Lenny RAG MCP Server

{ "episode": { "guest": "Dan Shipper", "expertise_tags": [ "AI Operations", "AI-First Companies", "Product Development", "Content Strategy", "Claude Code", "Prompt Engineering", "Allocation Economy", "Generalist Leadership" ], "summary": "Dan Shipper, co-founder and CEO of Every, discusses building a company at the bleeding edge of AI adoption. Every is a 15-person company operating a daily AI-focused newsletter with 100,000 subscribers, building multiple AI-powered products (Cora, Sparkle, Spiral), and running a consulting arm helping enterprises adopt AI. Dan shares tactics for operating efficiently with AI, including hiring a dedicated AI operations lead, using Claude Code for no-code product development, and building a generalist team. He argues that AI will reshape jobs rather than eliminate them, introduces the concept of the 'allocation economy' where management skills become universally valuable, and provides insights on why some companies successfully adopt AI while others fail.", "key_frameworks": [ "Compounding Engineering", "Allocation Economy", "Context Engineering", "AI-First Company Operations", "Generalist vs. Specialist Skills", "CEO Usage as Adoption Predictor", "Vibe Checks for AI Evaluation" ] }, "topics": [ { "id": "topic_1", "title": "Hot Takes on AI and Job Displacement", "summary": "Dan presents contrarian views on AI's impact, arguing that AI may actually reshore American jobs by making services affordable for small companies and enabling workers to serve more customers. He challenges narratives about entry-level jobs disappearing and discusses how young people with ChatGPT can accelerate learning faster than previous generations.", "timestamp_start": "00:00:20", "timestamp_end": "00:07:10", "line_start": 10, "line_end": 59 }, { "id": "topic_2", "title": "Claude Code: Underrated Tool for Non-Technical People", "summary": "Dan highlights that Claude Code is significantly underutilized by non-technical people despite being incredibly powerful for text processing tasks. He demonstrates use cases like analyzing meeting notes, extracting character descriptions from literature, and comparing translations.", "timestamp_start": "00:07:10", "timestamp_end": "00:13:58", "line_start": 61, "line_end": 139 }, { "id": "topic_3", "title": "Defining AGI Through the Leash Metaphor", "summary": "Dan proposes a novel framework for defining AGI based on how long of a 'leash' you can give an AI to work autonomously. He compares this to child development psychology, where growing up involves gradually increasing periods of independence. AGI is achieved when it becomes economically profitable to run agents indefinitely without human intervention.", "timestamp_start": "00:14:38", "timestamp_end": "00:18:29", "line_start": 148, "line_end": 184 }, { "id": "topic_4", "title": "Technology Adoption and Skill Trade-offs", "summary": "Dan addresses misconceptions about AI eroding human capabilities by drawing parallels to historical technology adoption. He argues that like writing replaced memorization but enabled knowledge distribution, AI replaces certain skills while enhancing others. The key is using AI correctly to amplify human potential rather than diminish it.", "timestamp_start": "00:18:48", "timestamp_end": "00:21:27", "line_start": 193, "line_end": 207 }, { "id": "topic_5", "title": "Context Engineering and Model Benchmarking", "summary": "Dan discusses his finding that frontier models struggle to predict his meeting statements without proper context engineering. He introduces the concept of context engineering as critical to AI performance, building on his earlier work on knowledge orchestration. The CEO Benchmark demonstrates that getting the right context to models is at least 50% of performance.", "timestamp_start": "00:21:55", "timestamp_end": "00:24:18", "line_start": 209, "line_end": 236 }, { "id": "topic_6", "title": "Every: Structure and Core Business Model", "summary": "Dan explains Every's unique business shape: a daily newsletter (100,000 subscribers), a suite of AI-powered apps (Cora, Sparkle, Spiral), and a consulting arm. The company produces 'vibe checks' of new AI models and builds products by identifying expensive services that AI makes affordable.", "timestamp_start": "00:26:04", "timestamp_end": "00:29:47", "line_start": 241, "line_end": 308 }, { "id": "topic_7", "title": "AI Operations Role and Team Automation", "summary": "Dan describes hiring Katie Parrott as a dedicated AI operations lead who builds prompts and workflows to automate repetitive tasks across the company. This role identifies efficiency opportunities without burdening individual team members. The AI operations function bleeds into every department and enables the small 15-person team to scale.", "timestamp_start": "00:30:10", "timestamp_end": "00:34:14", "line_start": 322, "line_end": 391 }, { "id": "topic_8", "title": "AI Tool Stack and Model Selection", "summary": "Dan details his personal AI tool stack: o3/ChatGPT as primary tool for writing and self-reflection with memory, Claude Opus for judging writing quality and self-improvement, Gemini for cost-effective app development, and Granola for meeting transcription. Different models have different strengths and 'personalities' that affect use cases.", "timestamp_start": "00:35:42", "timestamp_end": "00:40:17", "line_start": 413, "line_end": 461 }, { "id": "topic_9", "title": "Compounding Engineering and Efficiency Scaling", "summary": "The Cora team invented 'compounding engineering': making each unit of work easier than the last by building prompts and automations. Instead of writing PRDs manually each time, they built a prompt that converts rambling thoughts into structured PRDs. This philosophy extends to all their development practices.", "timestamp_start": "00:41:39", "timestamp_end": "00:43:37", "line_start": 484, "line_end": 509 }, { "id": "topic_10", "title": "Using Multiple AI Agents with Different Personalities", "summary": "The Cora team uses multiple agents simultaneously (Claude, Friday, Charlie from GitHub) because different models have different personalities and perspectives. Charlie's integration into GitHub makes it useful for PR reviews. This approach mirrors the concept of hiring a diverse team with complementary strengths.", "timestamp_start": "00:41:43", "timestamp_end": "00:46:35", "line_start": 485, "line_end": 581 }, { "id": "topic_11", "title": "Cora Team Backgrounds and Multidimensional Talent", "summary": "Kieran (VP Eng, composer, baker) and Nityesh (learned coding with ChatGPT) represent diverse backgrounds. Dan values generalists who can evaluate agents, design interfaces, and understand taste. Nityesh's AI-era learning demonstrates how young people can accelerate faster with AI mentorship than traditional entry-level paths.", "timestamp_start": "00:46:42", "timestamp_end": "00:50:11", "line_start": 589, "line_end": 621 }, { "id": "topic_12", "title": "No-Code Development and Product Teams Not Writing Code", "summary": "Every's product teams use Claude Code and other AI agents instead of manually writing code. The workflow involves refining specifications using prompt libraries, reviewing code, and testing output. This represents a major shift from traditional development where engineers spend time in code.", "timestamp_start": "00:51:54", "timestamp_end": "00:54:08", "line_start": 637, "line_end": 651 }, { "id": "topic_13", "title": "When Non-Technical People Can Build Software", "summary": "Dan argues that while non-technical people can't yet build conventional SaaS applications, they can build custom GPTs, browser skills, and other software-as-content products. The timeline for fully non-technical software development is still far off, but intermediate forms are already viable.", "timestamp_start": "00:55:54", "timestamp_end": "00:57:17", "line_start": 664, "line_end": 677 }, { "id": "topic_14", "title": "Product Incubation Strategy: From Expensive Services to Apps", "summary": "Every identifies expensive services (chief of staff, ghost writer, lawyer, organizer) that AI makes affordable and accessible. They prototype with general tools, then unbundle successful use cases into dedicated products. This approach is enabled by AI making historically expensive services cheap.", "timestamp_start": "00:57:38", "timestamp_end": "01:01:45", "line_start": 685, "line_end": 699 }, { "id": "topic_15", "title": "Fundraising Philosophy and 'Sip Seed' Strategy", "summary": "Dan raised $700K pre-seed and $2M in a novel 'sip seed' round (draw down on demand). This structure maintains optionality and creative freedom while providing a safety net. He argues that AI enables building with much less capital than traditional VC models require.", "timestamp_start": "01:02:25", "timestamp_end": "01:08:41", "line_start": 709, "line_end": 746 }, { "id": "topic_16", "title": "Consulting Arm: Teaching Companies to Be AI-First", "summary": "Every's consulting business (~$1M last year, likely double this year) helps enterprise companies adopt AI. They research organizations, identify automation opportunities, build dashboards showing team readiness, and provide customized training. The consulting arm reveals what large companies are struggling with.", "timestamp_start": "01:09:18", "timestamp_end": "01:12:00", "line_start": 750, "line_end": 769 }, { "id": "topic_17", "title": "CEO Usage as #1 Predictor of AI Adoption Success", "summary": "The single strongest predictor of whether a company successfully adopts AI is whether the CEO personally uses ChatGPT or Claude daily. CEO adoption drives organizational excitement, sets realistic expectations, and enables leaders to guide rather than doubt the