Declares that AGI is functionally here. Long-horizon agents — AI systems that can sustain multi-step work, correct their own errors, and persist toward goals autonomously — are the realization of AGI. Coding agents are the first proof point. One litmus test: can you hire an agent? In 2023-2024 AI apps were chatbots; in 2026-2027 they will be doers that feel like colleagues. Usage shifts from a few queries per day to all-day, every-day autonomous work. Human roles shift from executor to manager of AI teams.
AI has reached its 'synesthesia moment' — models that natively understand and generate across modalities (text, image, code, video, audio, voice) in a unified latent space. AI synesthesia converts strengths in one cognitive domain into capabilities in another: if you write well but cannot code, AI bridges the gap through semantic representations; if you design beautifully but struggle to pitch verbally, AI transforms sketches into narratives. Creativity becomes translation, expression becomes multidimensional, and intelligence becomes fluid.
If 2024 was the 'primordial soup' year for AI, the building blocks are now firmly in place. Data centers are the new rails of the digital economy and will be securely built by end of 2025. Five finalists emerged from the big model race (Microsoft/OpenAI, Amazon/Anthropic, Google, Meta, xAI). AI search will proliferate (Perplexity hit 10M MAU). The key question shifts from 'can we build it?' to 'what freight will ride on those rails?'
AI is progressing from 'thinking fast' (rapid pre-trained pattern matching, System 1) to 'thinking slow' (deliberate reasoning at inference time, System 2). The reasoning layer — inspired by AlphaGo-style approaches — endows AI with problem-solving that goes beyond pattern recognition. This enables 'service-as-a-software,' where the addressable market expands from the $1T software market to the $10T+ services market. Foundation layer has stabilized to five scaled players. Reasoning models are strong on logic-proximate domains (coding, math, science) but still developing on open-ended domains (writing, strategy).
Identified a massive gap between the revenue expectations implied by the AI infrastructure build-out (projected from NVIDIA's data center revenue run rate) and actual revenue growth in the AI ecosystem. The AI industry needs to generate $600 billion annually in revenue to justify current hardware spending levels — raising the question of whether this is sustainable or a bubble.
Act 1 came from the technology-out — foundation models as a new hammer generating a wave of novelty apps. Act 2 must come from the customer-back, solving real human problems end-to-end. The biggest challenge is not finding use cases but proving lasting value — retention, not novelty. Invoked Amara's Law: we overestimate technology in the short run and underestimate it in the long run. As foundation models commoditize, the real value shifts to the application layer (product, UX, workflows). Published alongside the V3 generative AI market map.
Introduced the generative AI market map and thesis. Argued that a new class of AI had emerged that creates rather than merely analyzes — shifting the marginal cost of creation toward zero. Every industry requiring original human work (coding, design, law, marketing, gaming) is up for reinvention. Outlined four waves of development from transformer breakthroughs to killer app emergence. Became the foundational reference document for the entire generative AI ecosystem.
In crypto, value accrues at the protocol layer rather than the application layer (originally by Joel Monegro, but Olaf popularized the counter-thesis).
Framework for evaluating consumer companies: Level 1 is growing engaged users around a 'core action,' Level 2 is retaining users through accruing benefits and mounting loss, Level 3 is self-perpetuation through network effects and virtuous loops.
Updated version clarifying how the framework impacts product roadmaps, with added layers on what makes a core action strong and how to measure retention meaningfully.
Reaching a new happiness threshold where you're so much better than any substitute that the market 'tips' in your direction. Find scalable, systematic ways to create better buyer-seller matches over time.
Three vectors to domination: outrun to become #1 in your original market, expand buyer use cases beyond your initial thimble, and pursue multi-threaded domination of adjacent markets.
To measure Happy GMV, ask 'what is my best guess at the buyer and seller experience that will lead to retention?' then measure the percentage that get that experience. NPS is not a useful proxy.
Create a 10x product and recast incumbent cost structures. The combination of dramatically better experience at dramatically lower cost is the formula for enduring businesses.
