Lenny RAG MCP Server

{ "episode": { "guest": "Dhanji R. Prasanna", "expertise_tags": [ "AI/LLM adoption", "Engineering leadership", "Organizational transformation", "Open source", "Product strategy", "AI agents" ], "summary": "Dhanji R. Prasanna, CTO of Block (formerly Square), discusses Block's transformation into an AI-native company. He shares how the organization shifted from a GM structure to functional hierarchy, enabling deeper technical focus. Prasanna explains Goose, Block's open-source AI agent built on the Model Context Protocol, which delivers 8-10 hours of weekly productivity gains for engineering teams and 20-25% manual hour savings across the company. The conversation covers AI's impact on different engineer levels, the importance of organizational structure over code quality, lessons from failed products (Google Wave, Google+, Secret), and how AI is reshaping how teams work—from autonomous agents running overnight to exploring experimental multi-feature builds. Throughout, Prasanna emphasizes starting small, questioning base assumptions, and optimizing for company purpose over technological trends.", "key_frameworks": [ "Conway's Law", "GM vs functional organizational structure", "Model Context Protocol (MCP)", "Vibe coding and AI agents", "Code quality vs product success", "Controlled chaos in engineering culture", "80/20 rule for maintenance burden", "Start small philosophy", "Portfolio approach to optimization" ] }, "topics": [ { "id": "topic_1", "title": "AI productivity gains at Block", "summary": "Dhanji shares concrete metrics on productivity improvements from AI tools at Block, including 8-10 hours saved per week for AI-forward engineering teams using Goose, and 20-25% of manual hours saved across the entire company. He emphasizes this is just the baseline and value will continue improving.", "timestamp_start": "00:00:00", "timestamp_end": "00:20:18", "line_start": 1, "line_end": 173 }, { "id": "topic_2", "title": "What is Goose and how it works", "summary": "Dhanji explains Goose, Block's open-source AI agent that uses the Model Context Protocol (MCP) to connect to existing tools and systems. Goose can perform tasks from organizing photos to writing software, orchestrating across multiple systems like Snowflake, Tableau, and enterprise applications.", "timestamp_start": "00:21:49", "timestamp_end": "00:27:15", "line_start": 175, "line_end": 207 }, { "id": "topic_3", "title": "Extreme AI experimentation: engineer with Goose watching", "summary": "An engineer at Block has built an experimental system where Goose continuously watches his screen and listens to his voice, anticipating work needs. The agent proactively opens PRs for discussed features, reschedules meetings, and automates repetitive tasks without explicit prompts.", "timestamp_start": "00:28:41", "timestamp_end": "00:32:16", "line_start": 224, "line_end": 254 }, { "id": "topic_4", "title": "Future of AI-driven engineering and autonomous work", "summary": "Dhanji envisions engineering where AI agents work autonomously for hours or overnight, building multiple experimental features in parallel that engineers can evaluate in the morning. He discusses pushing from ping-pong vibe coding toward true autonomy, and reimagining workflows where entire apps get rebuilt from scratch with each release.", "timestamp_start": "00:32:32", "timestamp_end": "00:37:43", "line_start": 257, "line_end": 294 }, { "id": "topic_5", "title": "AI's limitations and human judgment", "summary": "Dhanji identifies areas where AI still underperforms humans: architecture decisions, complex design patterns, and portfolio-level business judgment. He emphasizes the need for human taste to keep AI grounded and prevent 'AI slop,' noting AI succeeds at 60% of ambitious feature requests with clear specifications.", "timestamp_start": "00:36:39", "timestamp_end": "00:40:11", "line_start": 286, "line_end": 309 }, { "id": "topic_6", "title": "Build vs buy decisions in the AI era", "summary": "Dhanji discusses how AI changes build-vs-buy calculus. While teams can now quickly build internal tools with Goose, he cautions against losing focus on core business purpose. The real productivity gains sometimes come from questioning whether processes need to exist at all, not just automating them.", "timestamp_start": "00:40:31", "timestamp_end": "00:43:04", "line_start": 313, "line_end": 327 }, { "id": "topic_7", "title": "Block's transformation from GM to functional structure", "summary": "Dhanji shares how Block shifted