technology.", "timestamp_start": "01:12:00", "timestamp_end": "01:13:20", "line_start": 769, "line_end": 775 }, { "id": "topic_18", "title": "Tactics for Driving Company-Wide AI Adoption", "summary": "Successful AI adoption requires: CEO sends memo stating 'We're AI-first' (ideally written with ChatGPT), weekly meetings where employees share prompts, weekly stats email showing company ChatGPT usage and contributors, and recognition of early adopters. These tactics create awareness and transfer learning across the organization.", "timestamp_start": "01:13:20", "timestamp_end": "01:15:34", "line_start": 773, "line_end": 803 }, { "id": "topic_19", "title": "The Allocation Economy and Future of Work", "summary": "Dan's allocation economy thesis posits that as AI makes intelligence cheap, management becomes the valuable skill. The 8% of the workforce that manages will grow because managing AI agents becomes economically viable. Entry-level workers must learn management alongside their craft, creating new career paths.", "timestamp_start": "00:50:33", "timestamp_end": "00:51:43", "line_start": 625, "line_end": 634 }, { "id": "topic_20", "title": "Generalists vs. Specialists in the AI Era", "summary": "AI enables and rewards generalists by providing expert knowledge on demand. Dan argues that Ancient Athens' model of generalist citizens gave way to specialists as civilization scaled, but AI inverts this by letting individuals span many domains. Smaller organizations with generalists may become more prevalent.", "timestamp_start": "01:20:20", "timestamp_end": "01:23:44", "line_start": 841, "line_end": 864 }, { "id": "topic_21", "title": "Writing as the Core of Every's Identity", "summary": "Dan initially stopped writing to focus on building Every, but discovered the business flattened without his writing. By returning to writing as the core and asking ChatGPT for examples of writer-builders, he realized this model works (Joel Spolsky, Jason Fried, Bill Simmons). The unique business shape emerges from leaning into authentic desires.", "timestamp_start": "01:29:23", "timestamp_end": "01:33:08", "line_start": 973, "line_end": 996 } ], "insights": [ { "id": "insight_1", "text": "AI may reshore American jobs by making expensive services (in-house counsel, call centers) affordable for small companies and allowing workers to serve many more customers from the US at lower cost than outsourcing.", "context": "Response to question about AI's effect on employment", "topic_id": "topic_1", "line_start": 50, "line_end": 56 }, { "id": "insight_2", "text": "When a kid with ChatGPT sits down with mentorship, they make a year's worth of progress in two months because they can record instructions into prompts and never make the same mistake twice.", "context": "Example of Alex Duffy's acceleration through AI-assisted learning", "topic_id": "topic_1", "line_start": 617, "line_end": 618 }, { "id": "insight_3", "text": "Claude Code is incredibly underrated for non-technical people because they can give it local files and ask it to process large amounts of text autonomously without uploading to chat interfaces.", "context": "Hot take about underutilized tools", "topic_id": "topic_2", "line_start": 61, "line_end": 71 }, { "id": "insight_4", "text": "The future of AI interfaces is moving away from showing users the underlying work—instead of seeing the code or intermediate steps, you'll just delegate and let the agent return results.", "context": "Discussion of interface evolution", "topic_id": "topic_2", "line_start": 128, "line_end": 131 }, { "id": "insight_5", "text": "A good definition of AGI is when it becomes economically profitable to run agents indefinitely without human intervention, similar to how developing independence is the mark of human maturation.", "context": "Novel framework for understanding AI advancement", "topic_id": "topic_3", "line_start": 167, "line_end": 176 }, { "id": "insight_6", "text": "Using AI is a skill. Studies showing AI outperforms doctors only demonstrate what happens when you compare specialized AI to untrained doctors; doctors plus AI trained in AI usage will likely continue to perform better.", "context": "Critique of oversimplified AI comparison studies", "topic_id": "topic_4", "line_start": 197, "line_end": 204 }, { "id": "insight_7", "text": "Throughout history, technology has always required trading certain skills for others—writing replaced memorization but enabled knowledge distribution. The key is whether the trade is worthwhile.", "context": "Historical perspective on technology adoption", "topic_id": "topic_4", "line_start": 200, "line_end": 203 }, { "id": "insight_8", "text": "Context engineering (getting the right context to the model at the right time) accounts for at least 50% of AI performance, making it a critical and underestimated problem.", "context": "Discussion of why frontier models struggle without proper