For the past 25 years, software startups sold productivity improvements. AI enables selling the completed work itself — a 95% improvement instead of 15%. Incumbents are stuck selling software; startups can leapfrog.
LLM-based startups can build enduring value by accruing data assets as a positive externality of users using the application. This externalizes the moat beyond what's possible with LLMs alone.
A new wave of AI startups (DeepL, HeyGen, ElevenLabs) unlock new use cases by providing 10x better experiences at a fraction of the cost, recasting entire cost structures.
Companies like Midjourney, DeepL, and ElevenLabs demonstrate how AI creates entirely new markets by making previously expensive capabilities accessible and dramatically better.
Explores whether AI will produce as many breakthrough consumer companies as the mobile revolution did, concluding there will be countless consumer AI companies more akin to Uber or Google Maps that give consumers new superpowers.
As foundation models become more powerful, LLM companies will move up the stack and compete with their API developers. B2B AI startups need to build moats beyond API access.
Being distributed at the early stages is like sprinting with a parachute strapped to your back. Collaboration, trust, creativity from hallway conversations are all highest in person.
Money without context doesn't work. The next wave of fintech will be about contextual financial services that understand who you are, what you need, and when you need it — not one-size-fits-all products.
Annual letters to LPs that outline the state of global fintech disruption. Known for being deeply thoughtful about macro trends in financial services and regulatory evolution.
Weekly newsletter curating the best in science, technology, and deep tech. Cult following among deep tech enthusiasts. Each edition weaves together threads from scientific papers, defense developments, manufacturing breakthroughs, and frontier technology. Co-authored with team.
Science only moves in one direction — forward. The key to deep tech investing is identifying which 'arrows' are moving fastest and betting on founders who are riding them. You can't uninvent knowledge — so technical breakthroughs create permanent, irreversible advantages.
The best investments happen when consensus narrative is about to be violated by reality. Look for moments where everyone believes X but evidence points to Y. The gap between narrative and reality is where alpha lives.
The definitive guide to scaling startups from Series B to IPO. Covers CEO role evolution, board management, late-stage fundraising, M&A, product management at scale, and hiring executives. Based on interviews with leaders at Airbnb, Stripe, Dropbox, Twitter, and others. Free online and in print.
AI is the most significant platform shift since mobile. Every industry will be transformed. The biggest opportunities are in AI infrastructure and in vertical applications where AI can replace entire workflows, not just augment them.
Analysis of technology market cycles and when to invest aggressively vs. defensively. The best companies are built during downturns because talent is available and competition is reduced.
Framework for thinking about what makes companies defensible. Network effects, switching costs, technical complexity, and regulatory moats. Not all moats are created equal.
The co-founder relationship is the most important decision a startup founder makes. Complementary skills matter less than aligned values, work ethic, and communication style.
Being right about a market but wrong about timing is the same as being wrong. The best investors and founders have an instinct for when a market is ready — not too early, not too late.
Series of posts analyzing the AI landscape — foundation model economics, open vs. closed models, infrastructure layer opportunities, and where value will accrue in the AI stack.
Fireside Chat: Technology Will Replace 80% of What Doctors Do
AI and technology will replace the vast majority of what doctors currently do — diagnosis, treatment recommendations, monitoring. The remaining 20% (empathy, complex judgment) will be augmented. Healthcare costs will drop 10x.
Startups should 'engineer their gene pool' by carefully selecting early team members, advisors, and board members. The founding team's DNA determines the company's trajectory. Bring in expertise through advisors, not just hires.
Experts in a field are often the worst predictors of disruption in that field because they're anchored to existing paradigms. The best innovations come from outsiders who aren't constrained by 'how things are done.' This is why young, non-expert founders often outperform industry veterans.
Reinventing Societal Infrastructure with Technology
Every major societal system — energy, healthcare, education, food, finance — was designed for a pre-technology era. All of them will be completely reinvented using AI and other technologies within 20-30 years.