from a portfolio model (GM structure with independent units like Square, Cash App, Afterpay) to a functional organization where all engineers report to one head of engineering. This restructuring was crucial for becoming AI-native and aligning technical strategy across the company.", "timestamp_start": "00:07:53", "timestamp_end": "00:13:58", "line_start": 79, "line_end": 116 }, { "id": "topic_8", "title": "Getting company alignment on AI strategy", "summary": "Dhanji describes writing an AI manifesto to Jack Dorsey that convinced leadership to prioritize AI. This led to executive task forces, rebranding as a technology company, and initiating special projects, hack weeks, and IC summits to rebuild engineering culture around technology leadership.", "timestamp_start": "00:06:00", "timestamp_end": "00:08:55", "line_start": 64, "line_end": 86 }, { "id": "topic_9", "title": "Who benefits most from AI tools", "summary": "Contrary to expectations, the biggest beneficiaries aren't senior or junior engineers alone, but non-technical employees using Goose to build internal tools. Enterprise risk teams, legal, and other departments are gaining massive productivity from being able to self-service software creation, compressing weeks of work into hours.", "timestamp_start": "00:48:56", "timestamp_end": "00:51:22", "line_start": 367, "line_end": 381 }, { "id": "topic_10", "title": "Hiring and cultural shifts for AI adoption", "summary": "Block is hiring for learning mindset and openness to AI tools rather than AI expertise. Interview processes now include vibe coding exercises. The organizational shift from GM to functional structure actually had more impact on productivity than AI tools themselves, challenging the assumption that headcount needs to scale with features.", "timestamp_start": "00:44:34", "timestamp_end": "00:52:12", "line_start": 337, "line_end": 386 }, { "id": "topic_11", "title": "Code quality doesn't determine product success", "summary": "Dhanji shares the counterintuitive insight that code quality is largely irrelevant to product success, using YouTube as the prime example. YouTube had terrible architecture (storing videos as MySQL blobs, slow Python stack) but dominated Google Video. What matters is solving customer problems, not code elegance.", "timestamp_start": "01:02:02", "timestamp_end": "01:04:19", "line_start": 472, "line_end": 483 }, { "id": "topic_12", "title": "Controlled chaos and engineer freedom", "summary": "At Cash App, Dhanji employed a philosophy of controlled chaos—allowing engineers freedom to experiment and even waste time on unpromising ideas, while maintaining non-negotiable foundations around security, liability, and financial integrity. This balance bred creativity and high-impact shipping.", "timestamp_start": "01:06:13", "timestamp_end": "01:07:26", "line_start": 500, "line_end": 507 }, { "id": "topic_13", "title": "Start small principle and examples", "summary": "Dhanji emphasizes the core principle 'start small with everything,' using Goose, Cash App, and Bitcoin as examples of major products that began as small experiments or hack week projects. Google Wave failed by starting big; successful products iterate from small prototypes.", "timestamp_start": "01:08:26", "timestamp_end": "01:11:22", "line_start": 517, "line_end": 558 }, { "id": "topic_14", "title": "Failed products and lessons learned", "summary": "Dhanji's career includes multiple significant failures: Google Wave (overambitious, 70-80 engineers before finding product-market fit), Google+ (social networking), Secret (anonymous social app), and an email startup with Canva co-founder. These failures taught him never to repeat certain classes of mistakes and instilled humility.", "timestamp_start": "01:14:08", "timestamp_end": "01:15:36", "line_start": 595, "line_end": 605 }, { "id": "topic_15", "title": "Open source philosophy and giving back", "summary": "Dhanji explains why Block open-sourced Goose despite it being valuable IP. Block is built on open source (Linux, Java, MySQL) and believes in contributing back to the ecosystem. Open protocols and open source outlast individual companies and serve the broader community.", "timestamp_start": "00:27:23", "timestamp_end": "00:28:18", "line_start": 211, "line_end": 216 }, { "id": "topic_16", "title": "Practical AI adoption at Block: personal use cases", "summary": "Dhanji shares personal examples of using Goose to solve real problems: organizing therapy receipts into Apple Notes using AppleScript, and helping a