context", "topic_id": "topic_5", "line_start": 218, "line_end": 221 }, { "id": "insight_9", "text": "The best technology writing comes from people actually using and building with the technology, which is why vibe checks matter more than benchmarks for understanding real-world usability.", "context": "Philosophy behind Every's vibe checks", "topic_id": "topic_6", "line_start": 278, "line_end": 281 }, { "id": "insight_10", "text": "Having a dedicated AI operations lead solves the problem where busy team members choose to do work the old way rather than spend time setting up AI automations.", "context": "Explanation of why the AI operations role is critical", "topic_id": "topic_7", "line_start": 335, "line_end": 341 }, { "id": "insight_11", "text": "Claude Opus 4 can judge whether writing is good (instead of always giving B+/A-/A grades), which enables new use cases where you need an AI to evaluate other AI's output before returning to humans.", "context": "Unique capability of Claude Opus that unlocked Spiral product development", "topic_id": "topic_8", "line_start": 443, "line_end": 452 }, { "id": "insight_12", "text": "Different AI models have distinct personalities and perspectives—ChatGPT feels terse and professional while Claude has a different style—making it valuable to use multiple agents simultaneously.", "context": "Discussion of why Cora team uses multiple AI agents", "topic_id": "topic_10", "line_start": 560, "line_end": 563 }, { "id": "insight_13", "text": "Compounding engineering means making each unit of work easier than the last by building automations and prompts that reduce effort for future similar tasks.", "context": "Core principle of how Cora team scales without more engineers", "topic_id": "topic_9", "line_start": 497, "line_end": 506 }, { "id": "insight_14", "text": "Entry-level workers learning in the AI era are accelerating faster than ever because they can record feedback into prompts and avoid repeating mistakes, creating a different entry-level experience.", "context": "Discussion of how ChatGPT changes early career trajectories", "topic_id": "topic_11", "line_start": 614, "line_end": 620 }, { "id": "insight_15", "text": "For a long time, it will remain valuable to understand how code works even if you don't write it, just as understanding C remains useful for Python developers even though you rarely write C.", "context": "Important caveat about staying with Every's no-code approach", "topic_id": "topic_12", "line_start": 650, "line_end": 659 }, { "id": "insight_16", "text": "Software is becoming content—there will be forms of software that look different from traditional SaaS but can be built and run as businesses by non-technical people.", "context": "Vision for what non-technical people can build soon", "topic_id": "topic_13", "line_start": 674, "line_end": 674 }, { "id": "insight_17", "text": "Identifying products starts with noticing what expensive services you wish you could afford, then prototyping with general-purpose AI tools, then unbundling into dedicated apps if successful.", "context": "Every's product discovery and incubation approach", "topic_id": "topic_14", "line_start": 687, "line_end": 698 }, { "id": "insight_18", "text": "Having the right investor alignment matters more than the amount raised—investors who understand and believe in your vision enable more creative freedom than those pushing for growth-at-all-costs.", "context": "Philosophy behind working with Reid Hoffman", "topic_id": "topic_15", "line_start": 725, "line_end": 726 }, { "id": "insight_19", "text": "AI enables building with dramatically less capital because tools that were technically impossible three years ago are now accessible and one engineer with AI can accomplish what previously required 20.", "context": "How Cora cost only $300K all-in to build including salaries", "topic_id": "topic_15", "line_start": 734, "line_end": 740 }, { "id": "insight_20", "text": "The single strongest predictor of whether a company will successfully adopt AI is whether the CEO personally uses ChatGPT or Claude daily.", "context": "Key finding from consulting work with large companies", "topic_id": "topic_17", "line_start": 770, "line_end": 770 }, { "id": "insight_21", "text": "Most people (80%) will adopt AI if shown exactly how to use it for their specific job, unlike early adopters (10%) who explore on their own or resisters (10%) who won't engage.", "context": "Distribution of adoption types in organizations", "topic_id": "topic_18", "line_start": 764, "line_end": 765 }, { "id": "insight_22", "text": "In the allocation economy, management skills become valuable for everyone because AI makes it economically viable to delegate to agents, and managing agents requires the same skills as managing people.", "context": "Vision for how work changes as intelligence becomes cheap", "topic_id": "topic_19", "line_start": 