Chronological feed of everything captured from Garry Tan.
This post from Garry Tan, "Many such cases," without further context, suggests a recurring phenomenon or trend within a domain known to his audience. The brevity indicates a shared understanding of the situation being referenced, likely among those familiar with his prior commentary or the specific community he addresses. It functions as a concise, insider observation rather than a detailed explanation.
gstack is an open-source framework that overlays a structured software development lifecycle (SDLC) onto AI coding agents, specifically Claude Code and others. By replacing generic prompting with specialized roles (e.g., /cso for security, /qa for testing) and a stateful pipeline, it enables massive increases in LOC throughput and reduces 'AI slop' through rigorous, multi-stage review gates. The system integrates advanced capabilities like headless browser control, cross-model verification via OpenAI Codex, and parallel session management.
Marc Andreessen posits that Artificial General Intelligence (AGI) is currently in existence but its distribution across various sectors and applications is not uniform. This suggests a phase where advanced AI capabilities are present but not yet universally accessible or recognized, implying a nascent stage of integration rather than a total absence of the technology.
Programming is experiencing a period of significant vitality and growth. This resurgence is characterized by continuous innovation and expanding applications, making the field more dynamic and engaging than in previous eras.
Garry Tan asserts a direct relationship between a system's purpose and its actual outcomes. He posits that if a system's outcome does not serve the public good, it constitutes fraud. This framework suggests an ethical imperative for system design and evaluation, linking functionality directly to societal benefit.
An open-source AI system built for Claude Code successfully processed over 700 job applications, leading to a job offer. This system automates critical job search tasks, including scanning career pages, tailoring CVs for specific roles, and completing application forms. Its modular design and pre-configured company listings demonstrate its practical applicability for technical job seekers.
The deployment of autonomous vehicles, specifically Waymo in San Francisco, faces significant political obstacles despite technical readiness. The inability for Waymo to perform basic rideshare functions like curbside drop-off and pickup at major transportation hubs points to regulatory and political friction rather than technological limitations. This suggests that the advancement of AV integration into urban infrastructure is currently more constrained by governance than by engineering.
Garry Tan is utilizing a combination of GStack and OpenClaw to enhance his AI, Wintermute. This integration has enabled the rapid development of an early warning system for potentially viral posts on the X platform, demonstrating a quick turnaround from AI development to practical application.
AI-powered platforms are enabling just-in-time software development, drastically reducing project timelines. This advancement is exemplified by a project that traditionally required weeks being completed in minutes through a single command, leveraging integrated AI context and code generation capabilities. This marks a significant shift in software engineering paradigms.
GStack Browse is a new AI-powered browser designed for developers, offering a significantly faster alternative to existing AI browsers. It integrates with Claude Code for interactive navigation and operations, supporting use cases like page debugging and CSS interaction. The project is open-source and leverages a Playwright CLI for its performance.
AI, when directed with clear optimization goals, can significantly improve web performance metrics. This is exemplified by the improvement of "Garry's List" page speed to a 95. The key is to provide AI with specific targets and allow it to manage the optimization process.
GStack facilitates immediate integration of GitHub repositories as native OpenClaw skills, even without explicit OpenClaw support. This allows developers to leverage existing code in OpenClaw without modification, streamlining skill development and deployment. The system automatically interprets the skill file within the pointed GitHub repository.
Garry Tan highlights the destructive nature of the "drama triangle" in organizational behavior. He argues that an inability to confront issues directly leads to triangulation, involving a third party unnecessarily, which ultimately corrodes company culture. This dynamic prevents transparent communication and problem-solving, fostering an environment of indirectness and organizational decay.
Garry Tan, a prominent figure, has publicly labeled content on the X platform as "dishonest." This direct critique suggests a potential concern regarding misinformation or manipulative content within the platform's ecosystem. The statement, while brief, highlights a noteworthy stance from an influential voice.
Garry Tan is entering a "honing phase" for GStack, indicating a focus on refactoring and reducing code bloat. Concurrently, he is also initiating work on new, unspecified projects. This suggests a strategic reallocation of development resources towards optimization and innovation.
This post, while brief, indicates that Garry Tan is performing exceptionally well as a founder, purportedly outperforming 99.9% of his peers. The core insight is about his high performance relative to other founders.