colleague pull images from Google Docs. These hands-on experiences teach leaders about AI strengths and how to deploy them effectively.", "timestamp_start": "00:56:56", "timestamp_end": "00:58:02", "line_start": 427, "line_end": 431 }, { "id": "topic_17", "title": "Process optimization and questioning base assumptions", "summary": "Dhanji emphasizes that real productivity gains often come from questioning whether processes should exist at all, not just optimizing them. Example: shortening CI/CD pipelines by deleting unnecessary tests saved 2-3x more time than test selection tools. This portfolio thinking is where AI currently underperforms.", "timestamp_start": "00:52:21", "timestamp_end": "00:54:04", "line_start": 391, "line_end": 407 }, { "id": "topic_18", "title": "CTO reflections and lessons learned", "summary": "After two years as CTO, Dhanji's biggest learnings are: Conway's Law is powerful (structure determines output), and organizations need quiet time to step back and assess holistically—not just react when things go wrong. Leadership also requires humility and openness to others' viewpoints.", "timestamp_start": "01:00:13", "timestamp_end": "01:01:15", "line_start": 457, "line_end": 461 }, { "id": "topic_19", "title": "AI leadership: using tools yourself first", "summary": "Dhanji's core advice for leaders: use AI tools yourself daily before directing teams to use them. Jack Dorsey, leadership team, and Dhanji all use Goose regularly, which informs organizational adoption far better than reading think pieces or mandates. Personal understanding of ergonomics and limitations is crucial.", "timestamp_start": "00:54:04", "timestamp_end": "00:55:06", "line_start": 403, "line_end": 407 }, { "id": "topic_20", "title": "Purpose-driven technology and staying grounded", "summary": "Dhanji emphasizes that in an era of rapid change and hype, leaders should focus on their company's core purpose rather than chasing every technology trend. Block's purpose is economic empowerment; technology should serve that, not be an end unto itself. This clarity prevents wasteful tool-building and keeps teams focused.", "timestamp_start": "01:16:42", "timestamp_end": "01:17:51", "line_start": 611, "line_end": 620 } ], "insights": [ { "id": "insight_1", "text": "This is the worst it will ever be. This is the baseline. The value is changing every day, so you need to ride that wave along with it.", "context": "When discussing AI productivity metrics, Dhanji emphasizes that 8-10 hours saved per week is just the starting point—the technology is improving daily.", "topic_id": "topic_1", "line_start": 5, "line_end": 5 }, { "id": "insight_2", "text": "Where you have a lot of depth and a lot of really strong people come together is where AI still underperforms humans. Architecture, design, race conditions, orchestration—that's still an area where AI isn't quite there.", "context": "Dhanji identifies the current limitations of AI in complex technical decision-making despite its strength in routine tasks.", "topic_id": "topic_5", "line_start": 149, "line_end": 152 }, { "id": "insight_3", "text": "Code quality is important to building a successful product. The two have nothing to do with each other.", "context": "Dhanji's counterintuitive lesson learned from his career, supported by YouTube's messy architecture but massive success.", "topic_id": "topic_11", "line_start": 29, "line_end": 29 }, { "id": "insight_4", "text": "All these LLMs are sitting idle overnight and on weekends while humans aren't there. They should be working all the time, trying to build in anticipation of what we want.", "context": "Dhanji describes the next frontier of AI work—autonomous agents running continuously without human supervision.", "topic_id": "topic_4", "line_start": 23, "line_end": 23 }, { "id": "insight_5", "text": "Non-technical people using AI agents and programming tools to build things is really what's been surprising. The lines are going to be blurred between whether you're in legal, risk, engineering, or design.", "context": "Enterprise risk management and other non-technical teams are seeing the biggest productivity gains from Goose.", "topic_id": "topic_9", "line_start": 17, "line_end": 17 }, { "id": "insight_6", "text": "Structure matters more than the efficacy of the tools you have. Conway's Law is really powerful—you ship your org structure.", "context": "Dhanji emphasizes that the functional reorganization had as much impact as any AI tool.", "topic_id": "topic_7", "line_start": 110, "line_end": 