626, "line_end": 627 }, { "id": "insight_23", "text": "Critical management skills that become valuable are: evaluating talent, vision setting, taste development, knowing when to get into details, and when to delegate.", "context": "Skills that transfer from managing people to managing AI", "topic_id": "topic_19", "line_start": 833, "line_end": 833 }, { "id": "insight_24", "text": "Having AI in your pocket functioning like 10,000 PhDs empowers people to stay generalists longer and jump between domains, potentially reversing the specialization trend of the industrial age.", "context": "How AI enables and rewards generalist skills", "topic_id": "topic_20", "line_start": 860, "line_end": 864 }, { "id": "insight_25", "text": "The shape of a business should emerge from authenticity and leaning into what genuinely excites the founder rather than following Silicon Valley playbooks.", "context": "Dan's experience of stopping writing then realizing it was core to Everything's success", "topic_id": "topic_21", "line_start": 992, "line_end": 995 }, { "id": "insight_26", "text": "Media businesses follow different scaling patterns than tech startups—if founder-created content is the initial product-market fit, hiring to replace the founder's output is likely to destroy it.", "context": "Lesson from stopping his writing at Every", "topic_id": "topic_21", "line_start": 980, "line_end": 980 }, { "id": "insight_27", "text": "GPT wrappers are far more valuable than conventional wisdom suggests because they apply general models to specific use cases with better UX and distribution than the underlying model alone.", "context": "Defense of GPT wrapper business model", "topic_id": "topic_14", "line_start": 704, "line_end": 704 }, { "id": "insight_28", "text": "Asking AI to help you understand yourself (e.g., 'What's my life motto?') with memory creates a feedback loop where it inspires rather than directly answers, extending rather than replacing your thinking.", "context": "How Dan uses ChatGPT with memory for self-reflection", "topic_id": "topic_8", "line_start": 956, "line_end": 962 } ], "examples": [ { "id": "example_1", "explicit_text": "At Every, one of the big parts of Every is we have a daily newsletter. And I'm spending a lot of time giving feedback on headlines, or giving feedback on, 'How do you write an intro,' or 'Is this idea any good'", "inferred_identity": "Every's daily newsletter operations", "confidence": 1.0, "tags": [ "Every", "Newsletter", "Content", "Editorial", "Daily publication", "AI at work" ], "lesson": "Codifying founder feedback into prompts that push taste standards to the entire team, enabling writers to meet quality bars without direct feedback loops.", "topic_id": "topic_5", "line_start": 233, "line_end": 233 }, { "id": "example_2", "explicit_text": "We have this guy Alex Duffy who works with us, he writes for Context Window and he just launched, we taught AIs how to play diplomacy with each other, which is really cool.", "inferred_identity": "Alex Duffy, Every employee, Context Window writer", "confidence": 0.95, "tags": [ "Every", "Context Window", "AI research", "Diplomacy simulation", "Accelerated learning", "Entry-level growth" ], "lesson": "Entry-level talent can accelerate dramatically when mentored with AI tools—Alex went from struggling with writing to shipping complex AI research in months.", "topic_id": "topic_11", "line_start": 614, "line_end": 620 }, { "id": "example_3", "explicit_text": "Cora, which is Kieran and Nityesh, basically... That's the team, yeah. Well, with Cora, it's Kieran, Nityesh, and 15 Claude Code instances", "inferred_identity": "Kieran and Nityesh, Every engineers building Cora", "confidence": 1.0, "tags": [ "Every", "Cora", "Email AI", "Claude Code", "Small teams", "AI-augmented development" ], "lesson": "Two engineers with AI agents can build and ship sophisticated products like an AI email chief of staff with thousands of active beta users.", "topic_id": "topic_9", "line_start": 485, "line_end": 509 }, { "id": "example_4", "explicit_text": "Nityesh... he only started learning to code when ChatGPT came out. He had wanted to learn to code forever, and he's only known how to code in an AI era.", "inferred_identity": "Nityesh, Every engineer", "confidence": 0.9, "tags": [ "Every", "Cora", "ChatGPT", "Entry-level", "Learning acceleration", "AI-native programmer" ], "lesson": "People learning to code in the AI era are accelerating far faster than those who had to learn from books and Stack Overflow, creating a new type of entry-level talent.", "topic_id": "topic_11", "line_start": 605, "line_end": 620 }, { "id": "example_5", "explicit_text": "Kieran's got this crazy background, where he was previously VP Eng at a startup, so was effectively the CTO of a startup, or maybe two startups, and was one of the founders. But before that, he was a composer, a professional