Garry Tan reports generating 37,000 lines of code (LOC) daily across five projects using "agentic engineering," a process he claims significantly boosts productivity by leveraging AI to autonomously generate code from commands. This approach contrasts with traditional software development metrics that favor minimal, maintainable code. The debate centers on whether raw LOC output remains a valid productivity indicator in an AI-driven development paradigm where code generation is highly automated and focused on rapid capability delivery.
GStack, an open-source AI framework, now integrates with Pretext, an AI skill, to facilitate rapid, AI-driven HTML design. This integration allows users to quickly generate and iterate on responsive HTML designs through simple commands within a Claude Code environment. The creator claims a significant personal productivity boost, suggesting the framework fosters efficient software development.
GStack is described as a versatile AI agent designed to streamline software development. It functions as an entire development team, offering roles from UX design to release engineering, enabling users to build and ship software rapidly. The core value proposition is its ability to facilitate the creation of functional software from an idea within an hour.
Richmond experienced a significant increase in car thefts following the deactivation of Flock cameras, a decision initially made to protect immigrants. The immigrant shopkeeping community, disproportionately affected by the rise in crime, actively lobbied for the cameras' return. The city council ultimately reversed its decision, highlighting the practical need for security measures over ideological opposition in the face of demonstrable negative consequences.
Garry Tan, a prominent figure, expresses well wishes for Jackie Fielder, a political opponent, highlighting the importance of community and civility despite political differences. This demonstrates a call for respectful engagement within the San Francisco political landscape, prioritizing human well-being over ideological divides. The sentiment suggests a potential for bridging divides by acknowledging shared community values.
A street-level poll in San Francisco indicates a strong preference for Command Line Interfaces (CLI) over Menu-Controlled Programs (MCP), particularly among younger demographics. This sentiment aligns with efficiency-focused development trends, with MCPs being criticized for their perceived bloat. The findings informed Composio's decision to launch a universal CLI.
Resend Robot offers a development tool for Ruby on Rails applications that intercepts Resend API calls, stores emails as JSON, and provides a web UI for browsing, previewing, and simulating replies. This enables developers to test email functionalities thoroughly without sending actual emails, ensuring consistent behavior between development and production environments. The tool also integrates with Claude AI for conversational interaction with the development mailbox and provides robust testing utilities.
Garry Tan's "Mega Plan Review" is a comprehensive framework for rigorously evaluating software development plans. It emphasizes three core modes: SCOPE EXPANSION, HOLD SCOPE, and SCOPE REDUCTION, each with distinct objectives. The framework includes a pre-review system audit, a "Step 0" for challenging premises and selecting a review mode, and ten detailed sections covering architecture, error handling, security, testing, performance, observability, and long-term implications. It mandates detailed documentation, explicit error mapping, and a strong bias towards preventing silent failures and ensuring robust, maintainable systems.
This content explores the critical role of resilience, well-being, and self-awareness for entrepreneurial success, moving beyond traditional hustle culture. It examines how early life experiences and internal "bugs" can impact founder effectiveness and offers practical advice on building support systems, leveraging AI for self-improvement, and understanding motivation beyond extrinsic factors. The synthesis emphasizes that personal well-being is not a luxury but a foundational requirement for sustained achievement in the demanding startup environment.
Many high-profile individuals, despite achieving immense financial and professional success, succumb to personal ruin due to an overreliance on conventional metrics like money and fame. True fulfillment and sustained well-being stem from intentionally redefining personal "games" to prioritize intrinsic values, such as craft, generosity, and strong personal relationships, rather than external validation or material wealth. Neglecting this leads to burnout, moral compromises, and a disconnect from purpose.
This document outlines a comprehensive and opinionated framework for reviewing engineering plans. It emphasizes challenging scope, assessing architectural and code quality, ensuring thorough testing, and evaluating performance. The framework prioritizes explicit recommendations, trade-off analysis, and iterative user feedback to optimize engineering outcomes.
Y Combinator (YC) is actively combating California's increasingly hostile policies towards the tech industry, such as the proposed 'billionaire tax' and increased gross receipts tax, which have driven businesses and capital out of the state. YC's president, Garry Tan, argues these policies threaten the Golden State's role as a global tech leader. Concurrently, YC is refining its accelerator program, re-emphasizing in-person collaboration, and pushing a 'founder mode' mentality, inspired by leaders like Airbnb's Brian Chesky, to foster resilience and rapid iteration among startups. This strategy aims to increase YC's success rate and ensure founders are equipped to navigate challenges beyond just securing funding, despite facing a high volume of rejections that underscore the intense competition and resilience required in the startup world.