Fusion, superhot geothermal, and advanced nuclear will provide virtually unlimited clean energy by the 2030s-2040s. The energy transition is not a sacrifice — it's an upgrade. Clean energy will be cheaper than fossil fuels even without subsidies.
AI: The Most Important Technology in Human History
AI will be more transformative than electricity, the internet, or any prior technology. It will replace 80% of jobs in 80% of occupations. This is not a crisis — it's an opportunity to reimagine work, education, and human purpose.
One of the longest-running and most influential VC blogs. Fred wrote almost daily for 20+ years covering startup ecosystems, investing philosophy, crypto, climate, music, and NYC tech. Transitioned to avc.xyz. Known for engaging comment sections that became a community. The blog itself embodied the 'network' thesis.
USV publishes its investment thesis publicly. Core: 'large networks of engaged users, differentiated through user experience, and defensible through network effects.' Updated over time to include crypto/web3 and climate.
Reflection on how USV's thesis evolved from 'large networks of engaged users' to 'trusted brands that broaden access' — the through-line is always access and networks.
Fred frequently references Carlota Perez's framework on technological revolutions — installation phase vs. deployment phase. Argued that crypto was in its installation phase and the deployment phase (real utility) was coming.
Multiple posts over years making the case for crypto/blockchain as the next computing platform. Argued for decentralization as both a technology architecture and a political philosophy. Maintained conviction through multiple crypto winters.
Long-running weekly series explaining business concepts: cap tables, term sheets, burn rate, unit economics. Became essential reading for first-time founders. Later compiled into a book.
The foundational essay for a16z's American Dynamism practice. Argues that Silicon Valley spent a decade building consumer social apps while ignoring the hard infrastructure, defense, and industrial problems that actually matter for the country. The pendulum is swinging back — the most ambitious founders now want to build things that strengthen America.
Optimistic case for American renewal through technology. Argues that a new generation of founders is choosing to build defense companies, infrastructure companies, and manufacturing companies — not because it's trendy, but because it matters.
The best founders are 'builders' in the deepest sense — they want to create things that exist in the physical world, that defend the country, that house people, that power cities. This is a return to the builder ethos that created American prosperity.
Defense budgets are shifting toward technology. Government procurement is modernizing. A new generation of defense founders (many with military experience) are building companies that can move at startup speed while meeting government requirements.
Annual list of the 50 most important American Dynamism companies — spanning defense, infrastructure, housing, energy, public safety, and manufacturing. Signals where the fund sees opportunity.
Prolific video content on startups, investing, product design, AI, and San Francisco. Videos range from startup advice to political commentary to interviews with founders. One of the most-followed VC YouTube channels.
AI enables a single person or tiny team to build what used to require 20 people. This means the unit economics of startups have fundamentally changed. YC is seeing solo founders and 2-person teams building products of remarkable quality. The implication: smaller teams, faster iteration, higher per-person output.
YC's official podcast, hosted by Garry and YC partners. Deep discussions on startup building, AI, market trends, and interviews with successful YC founders. Covers both tactical advice and big-picture thesis.
Unconventional Advice for Founders (Stanford Talk)
Advice that goes against conventional startup wisdom: sometimes you should be a solo founder, sometimes you should bootstrap first, taste matters more than experience, and the best founders are often the ones who break the rules.
How AI tools are changing the startup building process. You can now prototype in hours what used to take weeks. This changes who can be a founder and how fast you can validate ideas.
Prolific Twitter presence with threads on product design, startup building, SF politics, and the startup ecosystem. Known for being direct and opinionated.
Zero to One: Notes on Startups, or How to Build the Future
The intellectual foundation of Founders Fund. Key ideas: (1) Going from 0 to 1 (creating something new) is fundamentally different from going from 1 to n (copying what works). (2) Competition is for losers — the goal is to build a monopoly. (3) Every great company is built on a 'secret' — something true that no one else believes. (4) Definite optimism beats indefinite optimism. (5) The best businesses have characteristics of monopolies: proprietary technology, network effects, economies of scale, and branding.