110 }, { "id": "insight_7", "text": "You only hear about it when things are going wrong. When things are going well, you have this eerie silence. You need time to step back and look at things holistically.", "context": "Lesson learned in first two years as CTO—leaders must create space for reflection, not just react to crises.", "topic_id": "topic_18", "line_start": 458, "line_end": 461 }, { "id": "insight_8", "text": "A learning mindset is what matters. Everything we do, every experiment, what can we learn from it? Did we give it our best shot?", "context": "Jack Dorsey's philosophy that drives hiring and culture at Block—prioritizing learning over outcomes.", "topic_id": "topic_10", "line_start": 353, "line_end": 353 }, { "id": "insight_9", "text": "Question base assumptions on everything. Should we even build this at all? What's the purpose? Could we build something completely different that matters more?", "context": "Core principle for avoiding wasted effort and staying focused on core business.", "topic_id": "topic_17", "line_start": 575, "line_end": 578 }, { "id": "insight_10", "text": "If you try to boil the ocean to make a cup of tea, you'll never get there. Start small with everything.", "context": "Dhanji's most important leadership principle, demonstrated through successes and failures.", "topic_id": "topic_13", "line_start": 518, "line_end": 518 }, { "id": "insight_11", "text": "The fact that everyone's building software means there's more software to be built, more coordination to happen. We're seeing an overall uptake in velocity and asks for more features.", "context": "The productivity paradox—making tools available increases demand rather than reducing headcount needs.", "topic_id": "topic_9", "line_start": 380, "line_end": 380 }, { "id": "insight_12", "text": "A certain amount of creativity that chaos breeds and you have to know how to build controlled chaos. Create a foundation that isn't liable to rupture, then allow engineers freedom to experiment.", "context": "Balance needed in engineering culture—safety guardrails plus experimentation freedom.", "topic_id": "topic_12", "line_start": 506, "line_end": 506 }, { "id": "insight_13", "text": "There's a trap in getting away from your core purpose as a company. If we're purely looking at dollars versus dollars building in-house tools, that's pulling us off purpose.", "context": "Warning against the temptation to build everything internally despite AI making it cheaper.", "topic_id": "topic_6", "line_start": 314, "line_end": 317 }, { "id": "insight_14", "text": "The way we've been able to drive adoption is Jack uses Goose, I use Goose, our executive team all use it daily. Learn about how your own workflow changes, then apply to teams.", "context": "Leaders must model AI tool adoption themselves to drive organizational change.", "topic_id": "topic_19", "line_start": 404, "line_end": 407 }, { "id": "insight_15", "text": "If you're not waking up in the morning feeling energized about what you're going to do that day in your professional life, then change something. Don't accept what's meted out to you.", "context": "Personal motto guiding career decisions and transitions.", "topic_id": "topic_20", "line_start": 662, "line_end": 662 }, { "id": "insight_16", "text": "A year from now you're going to look back on what looks like a monumental problem and be like 'oh, that was so trivial.' There's always more to the world and it's never too late.", "context": "Advice for overcoming fear of major career changes or decisions.", "topic_id": "topic_20", "line_start": 668, "line_end": 671 }, { "id": "insight_17", "text": "We need to demand more of their companies, employers, teams. Default to making things open source, build for everyone. Demand that people unlock AI for shared benefit, not walled gardens.", "context": "Final call to action about how the community should push back on corporate AI lock-in.", "topic_id": "topic_15", "line_start": 692, "line_end": 695 }, { "id": "insight_18", "text": "I'm much more keen on looking for people eager to learn about these tools and open to them, not people amazing at AI on day one. It's about learning mindset.", "context": "Hiring philosophy in the AI era—attitude and adaptability matter more than current AI skills.", "topic_id": "topic_10", "line_start": 344, "line_end": 347 }, { "id": "insight_19", "text": "You don't need to be deep with the technology or at the forefront of every trend. Technology is here to serve us, and if we have important