composer. And before that, he was a baker.", "inferred_identity": "Kieran, Every engineer, VP Eng at previous startup", "confidence": 0.95, "tags": [ "Every", "Cora", "Multidimensional talent", "VP Engineering", "Composer", "Generalist skills", "Rails developer" ], "lesson": "Multidimensional backgrounds create better taste and broader capabilities—Kieran's music and craft backgrounds inform product decisions beyond pure engineering.", "topic_id": "topic_11", "line_start": 596, "line_end": 602 }, { "id": "example_6", "explicit_text": "Katie Parrott... she actually does a lot of ghostwriting for us... She still does that, but she also spends a lot of time doing the AI operation stuff... she worked at Animalz, which is a content marketing agency, one of the top content marketing agencies.", "inferred_identity": "Katie Parrott, Every AI Operations Lead, former Animalz", "confidence": 0.95, "tags": [ "Every", "AI Operations", "Content marketing", "Animalz", "Process optimization", "Automation" ], "lesson": "The ideal AI operations person combines process orientation, writing craft, and genuine excitement about AI tinkering—this combination is key to building prompts people actually use.", "topic_id": "topic_7", "line_start": 383, "line_end": 389 }, { "id": "example_7", "explicit_text": "Katie is so good is because she's incredibly good at that kind of process stuff or thinking about that, but she's also a great writer and she's also just incredibly excited about AI. She just wants to tinker and wants to use it.", "inferred_identity": "Katie Parrott's qualities", "confidence": 0.95, "tags": [ "Every", "AI Operations", "Process-oriented", "Writing", "Tinkering mindset", "Automation culture" ], "lesson": "The core requirement for an AI operations lead is the desire to tinker and build combined with understanding of the craft they're optimizing for.", "topic_id": "topic_7", "line_start": 386, "line_end": 389 }, { "id": "example_8", "explicit_text": "Nityesh, who's one of the engineers on Cora, built a Claude Code command that just uses that prompt, and checks through the entire code base for all the copy edits, and then creates a pull request on GitHub, and then sends the pull request to Kate.", "inferred_identity": "Nityesh creating Claude Code automation for Cora", "confidence": 0.95, "tags": [ "Every", "Cora", "Claude Code", "Automation", "GitHub", "Copy editing", "Style consistency" ], "lesson": "Translate editorial standards into engineering-compatible formats via Claude Code commands to scale quality control across products.", "topic_id": "topic_7", "line_start": 401, "line_end": 404 }, { "id": "example_9", "explicit_text": "We work with a hedge fund called Walleye, which I had the founder on my podcast, AI and I, a few weeks ago, their gigantic $10 billion hedge fund.", "inferred_identity": "Walleye hedge fund, Every consulting client", "confidence": 0.9, "tags": [ "Every", "Consulting", "Walleye", "Hedge fund", "$10 billion", "AI adoption", "Enterprise" ], "lesson": "Successful large-scale AI adoption requires CEO leadership, weekly prompt-sharing meetings, and public recognition of AI contributions to drive organizational momentum.", "topic_id": "topic_18", "line_start": 773, "line_end": 783 }, { "id": "example_10", "explicit_text": "I downloaded War and Peace to my computer, which you can do because it's public domain. And then I had Claude read the first three chapters of War and Peace and pull out all of those descriptions, and then make a guide for itself for how to do character descriptions like Tolstoy.", "inferred_identity": "Dan Shipper using Claude Code with War and Peace", "confidence": 1.0, "tags": [ "Claude Code", "Literature analysis", "Writing improvement", "Tolstoy", "Character analysis", "AI-assisted learning" ], "lesson": "Claude Code can process large complex documents to extract patterns and style guides that inform creative writing, extending literary analysis beyond what chat interfaces enable.", "topic_id": "topic_2", "line_start": 103, "line_end": 107 }, { "id": "example_11", "explicit_text": "I built this thing over a weekend a month ago that was, '0.3, can it predict what I'm going to say in a meeting?' That's a benchmark. It's the CEO benchmark.", "inferred_identity": "Dan Shipper's CEO benchmark experiment", "confidence": 1.0, "tags": [ "Benchmark", "CEO", "Meeting prediction", "Model evaluation", "Context engineering", "Personalization" ], "lesson": "Using personal data (meeting transcripts) as a benchmark reveals how poor general models are without context engineering, highlighting the true difficulty of personalization.", "topic_id": "topic_5", "line_start": 212, "line_end": 224 }, { "id": "example_12", "explicit_text": "I had it download a Russian version of War and Peace and the English version, and then start comparing different scenes that I love to tell me about things