The Stanford lectures that became 'Zero to One.' Notes were taken and published by Blake Masters. Covered monopoly theory, secrets, founder archetypes, distribution, and the importance of definite planning.
Famous and controversial essay arguing that democracy and freedom have become incompatible, and that the future of freedom lies in new frontiers — cyberspace, outer space, and seasteading. Catalyzed significant debate about tech libertarianism.
Published in the Wall Street Journal. Argues that the most successful businesses avoid competition entirely by creating new categories. Companies that compete head-to-head destroy value. The goal is monopoly — and most founders don't think big enough.
Founders Fund manifesto arguing that technological progress has stagnated in the physical world. We have smartphones but not flying cars, fusion energy, or cures for cancer. The venture industry has become too focused on incremental software and not ambitious enough about transforming the physical world.
Philosophical essay arguing that the post-9/11 world requires a new framework for understanding politics and technology. References Leo Strauss, Carl Schmitt, and René Girard. Shows the philosophical depth behind Thiel's investing worldview.
The definitive annual report on the cloud/SaaS industry. Covers market size, growth trends, public and private cloud benchmarks, IPO activity, M&A, and emerging trends. Each edition introduces new frameworks and metrics. The 2024/2025 editions focus heavily on AI's impact on cloud economics.
Interactive benchmarking tool that lets SaaS companies compare their metrics (ARR, NRR, gross margin, growth rate, efficiency) against thousands of cloud companies. Has become the standard reference for SaaS benchmarking.
Legendary page listing every major company BVP passed on — Apple, Google, Facebook, Intel, eBay, FedEx, etc. A masterclass in VC humility. Shows that even the best investors miss deals, and that what matters is the portfolio you build, not the individual misses.
A basket of publicly traded cloud companies used as an industry benchmark. Tracks the performance of emerging cloud companies and provides data on cloud adoption trends.
AI is the biggest inflection point in cloud since the transition from on-prem to SaaS. The next generation of cloud companies will be 'AI-native' — built from the ground up with AI as the core, not bolted on as a feature. This will create a new wave of category leaders.
BVP's framework for evaluating cloud companies. Includes metrics on growth efficiency, retention, unit economics, and market positioning. Widely referenced by SaaS founders and investors.
Unscaled: How AI and a New Generation of Upstarts Are Creating the Economy of the Future
The economy is 'unscaling' — for a century, bigger was better (economies of scale). Now, technology allows small companies to rent scale (via cloud, AI, platforms) and compete with giants. This creates opportunities in every industry for nimble, technology-native upstarts to challenge incumbents. The unscaling thesis explains why Stripe can challenge banks, why Livongo can challenge healthcare systems.
Healthcare needs to shift from 'sick care' (treating illness) to 'health assurance' (preventing illness and maintaining health). Technology — AI, wearables, telemedicine — enables this shift. The current healthcare system is designed around episodes of illness; the future system should be designed around continuous health maintenance.
Technology companies have a responsibility to consider the societal impact of their products. 'Move fast and break things' is irresponsible when the things you're breaking are people's lives, privacy, and livelihoods. Responsible innovation means thinking about second-order effects from the beginning.
The traditional VC model of writing checks and sitting on boards is necessary but insufficient. The best companies need operational support, strategic guidance, talent networks, and sometimes adjacent acquisitions. GC is evolving to provide all of these under one roof.
AI will reduce healthcare costs by 10x while improving outcomes. The technology exists today — the barriers are regulatory, institutional, and cultural. Companies that navigate these barriers will build generational businesses.
One of the best startup knowledge bases on the internet. Long-form, deeply tactical content on hiring, product management, engineering leadership, go-to-market strategy, fundraising, and company building. Based on interviews with experienced operators and founders. Each article is a masterclass. Examples: 'How Superhuman Built an Engine to Find Product/Market Fit', 'The Science of Speaking Up', and hundreds more.
