Lenny’s Podcast: How Block is Becoming the Most AI-Native Enterprise in the World
Guest: Dhanji R. Prasanna, CTO at Block
Host: Lenny Rachitsky
Date: October 26, 2025
Episode Overview
In this episode, Lenny interviews Dhanji R. Prasanna, the Chief Technology Officer at Block (formerly Square), to uncover how Block became one of the world’s most AI-native large companies. Dhanji shares concrete lessons for leaders seeking to make their organizations truly technology—and especially AI—driven. They dig into the genesis and impact of Block’s internal open source AI agent (Goose), organizational restructuring, AI adoption across technical and non-technical teams, and broader leadership and product lessons from Dhanji’s career, including pivotal experiences at Google and Cash App.
Key Topics & Insights
1. Laying the Foundation for AI-Native Transformation
1.1 The AI Manifesto and Cultural Shift
- Dhanji wrote an "AI manifesto" letter to Jack Dorsey to urge AI as a central company focus.
“No one was really paying attention to AI. That’s when I wrote the letter… I think we should do this, do it centrally, and be ahead of the game and be AI-native.”
— Dhanji, [06:00] - Dhanji wasn’t CTO at the time, but his vision convinced Jack Dorsey, leading to his appointment after deep discussions.
“Jack came over to Sydney and spent two days with me. …He offered me the job.”
— Dhanji, [07:00]
1.2 Restructuring for Tech-First Focus
- Block shifted from a GM-based structure (siloed, product-based) to a functional one (engineering, design, etc. centralized).
“When you really want to go deep in technology… you need a singular focus. …That was the big transformation.”
— Dhanji, [09:23] - This enabled unified technical direction, facilitated collaboration, and transformed company culture back to its “technology company” roots.
“Find your DNA and really try to optimize for what that is in a very simple and clear way.”
— Dhanji, [11:19]
2. How AI is Actually Changing Work at Block
2.1 Transformation in How Teams Operate
- AI-Forward Teams: Those closely embracing AI tools report the most dramatic changes, such as building software without hand-coding (vibe coding).
“Teams that are very AI-native are building without writing lines of code by hand.”
— Dhanji, [14:12] - Baseline Productivity Gains: Self-reported and data-validated—up to 8-10 hours saved per week for engineering teams; ~20-25% manual hours saved company-wide.
“We feel across the whole company we’re probably trending towards 20-25% of manual hours saved… and that’s just the start.”
— Dhanji, [16:02] - AI Use by Non-technical Staff: Notably, non-engineering teams—like legal and risk—have compressed weeks of work into hours using AI agents to build self-service tools.
“One of the most surprising and energizing uses of Goose within Block is... non-technical teams… that’s compressing weeks of work into hours.”
— Dhanji, [16:02, 47:33]
3. Deep Dive on Goose: Block’s Open Source AI Agent
3.1 What is Goose?
- Goose is an open source, general-purpose AI agent and orchestration platform that integrates with internal and third-party systems via the Model Context Protocol (MCP).
- Can automate tasks ranging from photo organization to code generation to custom workflow automation.
“Goose gives these [AI] brains arms and legs to go out and act in our digital world… [it’s] entirely open source.”
— Dhanji, [21:49], [23:54]
3.2 Notable Features and Stories
- Plug-and-play with multiple models (Claude, OpenAI, open source), easy extension via MCPs, and capable of writing its own MCPs.
- Anticipation and Autonomy: Goose can “watch” an employee work (listening and screen watching), and autonomously draft PRs for in-progress ideas.
“He’ll be talking to a colleague about a feature… a few hours later, [Goose] has already tried to build that feature and opened a PR for it.”
— Dhanji, [28:50]
3.3 Open Sourcing and Community Impact
- Goose is widely adopted, not just internally but by partners and competitors alike.
“We have a lot of companies using Goose pretty actively... Databricks talks about it a lot.”
— Dhanji, [26:51] - It reflects Block’s commitment to open protocols and giving back to the ecosystem.
“We feel like we have a strong imperative to give back… That’s certainly a core value for us.”
— Dhanji, [27:23]
4. AI, Productivity, and the Future of Engineering
4.1 The Future of Software Development
- High autonomy is coming: LLM-powered agents will utilize “idle” time and run experiments/implement ideas overnight.
“All these LLMs are sitting idle overnight… there’s no need for that. They should be working all the time, building in anticipation of what we want.”
— Dhanji, [32:32] - Regularly rewriting and deleting code can become the norm—AI can spin up multiple versions for human review and selection.
“What would our world look like if every single release we deleted the entire app and rebuilt it from scratch?”
— Dhanji, [32:32]
4.2 Combining Human Judgment with AI Power
- AI is not yet able to replace nuanced human judgment, particularly across entire systems or in process prioritization.