that I might've missed in the translations", "inferred_identity": "Dan Shipper's War and Peace comparison project", "confidence": 1.0, "tags": [ "Claude Code", "Language analysis", "Translation comparison", "Russian literature", "Tolstoy", "Deep learning" ], "lesson": "Claude Code enables deep textual analysis that would be impractical manually—comparing translations across languages for nuance.", "topic_id": "topic_2", "line_start": 106, "line_end": 107 }, { "id": "example_13", "explicit_text": "Cora, I think all in to build Cora, we've spent maybe 300K, Maybe. That's crazy because... Includes salaries. Yeah.", "inferred_identity": "Cora development budget at Every", "confidence": 0.95, "tags": [ "Every", "Cora", "Product cost", "AI efficiency", "Budget", "Small team scaling", "$300K" ], "lesson": "Building a sophisticated AI product with two engineers costs ~$300K including salaries—a dramatic reduction from the 20-engineer teams that would have been required pre-AI.", "topic_id": "topic_15", "line_start": 734, "line_end": 740 }, { "id": "example_14", "explicit_text": "This product was not even technically possible even if you had billions of dollars three years ago. Not possible because you can't do email summarizing and automatic responses and all that kind of stuff without GPT.", "inferred_identity": "Cora product capabilities timeline", "confidence": 0.95, "tags": [ "Cora", "Email AI", "GPT", "AI capability evolution", "Product feasibility", "Technology timing" ], "lesson": "Many AI products are only newly possible because of recent capability improvements—the business model depends on technological breakthroughs that make old problems solvable.", "topic_id": "topic_15", "line_start": 734, "line_end": 740 }, { "id": "example_15", "explicit_text": "Base44... he's been around for six months, the company. For the last three months, he hasn't touched a single line of front-end code, all Cursor and other tools he's using.", "inferred_identity": "Base44 founder (sold to Wix for $80M), using no-code approach", "confidence": 0.95, "tags": [ "Base44", "Wix", "$80M acquisition", "Cursor", "No-code development", "Founder building", "AI tools" ], "lesson": "Successful founders can now avoid writing code entirely using AI tools like Cursor, representing a fundamental shift in what's required to build products.", "topic_id": "topic_12", "line_start": 140, "line_end": 140 }, { "id": "example_16", "explicit_text": "we originally incubated Lex, which is an AI document writer, which we spun out into its own company, and my Every co-founder runs that.", "inferred_identity": "Lex, Every spinout company", "confidence": 0.95, "tags": [ "Every", "Lex", "AI document writer", "Spinout", "Product incubation", "Co-founder venture" ], "lesson": "Every's incubation model can spin out successful products into standalone companies run by co-founders, creating portfolio approach to building.", "topic_id": "topic_6", "line_start": 290, "line_end": 290 }, { "id": "example_17", "explicit_text": "we started to codify all of that into prompts that basically... It's not the same as mimicking me. It can't exactly say exactly what I'm going to say in a meeting, but it pushes my taste out to the edge so that writers who are not able to talk to me, by the time I see it, they've already talked to some simulation of a simulation of me.", "inferred_identity": "Every's feedback codification process", "confidence": 0.95, "tags": [ "Every", "Editorial", "Prompts", "Feedback automation", "Taste distribution", "Scaling leadership" ], "lesson": "Codifying founder feedback into prompts scales editorial taste without requiring founder time on every piece of content.", "topic_id": "topic_5", "line_start": 236, "line_end": 236 }, { "id": "example_18", "explicit_text": "Joel Spolsky, who built Trello and Stack Overflow... Jason Fried who I've known for a long time... Sam Harris who's got a great podcast, and he's got a gigantic meditation app... Bill Simmons... incredible podcaster and also built The Ringer, sold to Spotify for a couple hundred million bucks.", "inferred_identity": "Writer-builders: Spolsky, Fried, Harris, Simmons", "confidence": 1.0, "tags": [ "Writer-founders", "Joel Spolsky", "Jason Fried", "Sam Harris", "Bill Simmons", "Trello", "Stack Overflow", "The Ringer", "Content business", "Spotify" ], "lesson": "There are successful precedents for writer-founders building substantial businesses—this model isn't unique but requires leaning into the writing as core rather than auxiliary.", "topic_id": "topic_21", "line_start": 983, "line_end": 984 }, { "id": "example_19", "explicit_text": "For example, Plato is famously very skeptical of writing because he thought it would harm your memory. And it did.", "inferred_identity": "Plato's concerns about writing technology", "confidence": 1.0, "tags": [ "Plato", "Writing", "Memory", "Technology