Gary Tan, CEO of Y Combinator, discusses the critical role of the tech community in San Francisco's political and social revival. He highlights the impact of engagement in local politics on issues like public safety, education, and housing, advocating for a more integrated tech presence within the city's broader community to foster widespread prosperity.
The novel "The Veil" offers a narrative framework to explore entrepreneurial realities within a fictional context. It delves into the cyclical nature of startup failure, the importance of intellectual property protection, and the critical role of ethical considerations in AI development. The story also highlights how fictional narratives can make complex technological concepts accessible to a broader audience, demonstrating the evolving landscape of AI comprehension.
Garry Tan uses Jiro Ono's sushi mastery as a framework for startup product craft, arguing that compounding micro-improvements — not marketing — is the primary driver of perceived quality and durable competitive advantage. The "Pull, Play, Outlast" model maps customer urgency, founder flow state, and sustained commitment as the three necessary conditions for building something transcendent. Zero-bug discipline (per Carmack and Gates-era Microsoft) is positioned as a structural prerequisite, not a nice-to-have. The core thesis: fusing intrinsic motivation with externally painful problems creates a self-reinforcing loop where compounding does the competitive heavy lifting.
Garry Tan argues that a founder's unexamined psychology — particularly Jung's concept of the "shadow" and its positive counterpart, the "golden shadow" — is an underweighted variable in startup outcomes. Suppressed strengths (the golden shadow) manifest as imposter syndrome, conflict avoidance, and deferred leadership, quietly degrading decision quality and team dynamics. Tan's own failure to claim a CEO-level leadership identity at Posterous contributed to the company's stagnation and eventual acqui-hire by Twitter. He posits that systematic inner work — therapy, coaching, and deliberate practice of admired traits — is a strategic lever, not a soft-skills afterthought.
Y Combinator (YC) is experiencing an unprecedented surge in AI-centric startups, with 70% of current batches focused on AI. This boom is attributed to the transformative capabilities of models like GPT-4, which enable practical, revenue-generating applications. YC continues to back highly technical founders, emphasizing direct customer engagement and rapid iteration to capitalize on this dynamic market.
Y Combinator's 3-day onboarding retreat at Fairmont Sonoma is designed to establish deep trust among founders before the batch begins — a deliberate infrastructure investment that Garry Tan argues is YC's primary competitive differentiator. While YC freely distributes all its tactical knowledge (PMF frameworks, startup mechanics) via public channels, the irreproducible asset is a high-trust network where founders share source code, business secrets, and hard-won market insights with peers they treat as co-conspirators. The retreat combines structured activities (pitches, partner sessions, unconference talks) with psychological safety engineering — giving founders a rare space to be candid rather than performative. Tan's thesis: knowledge is a commodity; trusted networks among exceptionally capable people are not.
Post-1971 Nixon shock ended dollar-gold convertibility, unleashing fiat money printing that inflated money supply by 18% in 2020 alone, from $15.3T to $18.7T. This created trillions in idle corporate cash—$17T+ in commercial deposits—that banks reject and firms like Google ($115B hoard) struggle to deploy due to lack of high-return ideas. Tan argues this zero-interest era is ideal for entrepreneurs to channel capital into startups solving real problems in climate tech, robotics, fintech, and beyond, fostering exponential value creation.
Garry Tan critiques the advice to constantly upgrade peer groups to smarter, more successful people as sociopathic networking that repels true creators. Communities of creators (geeks) who riff off each other attract fanatics for support, mature into subcultures, and outperform sociopath-driven groups through sustained innovation. Post-mainstream success, prioritize grounded, authentic relationships with diverse viewpoints over status-chasers to maintain upward momentum without losing core humanity.
Brian Armstrong recounts founding Coinbase after reading the Bitcoin whitepaper, driven by Airbnb fraud prevention insights into global payment inefficiencies and a vision for a unified, bureaucracy-free money system akin to the internet. He prioritized public identity, legitimacy indicators like professional design and real contact info from day one, and bootstrapped solo to achieve product-market fit despite initial self-doubt. Crypto evolves in phases: digital gold (Bitcoin), DeFi financial primitives, and internet-scale apps via DAOs enabling tokenized research, customer-owned firms, and forkable governance to counter centralization risks.