The biggest conversion barrier isn't price — it's the gap between free and any price at all. Going from $0 to $0.01 is harder than going from $10 to $100. This has implications for pricing strategy, freemium models, and product-led growth.
Annual survey of hundreds of startup founders on fundraising, hiring, compensation, diversity, and market sentiment. Provides data-driven insights into what founders are actually experiencing.
Analysis of value distribution in the AI stack — where does value accrue? At the model layer, the infrastructure layer, or the application layer? Argues that the AI platform is still early and the value chain isn't settled. Both infrastructure and applications can capture significant value, unlike previous platform shifts where one layer dominated.
AI compute costs are the defining constraint for AI companies. Successful AI companies need strategies for managing compute costs — whether through more efficient models, inference optimization, or novel architectures. The compute cost curve will determine which business models are viable.
Framework for how AI companies should think about building, deploying, and scaling AI products. Covers model selection, fine-tuning strategy, evaluation, deployment, and business model considerations.
Analysis of consumer AI engagement patterns. Some of the most successful consumer AI products are deliberately constrained — doing one thing exceptionally well rather than trying to be everything. Character.ai's focused approach to AI conversation is an example.
Active Twitter presence sharing insights on seed investing, the NYC startup ecosystem, and founder advice. Less formal than long-form writing but consistently insightful on early-stage dynamics.
Extensive experience running TechStars NYC batches — has worked with hundreds of early-stage founders through the accelerator process. Shares insights on what separates companies that break out from those that don't.
Regular speaker and panelist on the evolution of New York's startup ecosystem. Has firsthand perspective on how NYC went from an afterthought in tech to one of the world's most important startup cities.
The best enterprise companies grow bottom-up. Individual users adopt the product because it's great, it spreads organically through the organization, and eventually the company signs an enterprise deal. This PLG motion produces better products, stronger retention, and more defensible businesses than traditional top-down enterprise sales.
Design and user experience are competitive moats in B2B, not just nice-to-haves. Slack won because it was delightful. Figma won because it was collaborative and beautiful. The next generation of enterprise winners will be the products that users love, not just tolerate.
Read Write Own: Building the Next Era of the Internet
Web 1.0 was read-only (static pages). Web 2.0 was read-write (social media, user-generated content — but platforms own everything). Web 3.0 is read-write-own (blockchain enables users to actually own their digital assets, identities, and communities). The 'own' is the revolutionary part — it changes the incentive structure of the internet from extractive to participatory.
Disruptive technologies always start as toys that seem useless to incumbents. Mainframes → PCs → mobile phones → social media → crypto. The pattern repeats because incumbents judge new technology by current use cases rather than future potential. This is one of the most cited blog posts in startup history.
The best network-effect businesses start with a single-player utility that's useful even without a network. Users come for the tool (Instagram's photo filters, Delicious's bookmarks) and stay for the network that forms around it. This solves the cold start problem.
Centralized platforms follow a predictable arc: attract users and developers → build network effects → extract value from participants. Decentralized protocols can't extract because they're governed by code and communities, not corporations. This makes them fundamentally more trustworthy for long-term building.
Crypto goes through recurring cycles: price increases → media attention → new people enter → some speculate, some build → builders create genuine innovation → innovation creates real utility → utility drives the next price cycle. Each cycle leaves behind permanent infrastructure and capabilities.
NFTs enable Kevin Kelly's '1,000 true fans' theory at scale. Creators can now sell directly to their most passionate fans and capture the full economic value. This is a fundamental shift in the creator economy from ad-supported (platforms extract value) to ownership-supported (creators capture value).
Throughout history, new computing technologies have expanded the creative palette. Each wave (PCs, internet, mobile, crypto, AI) enables new forms of expression that didn't exist before. This is what makes technology a humanistic force.
What smart people do on weekends (hobbies, side projects) predicts what everyone will do in 10 years. The internet started as a weekend project. Bitcoin started as a weekend project. Watch what hackers build for fun — that's where the future is.