purpose, technology can serve that.", "context": "Antidote to FOMO and hype-driven strategy—purpose-driven technology adoption.", "topic_id": "topic_20", "line_start": 614, "line_end": 614 }, { "id": "insight_20", "text": "We need a lot of human taste to anchor these AIs so they don't go off script. That's where design teams are pushing us, and that's a differentiator that pushes us beyond AI slop.", "context": "The critical role of design and human judgment in preventing low-quality AI outputs.", "topic_id": "topic_5", "line_start": 299, "line_end": 299 } ], "examples": [ { "id": "example_1", "explicit_text": "engineering teams that are very, very AI forward are reporting about eight to 10 hours save per week", "inferred_identity": "Block engineering teams using Goose", "confidence": "high", "tags": [ "Block", "engineering team", "Goose", "AI agent", "productivity gains", "metrics", "self-reported" ], "lesson": "AI agents can deliver measurable productivity gains of 8-10 hours per week in engineering teams actively adopting them, with further improvements expected.", "topic_id": "topic_1", "line_start": 5, "line_end": 5 }, { "id": "example_2", "explicit_text": "he'll find that Goose has already tried to build that feature and opened a PR for it on Git", "inferred_identity": "Block engineer working on Goose team", "confidence": "high", "tags": [ "Block", "engineer", "Goose", "autonomous AI", "screen watching", "PR generation", "proactive work" ], "lesson": "AI agents with screen watching capabilities can anticipate work needs from conversations and proactively generate code PRs, showing promise for near-autonomous development.", "topic_id": "topic_3", "line_start": 11, "line_end": 11 }, { "id": "example_3", "explicit_text": "YouTube is the most successful product at Google by a long way. YouTube's code base is horrible... they're storing videos as blobs in MySQL", "inferred_identity": "YouTube at Google", "confidence": "high", "tags": [ "YouTube", "Google", "product success", "code quality", "architecture", "database design", "messy codebase" ], "lesson": "Messy code and poor architecture don't determine product success; solving real customer problems matters far more than technical elegance.", "topic_id": "topic_11", "line_start": 473, "line_end": 476 }, { "id": "example_4", "explicit_text": "Google Video, which is product... It supported more formats, it supported higher resolution... YouTube had none of this", "inferred_identity": "Google Video vs YouTube", "confidence": "high", "tags": [ "Google", "Google Video", "YouTube", "feature comparison", "market competition", "product strategy", "simplicity vs features" ], "lesson": "A simpler product that solves the core use case better can win against a technically superior product with more features.", "topic_id": "topic_11", "line_start": 476, "line_end": 479 }, { "id": "example_5", "explicit_text": "At some point, I observed that we were talking about lots of deep things... but no one was really paying attention to AI, and so that's when I wrote that letter", "inferred_identity": "Dhanji R. Prasanna writing AI manifesto to Jack Dorsey at Block/Square", "confidence": "high", "tags": [ "Block", "Square", "Dhanji Prasanna", "Jack Dorsey", "AI strategy", "leadership manifesto", "organizational transformation" ], "lesson": "Bringing unaddressed strategic issues to leadership's attention through a focused memo can catalyze organizational transformation.", "topic_id": "topic_8", "line_start": 65, "line_end": 68 }, { "id": "example_6", "explicit_text": "When I started working at what was then known as Square, we were always thought of as a technology company just like Google or Facebook", "inferred_identity": "Square/Block under Dhanji's engineering leadership", "confidence": "high", "tags": [ "Square", "Block", "technology company", "identity", "culture shift", "financial services", "fintech" ], "lesson": "Reframing company identity from industry vertical (FinTech) to core capability (technology company) enables strategic focus on technical innovation.", "topic_id": "topic_7", "line_start": 80, "line_end": 80 }, { "id": "example_7", "explicit_text": "Cash App was like that... Cash App started more or less as a hack week sort of idea and grew into a bigger and bigger thing", "inferred_identity": "Cash App at Block/Square", "confidence": "high", "tags": [ "Cash App", "Block", "Square", "hack week", "product launch", "mobile payments", "peer-to-peer" ], "lesson": "Major products can grow from small hack week experiments; starting