adoption", "Historical skepticism", "Skill trade-offs" ], "lesson": "Historical technology adoption always involved real skill trade-offs; Plato was right that writing harmed memory, but the benefits vastly outweighed costs.", "topic_id": "topic_4", "line_start": 200, "line_end": 200 }, { "id": "example_20", "explicit_text": "when I was in middle school learning to code, the new hot thing was scripting languages, which is Python and JavaScript. But if you were a real programmer, you would understand the language underlying Python and JavaScript, which is written in C.", "inferred_identity": "Dan Shipper's programming education journey", "confidence": 0.95, "tags": [ "Programming history", "C programming", "Python", "JavaScript", "Scripting languages", "Skill stacking", "Technical depth" ], "lesson": "Each technological layer requires understanding the layer below it, but this requirement gradually decreases over time as abstractions mature—same will be true for AI tools.", "topic_id": "topic_12", "line_start": 656, "line_end": 659 }, { "id": "example_21", "explicit_text": "DIA has these things called skills, which are effectively little AI apps that you can run in the browser. You can prompt them and they run on the web page and do work for you. A non-technical person can build that", "inferred_identity": "DIA (browser company) skills feature", "confidence": 0.9, "tags": [ "DIA", "Browser", "AI skills", "No-code", "Custom GPT alternative", "Web automation" ], "lesson": "Browser-based AI skills enable non-technical people to build functional software-like products today, not years in the future.", "topic_id": "topic_13", "line_start": 671, "line_end": 671 }, { "id": "example_22", "explicit_text": "we will unbundle it into its own separate thing that becomes an app. And I think what's really special about this time is the entire game board has been totally reset in terms of things you can build. Where five years ago it was like you're going to build another Notes app.", "inferred_identity": "Every's product philosophy and market opportunity", "confidence": 0.95, "tags": [ "Every", "Product development", "Market opportunity", "AI-first products", "Unbundling", "New categories" ], "lesson": "The AI era creates entirely new product categories rather than incremental improvements on existing ones like notes apps.", "topic_id": "topic_14", "line_start": 692, "line_end": 692 }, { "id": "example_23", "explicit_text": "Tobi from Spotify coined this term called 'context engineering,' which is getting the context to the model, the right context at the right time", "inferred_identity": "Tobi (Spotify CEO) context engineering concept", "confidence": 0.9, "tags": [ "Tobi Lutke", "Spotify", "Context engineering", "AI optimization", "Model performance", "Terminology" ], "lesson": "Context engineering is a term now being widely adopted to describe the critical challenge of getting the right information to AI models at the right time.", "topic_id": "topic_5", "line_start": 218, "line_end": 218 }, { "id": "example_24", "explicit_text": "The tailor down the street from me doesn't accept credit cards. Credit cards have been around for a long time, so it takes a long time for technology like this to be adopted even in the best case.", "inferred_identity": "Dan Shipper's local tailor in Brooklyn", "confidence": 0.9, "tags": [ "Brooklyn", "Tailor", "Technology adoption", "Credit cards", "Slow adoption", "Real-world friction" ], "lesson": "Even simple, proven technologies like credit cards take decades to fully adopt, suggesting AI transformation will take far longer than hype suggests.", "topic_id": "topic_4", "line_start": 209, "line_end": 209 }, { "id": "example_25", "explicit_text": "I asked ChatGPT, 'I'm going on Lenny's podcast. What would my life motto be?' and it said, 'Your life motto is witness deeply, build bravely.'", "inferred_identity": "Dan Shipper asking ChatGPT to define his life motto", "confidence": 1.0, "tags": [ "ChatGPT", "Memory feature", "Self-reflection", "Life philosophy", "AI introspection", "Personal branding" ], "lesson": "ChatGPT with memory can help you articulate your own philosophy by reflecting back patterns in what you've shared, functioning as a mirror rather than oracle.", "topic_id": "topic_8", "line_start": 935, "line_end": 938 }, { "id": "example_26", "explicit_text": "'Do things worth writing about, and write things worth reading.' Seems like a pretty good summation... Pliny the Younger said", "inferred_identity": "Pliny the Younger's writing philosophy", "confidence": 1.0, "tags": [ "Pliny the Younger", "Writing philosophy", "Ancient Rome", "Content creation", "Founding principle", "Editorial standard" ], "lesson": "Ancient writing philosophy remains relevant for modern creators—focus on substance and craft rather than volume.", "topic_id": "topic_21", "line_start": 938, "line_end": 938 } ] }