it happened, and why you can do the same thing. So here's the thing about our company: from the outside it looks like machines and metal. But the actual company, the part that matters, isn't the machines. It's the knowledge. Who the customer is, what we quoted them in 2019, why that one machine needed that weird custom tweak. And for 3 generations, all of that lived in exactly 3 brains. Initially my grandfather's, then my father's, and now mine. Which is a genuinely terrifying way to run a company when you sit with it. A lot of people have joined us, people have left us, the revolving door never stopped. And every single time someone walked out, a chunk of our brain walked out with them. We weren't scared of the competitors. We were scared of forgetting. Of waking up one day and realizing the whole company only existed inside two increasingly tired heads. So I had an idea. I'll be honest, it sounded insane first. But what if, instead of writing the knowledge down in some document nobody ever reads, what if we grew a brain that just... held it? Not a chatbot you poke at. A twin of the company. I didn't hire a sales team. ### The Idea **Rushabh Doshi** [1:49] I tried to build one. A quick detour, because you need to know how messy this is. We make thermoforming machines. They heat up a plastic sheet and shape it. Same core machine, but it ends up making hydroponic farm trays, spa bathtubs, EV car panels, medical casings, and even packaging. Seven totally different worlds, seven totally different buyers. So this brain couldn't just memorize a brochure. It had to know which universe a given customer lives in. Step 1 was almost boringly simple: feed it everything. And I mean everything. Years of quotes, drawings, payment schedules, timelines, email threads, hundreds of gigabytes of our own private history. Not the public internet. Our internet. And here's the plot twist, the part that surprises every engineer I tell this to. We never trained a model. ### Data Diet **Rushabh Doshi** [2:50] No GPUs humming in the basement, no fine-tuning. We just looked at all the history, chopped it into bite-sized chunks, and let off-shore models read it and pull out the facts. We stored the meaning of each chunk as vectors and relationships. Who's connected to what as a graph. The brain isn't a smarter model. It's actually a really, really well-organized memory. Now, this is where it gets a little weird in a good way. We stopped thinking of Ira as a software and started thinking of it as something we were raising. So we gave it a body modeled on biology. Senses to figure out who it's talking to. A gut to digest the documents into facts. A memory. A dream cycle. An immune system to fight off bad information. Why biology? ### Agent Design **Rushabh Doshi** [3:44] Well, because evolution already spent a billion years solving how do you stay coherent over time. We just copied the homework. Okay, so the big question: why 36 agents instead of one genius mega-prompt? Because—and you already know this if you've ever tried it—one prompt that's supposed to do everything ends up doing everything badly. So Ira isn't one mind. It's a pantheon. A whole cast of specialists. Each one has exactly one job. Athena runs the room. Prometheus owns the sale. Plutus does pricing. Hephaestus knows every machine spec cold. Vera fact-checks everything. And Memon, my favorite, guards corrections. So the second a human fixes something, it stays fixed forever. One agent, one job. It's a team, not a hero. And here's the cool part: they hold meetings. Athena pulls in specialists. They actually argue. And a single answer comes out the other side. ### Nine Jobs **Rushabh Doshi** [4:54] It's like having a boardroom that never sleeps, never gets tired, and somehow has no ego. So what does all this actually run? Honestly, the whole front business. Everything between a stranger exists somewhere, and now they're a customer. Nine concrete jobs every single day. Outbound emails that actually reference my real world. Account briefs built from cross-checked truths before a call. Quotations. A swipe-left/swipe-right mode for outreach. Reviving dead leads, which I call "blast from the blast." Inbound replies. And figuring out, before we waste an hour, whether a company is even a fit. Nine jobs, one operator who never sleeps. Where does all this live? One Cursor tab. That's genuinely it. You type, and Ira reaches out with a dozen hands. Searches the knowledge base, reads the inbox, drafts the email, builds the code, and then shows you before anything actually goes out. Under the hood, it's genuinely a real stack. Not a demo held together with a tape. Databases for vectors, for relationship graph, for the CRM. Three different model providers, each picked for the job it's actually best for. Tools for Google, for swallowing documents for every communication channel. Plus monitoring. So we can see what it's thinking. All of it, every capability exposed as 213 tools over one protocol. And the golden rule, the one we never break: Ira drafts human sentences. Now, memory. And this is the part where most AI quietly lies to you. Because a raw language model is basically a goldfish. Brilliant for about 30 seconds, and then you close the tab and forget you ever existed. So we engineered memory on purpose, in layers. Working memory, for the last few minutes. ### Layered Memory **Rushabh Doshi** [6:59] Pinned facts, about someone who is. Episodes. Whole conversations as little stories. Relationships, with warmth that grows from stranger to trusted. And a bouncer at the door. A salience gate that decides what's even worth remembering so the brain doesn't fill up with junk. ### Dream Cycle **Rushabh Doshi** [7:19] When two facts disagree, corrections win. Continuity without making things up. And then, I genuinely love this part. At night, it dreams. Every night, Ira runs a sleep cycle. It replays the day, locks in useful stuff, hunts for contradictions, gently forgets the stale junk, and turns the day's work into reusable skills. In the morning, there's a little dream report waiting for me to read. Here's what I consolidated. Here's what I let go of. Here's what I figured out while you were asleep. The thing literally gets smarter overnight. And here's the part I care about the most: every agent has a conscience. ### Soul File **Rushabh Doshi** [8:06] And it is emphatically not to be helpful, be harmless. It's a soul file, written from the principles of a Jain family business that's been doing this for the last 3 generations. Five old ideas turned into engineering rules. No single source has the whole truth. So cross-check before you speak. Never state things absolutely. Cite the document and the date. Do your own job, not someone else's. Report the truth, even when the truth is ugly. And nobody works alone. Ancient philosophy running as guardrails in production. Now let's talk money. Because this is the part that should make the whole industry a little uncomfortable. There was no training bill. Zero. The expensive part was never compute. It was teaching a company to remember itself. An agency quoted us 230 grand to build this. ### BrainOS **Rushabh Doshi** [9:02] We built it for around 30. That's cheaper than a nice watch. And it runs on a couple of thousand dollars a month. So here's the move. We pulled the whole architecture out and made it forkable. We call it BrainOS. It ships as an empty nervous system. The agents, the memory, the dream cycle, the soul file. All there, completely blank. You pour your own company's truth into it, and from inside out. Because here's the thing nobody can outsource for you. Only you can build your company's brain. We are a 100-people factory with no data scientists. If we can grow a brain, you can too. We're not selling ours to you. We're helping you build your own. ForkMyBrain.org. Go build something that remembers. Thank you. --- This library is powered by PodHood (https://podhood.com), the podcast website platform.