small and iterating beats trying to build complete products from day one.", "topic_id": "topic_13", "line_start": 533, "line_end": 533 }, { "id": "example_8", "explicit_text": "We became the very first company that was a public company to launch a Bitcoin product. And that was again a hack week idea that actually Jack and me and another engineer worked on.", "inferred_identity": "Block/Square Bitcoin feature with Jack Dorsey and Dhanji Prasanna", "confidence": "high", "tags": [ "Block", "Square", "Bitcoin", "Jack Dorsey", "Dhanji Prasanna", "hack week", "public company", "cryptocurrency" ], "lesson": "Even groundbreaking financial innovations can start as small hack week experiments; CEO involvement in hands-on building creates strategic clarity.", "topic_id": "topic_13", "line_start": 533, "line_end": 539 }, { "id": "example_9", "explicit_text": "We went and bought a cup of coffee, a blue bottle, and it was bought using Bitcoin over cash card", "inferred_identity": "First Bitcoin transaction at Square with Jack Dorsey and Dhanji", "confidence": "high", "tags": [ "Square", "Bitcoin", "Blue Bottle Coffee", "Jack Dorsey", "first transaction", "product testing", "real world use" ], "lesson": "Testing new financial products with real transactions (even symbolic ones) validates viability and creates memorable proof points.", "topic_id": "topic_13", "line_start": 545, "line_end": 545 }, { "id": "example_10", "explicit_text": "I worked at Google on this product called Google Wave, which was trying to be everything to everyone... 70, 80 engineers building this thing before it even really had any users", "inferred_identity": "Google Wave", "confidence": "high", "tags": [ "Google", "Google Wave", "product failure", "over-engineering", "collaboration", "overbuild", "bloated scope" ], "lesson": "Starting with massive resources and trying to build everything from day one leads to feature bloat and product failure—start small and validate first.", "topic_id": "topic_13", "line_start": 563, "line_end": 563 }, { "id": "example_11", "explicit_text": "I worked for Hot Minute on Google+, which was another epic failure", "inferred_identity": "Google+", "confidence": "high", "tags": [ "Google", "Google+", "social network", "product failure", "Google failure", "competition with Facebook" ], "lesson": "Even tech giants with massive resources can fail to build successful social networks if they don't understand user behavior and network effects.", "topic_id": "topic_14", "line_start": 596, "line_end": 596 }, { "id": "example_12", "explicit_text": "I worked at this social networking startup called Secret, which burned hot for a bright minute and then blew up", "inferred_identity": "Secret (anonymous social network)", "confidence": "high", "tags": [ "Secret", "social network", "startup", "anonymous posting", "product failure", "failed experiment" ], "lesson": "Anonymous social networks face inherent challenges with abuse and sustainability that can lead to rapid failure despite early hype.", "topic_id": "topic_14", "line_start": 596, "line_end": 596 }, { "id": "example_13", "explicit_text": "There was an email startup that we did... the co-founder of Canva and I worked on that one", "inferred_identity": "Email startup co-founded by Dhanji and Canva co-founder", "confidence": "medium", "tags": [ "email", "startup", "Dhanji Prasanna", "Canva co-founder", "product failure", "email application" ], "lesson": "Even experienced founders with good co-founders can fail if they don't find product-market fit, requiring learning and iteration.", "topic_id": "topic_14", "line_start": 602, "line_end": 602 }, { "id": "example_14", "explicit_text": "In the early days of Cash App, I was head of engineering from when we were about 10 engineers to 200 plus... about 10 plus or 20 million users", "inferred_identity": "Cash App at Block/Square under Dhanji's engineering leadership", "confidence": "high", "tags": [ "Cash App", "Block", "Square", "head of engineering", "scaling", "growth", "mobile payments" ], "lesson": "Scaling engineering from 10 to 200+ people while hitting 10-20M users requires balancing structure with creative freedom.", "topic_id": "topic_12", "line_start": 500, "line_end": 500 }, { "id": "example_15", "explicit_text": "My son has a whole bunch of therapies... I was trying to gather the receipts... I asked Goose to do this... convert to HTML... sync to phone... share with wife", "inferred_identity": "Dhanji Prasanna's personal use of Goose", "confidence": "high", "tags": [ "Goose", "personal use", "receipts management", "AppleScript", "Apple Notes", "HTML conversion", "family use" ], "lesson": "AI agents can solve complex real-world problems combining multiple tools and formats by being given high-level goals and allowed to try different approaches.", "topic_id": "topic_16", "line_start": 416, "line_end": 420 }, { "id": "example_16", "explicit_text": "I was trying to pull images out of a Google Doc... I just went to Lovable and built an app... Google Doc URL and let me download the images", "inferred_identity": "Lenny Rachitsky using Lovable (AI code tool)", "confidence": "high", "tags": [ "Lovable", "Google Docs", "image extraction", "no-code", "personal tool", "quick build" ], "lesson": "AI code generation tools enable non-engineers and busy professionals to quickly build small utility applications for personal or team use.", "topic_id": "topic_16", "line_start": 413, "line_end": 413 }, { "id": "example_17", "explicit_text": "We have these tools that run 24/7 or run in the CI pipeline... analyzing vulnerabilities... looking at bugs filed on tickets and trying to build patches while engineers are asleep", "inferred_identity": "Block's background AI processes and CI/CD automation", "confidence": "high", "tags": [ "Block", "CI/CD", "automation", "vulnerability scanning", "bug fixing", "overnight processing" ], "lesson": "AI agents running in CI/CD pipelines can identify and propose fixes for vulnerabilities and bugs, enabling engineers to wake up to patch suggestions.", "topic_id": "topic_1", "line_start": 125, "line_end": 125 }, { "id": "example_18", "explicit_text": "Enterprise risk management team build a whole system for self-servicing enterprise risk... compressing weeks of work into hours", "inferred_identity": "Block's enterprise risk management team using Goose", "confidence": "high", "tags": [ "Block", "enterprise risk management", "Goose", "internal tools", "productivity", "self-service" ], "lesson": "Non-technical teams using AI agents can build internal tools that would normally require weeks of engineering coordination, in just hours.", "topic_id": "topic_9", "line_start": 143, "line_end": 143 }, { "id": "example_19", "explicit_text": "We have this other tool called Gosling... a goose for mobile... operates Android OS... automating UI tests", "inferred_identity": "Gosling - Block's mobile AI agent for Android", "confidence": "high", "tags": [ "Block", "Gosling", "Android", "mobile automation", "UI testing", "accessibility API" ], "lesson": "AI agents can operate at the OS level using accessibility APIs to automate mobile testing, replacing manual QA work.", "topic_id": "topic_1", "line_start": 146, "line_end": 146 }, { "id": "example_20", "explicit_text": "We cut down basically 50% of test runs... offloading tests to the cloud or simply just deleting tests that don't make sense anymore, probably save you two to three times that", "inferred_identity": "Block's test optimization strategy", "confidence": "high", "tags": [ "Block", "testing", "CI/CD optimization", "test selection", "test deletion", "build time", "environmental impact" ], "lesson": "Test optimization and deletion can deliver 2-3x more productivity gains than tool-based test selection, demonstrating the importance of questioning base assumptions.", "topic_id": "topic_17", "line_start": 392, "line_end": 395 }, { "id": "example_21", "explicit_text": "Brad, our creator of Goose... believed very early on that agents would be how we unlock value from LLMs... built a proof concept... shared it with Databricks and Anthropic", "inferred_identity": "Brad - Goose creator at Block", "confidence": "high", "tags": [ "Block", "Brad", "Goose", "AI agent", "creator", "Databricks", "Anthropic", "proof of concept" ], "lesson": "Individual engineers with conviction about emerging technology can drive company-wide adoption by building proofs of concept and sharing with partners.", "topic_id": "topic_13", "line_start": 530, "line_end": 530 }, { "id": "example_22", "explicit_text": "Jack came over to Sydney and spent two days with me... both of us like long walks... then he offered me the job", "inferred_identity": "Jack Dorsey recruiting Dhanji Prasanna as CTO", "confidence": "high", "tags": [ "Block", "Square", "Jack Dorsey", "Dhanji Prasanna", "CTO", "recruitment", "personal connection" ], "lesson": "Significant career opportunities can emerge from personal connections and conversations with company leadership.", "topic_id": "topic_8", "line_start": 74, "line_end": 74 } ] }