“AI isn’t good at [global judgment]… it’s better for a human to use judgment.”
— Dhanji, [40:11] - Anchoring AI work in design thinking and human “taste” will differentiate companies amid increasing automation.
“That’s a differentiator that will push us beyond this era of AI slop everyone’s talking about.”
— Dhanji, [37:54]
5. Organizational, Hiring, and Leadership Lessons
5.1 Impact on Hiring & Team Structure
- Organizational design (functional vs. GM) has a bigger productivity impact than AI alone.
“If you’re trying to be more productive, forget AI—just reorg into a functional structure.”
— Lenny, [52:11] - Growing importance of a "learning mindset" for new hires—valuing curiosity and willingness to adopt new tools over deep AI expertise.
“We’re very much looking for people that are embracing these tools and that are eager to try and learn from it… a learning mindset is how I would put it.”
— Dhanji, [46:55]
5.2 Whose Productivity Rises Most?
- Both juniors (eager, fearless adopters) and seniors (relieved to automate repeat tasks) benefit, but most remarkable is the blurring of technical/non-technical roles.
“The non-technical people using AI agents… are really what’s been surprising and really amazing.”
— Dhanji, [48:55]
6. Counterintuitive & Hard-Won Lessons
6.1 Code Quality vs. Product Success
- Code quality and product success are largely unrelated.
“A lot of engineers think that code quality is important to building a successful product. The two have nothing to do with each other.”
— Dhanji, [62:01] - E.g. YouTube’s early success despite a “terrible” codebase; focus instead on solving real user problems.
6.2 Controlled Chaos and Starting Small
- Creativity and speed thrive in “controlled chaos,” not rigid process.
- Best products start small and focused—Cash App and Goose both originated as small internal experiments.
“Goose started small. It was just an engineer working on their own time, trying to build something useful.”
— Dhanji, [69:23]
Notable Quotes & Moments
-
On AI-driven autonomy:
“You can give it your wishlist of ten things… it’s successful on like 60%.”
— Dhanji, [36:56] -
On humans’ long-term role:
“We’re going to need a lot of human taste to anchor these AIs so they don’t go off script.”
— Dhanji, [37:54] -
On questioning assumptions and “building for building’s sake”:
“Do we even need to do this process at all?”
— Dhanji, [52:21] -
On advice for AI adoption:
“Really try and use these tools yourself. …Feel it, use the product yourself, understand its strengths and weaknesses and its ergonomics, then figure out how to apply it to your teams.”
— Dhanji, [54:04] -
On counterintuitive product lessons:
“YouTube… most successful product at Google… had a terrible codebase. …Just focus on what we’re trying to build and for whom.”
— Dhanji, [62:01]
Timestamps for Important Segments
- [06:00] — Dhanji’s AI Manifesto and pathway to CTO
- [09:23] — Organizational transformation: From GM to functional structure
- [14:12] — AI-native engineering work and “vibe coding”
- [16:02] — Quantifying AI impact: hours saved, company-wide gains
- [21:49] — What is Goose? Capabilities, open source, and integrations
- [28:50] — The “Goose watches you” story: agent as co-worker
- [32:32] — The future: overnight AI agent autonomy, regular code rewrites
- [40:11] — Where humans must intervene; limits of AI’s judgment
- [46:55] — Hiring priorities: learning mindset, embracing AI
- [52:11] — Functional org structure vs. AI for productivity
- [62:01] — Code quality ≠ product success; YouTube vs. Google Video
- [69:23] — Starting small: Goose and Cash App originated as tiny projects
Leadership & Career Wisdom
- Question base assumptions: Don’t just optimize—ask if it’s worth doing at all.
- Leadership lesson: You only hear about things when they’re going wrong; make time to assess big-picture priorities.
- Managing chaos: Balance freedom and experimentation with foundational reliability to foster innovation.
Lightning Round — Personal Recommendations
- Books:
- The Master and Margarita by Mikhail Bulgakov
- Tennyson’s Poetry
- TV Shows: Alien Earth (Noah Hawley), Slow Horses
- Product: Steam Deck OLED
- Life Motto:
“If you’re not waking up in the morning feeling energized about what you’re going to do… change something.”
— Dhanji, [81:55] - On overcoming fear: "A year from now your monumental problem will seem trivial."
- Pop Culture Mad Scientist: Doc Brown from Back to the Future
Final Advice
- Use and experiment with AI tools yourself.
- Focus organizational energy on core purpose and openness over chasing trends.
- Don’t conflate code quality with delivering valuable products.
- Encourage experimentation, but anchor in what matters to people and your company’s mission.
For more and to access Goose:
- Block open source/GitHub: [link in show notes]
- Contact Dhanji: LinkedIn
This episode is essential listening for company leaders, tech managers, and anyone interested in building more nimble, future-proof organizations in the age of enterprise AI.