Garry Tan invested $300,000 in Coinbase in 2012 after receiving 0.05 BTC from founder Brian Armstrong, turning it into $2.4B amid Bitcoin's market cap surging from under $1B to over $1T. His decision stemmed from prior frustration with poor Bitcoin buying UX like Mt. Gox, preparedness from fiat currency skepticism via Peter Thiel's hedge fund reading, and Armstrong's proven engineering prowess from Airbnb anti-fraud. Key lessons emphasize contrarian foresight on fringe ideas, builder autonomy to minimize risks, and compelling outreach like sending free value to capture attention.
Wealth consists of valuable goods, services, and scalable products, not fiat money which merely accounts for economic energy. True wealth creation demands rare, high-leverage skills in engineering, design, or product—roles closest to production—enabling ownership of systems over mere participation. Leverage transcends time constraints via products that generate value autonomously, fueled by virtuous cycles of capital, talent, and customers in positive-sum games like startups.
Founders must treat raised capital as an investment with fiduciary duty, justifying every expense by its expected ROI and payback timeline. Develop a 24-month operating plan modeling net revenue growth (targeting 3x/year post-Series A), costs, and cash flow, grounded in input variables like leads and conversion rates rather than arbitrary extrapolation. Prioritize retention (e.g., 73% user retention at 6 months via cohort analysis; high NDR for enterprise) over growth to avoid a leaky bucket, and prepare contingency strategies like achieving default alive, layoffs, pivots, or investor bridges if growth stalls.
Remote interviews better simulate actual remote work by assessing Zoom proficiency and online communication, expanding talent pools 10-100x beyond coastal hubs. Inter-rater reliability in interviews matches movie critic agreement levels, underscoring noise as the primary challenge; brain teasers amplify this due to prior exposure risks, while continuous tasks like implementing Minesweeper provide observable processes without binary gotchas. Triplebyte Screen offers free, background-blind coding assessments to filter inbound applicants by skills, boosting diversity and efficiency in the current buyer's market for talent.
Successful founders invert common biases by adopting short-term pessimism to rigorously mitigate risks and long-term optimism to endure the decade-long grind of building a business. LSE research shows optimistic entrepreneurs earn 30% less post-startup than pessimistic ones, as short-term optimism blinds them to flaws while pessimism prompts proactive fixes via user feedback and criticism. This mindset, echoing Andy Grove's paranoia and the Stockdale Paradox, channels anxiety into action, enabling survival past the 96% failure rate in 10 years.
First principles thinking breaks societal conformity by starting from undeniable truths, enabled by curiosity (e.g., five whys analysis), empathy (direct user immersion), and serenity (accepting uncontrollable factors). Airbnb founders gained conviction from hosting users personally, scaling to $100B+ valuation. DoorDash achieved 50% US market share by retreating to profitable suburbs against better-funded rivals like Uber, funding urban dominance to $55B+ valuation. These traits enable outmaneuvering conventional paths in startups.
Traditional mentorship approaches are often ineffective. A more fruitful strategy involves proactively identifying individuals or organizations one believes in, and then contributing one's unique skills without explicit requests for mentorship. This "give first" approach can lead to unexpected opportunities and organic mentorship relationships, as demonstrated by the author's entry into Y Combinator.
Talented software engineers, designers, and builders must actively increase career risk to avoid stagnation, as the primary risk is inaction amid software's expanding role across sectors. Three paths elevate risk: internal entrepreneurship at big firms (shifting from "police" to "commando" roles), joining passion-aligned startups via targeted searches and cold emails, or founding a startup after building experience. Evaluate opportunities on learning (e.g., building from scratch, shipping to scale, team management) versus earning (salary + normalized equity value using % ownership, valuation, vesting), then apply Bezos' regret minimization framework by projecting to age 80.
A CoffeeScript React class component uses a HistoryJSMixin to manage browser history. The search method triggers an asynchronous index search, and upon receiving results matching the current state, updates the component state with search hits while invoking saveState to persist the search via a URL like '/foo?bar=foo'. This enables bookmarkable, shareable search states without page reloads.
Garry Tan adapts Leif Denby's HistoryJSMixin to sync React state changes with browser history via pushState, ensuring URLs are durable and shareable. Key modification passes explicit URLs to History API calls on lines 57 and 61. Enables faceted search (e.g., Algolia) and pagination without React Router overhead.
LazyGenerate is a Ruby module that extends ActiveRecord classes to define lazily generated properties. Accessing an unset property invokes a user-specified generator method, assigns the result, auto-saves the record, and returns the value. Existing values bypass generation, enabling efficient deferred computation with persistence guarantees.