SUMMARY

Alan Wells, CEO of Rocketable, pitches to Indie VC's Bryce Roberts an AI-driven holding company automating small SaaS acquisitions for hyper-efficiency, envisioning a one-person billion-dollar operation.

STATEMENTS

• Alan Wells is building Rocketable to acquire and automate small SaaS businesses using AI for maximum efficiency.
• He became convinced of AI's potential after feeding an entire app's source code into an LLM, resolving issues in seconds that took hours manually.
• Initially skeptical of LLMs due to his self-driving car experience at Uber and Cruise, Alan shifted after seeing rapid improvements in AI models.
• The golf app he bought generated cash flow but required constant manual effort, highlighting the need for automation.
• Alan realized that source code serves as the true knowledge base for software, outperforming traditional documentation.
• Off-the-shelf AI tools like Intercom's agent failed because they relied on incomplete human-written articles rather than code.
• By integrating source code and user data into Gemini's 1M token model, Alan achieved nuanced, accurate customer support responses.
• He envisions automating product management by analyzing customer feedback to generate specs, wireframes, and A/B tests automatically.
• Marketing automation involves AI-generated short-form videos using tools like Viva Labs for probabilistic testing on platforms like TikTok.
• Rocketable targets SaaS businesses with high customer counts and low ACV to maximize data for AI-driven insights.
• The J-curve of automation means initial productivity dips but leads to exponential gains after optimization.
• Supervised mode with human-in-the-loop is essential for capturing failures and iterating to superhuman performance.
• Alan's manifesto sparked reactions from skeptics doubting AI's readiness to supporters quietly endorsing efficiency.
• Labor market shifts will favor builders who own automated assets over traditional employees.
• Startups now prioritize revenue-per-employee metrics over headcount growth.
• Constellation Software's model of acquiring hundreds of businesses inspired Alan, but AI lowers the viable acquisition size threshold.
• Automating deal sourcing, diligence, and underwriting via APIs could enable acquiring thousands of small SaaS firms.
• Sellers of sub-$5M ARR businesses have limited liquidity options, making Rocketable an attractive buyer.
• Alan plans a vertically integrated holding company that builds its own AI platform for portfolio operations.
• AI diffusion in software will accelerate, but organizational challenges slow widespread adoption.
• Junior engineers should focus on building independent projects rather than seeking traditional jobs.
• Funding strategies now emphasize lean teams achieving $10M-$100M ARR without scaling headcount.
• Efficiency discussions carry political baggage, leading to private endorsements over public ones.
• Niche, handcrafted software will persist but remain marginal in an AI-dominated future.

IDEAS

• AI enables solo entrepreneurs to manage multiple SaaS businesses by treating source code as the ultimate knowledge base.
• Rapid LLM improvements contrast sharply with the slow progress in self-driving cars, signaling a paradigm shift.
• Feeding entire codebases into models unlocks instant debugging and explanation, democratizing complex software understanding.
• Automation's J-curve demands upfront investment but yields 100x efficiency once optimized.
• Custom workflows integrating code, user data, and feedback loops outperform generic AI tools.
• Product development can become probabilistic: generate multiple feature variations, A/B test via AI, and iterate on results.
• High-volume customer data from low-ACV SaaS fuels automated insights for feature prioritization.
• AI avatars for marketing videos allow cheap, scalable testing of content variations.
• Holding companies like Rocketable can acquire micro-SaaS (under $5M ARR) profitably only with AI automation.
• Sellers benefit from API-based underwriting, skipping manual financial prep for instant offers.
• Supervised human-in-the-loop modes prevent errors while building datasets for continuous improvement.
• AGI skeptics undervalue base model intelligence; the gap lies in workflow orchestration, not raw capability.
• Efficiency mantras risk brand backlash, fostering underground support among practitioners.
• Junior talent should pivot to indie building, as AI shrinks entry-level software roles.
• Constellation Software's permanent-hold strategy scales better with AI, targeting 10,000+ niche products.
• Future SaaS favors tailored niche apps over monoliths, exploding the number of viable acquisitions.
• Automating negotiations and diligence could make one-person deals at massive scale feasible.
• Owning automated assets insulates against job loss; employees face diffusion delays.
• Organizational politics hinder AI adoption, creating opportunities for agile solo operators.
• Manifesto reactions reveal a zeitgeist of hyper-efficiency obsession among ambitious founders.
• Vinyl-like "handcrafted" software niches will coexist with AI efficiency but won't dominate.
• YC's environment pushes "what if it works" thinking, overriding small-scale temptations.
• Labor diffusion in software outpaces other sectors, but full job displacement takes years.
• Vanilla metrics like headcount growth are obsolete; lean operations define success.
• Private messages endorsing efficiency highlight a cultural shift toward quiet ambition.

INSIGHTS

• AI's true leverage emerges not from base models but from bespoke integrations of data and feedback loops.
• Source code as ground truth revolutionizes knowledge management, bypassing inefficient human documentation.
• The automation J-curve filters committed builders, rewarding those who endure initial setbacks with outsized gains.
• High-customer-volume SaaS maximizes AI's data-hungry nature for self-improving operations.
• Probabilistic product development via AI variants accelerates innovation beyond human bandwidth.
• Vertically integrated holding companies exploit AI to lower acquisition thresholds, capturing untapped micro-markets.
• Supervised modes bridge the gap from subpar to superhuman AI, mirroring successful engineering paradigms.
• Efficiency narratives carry sociopolitical weight, demanding strategic public positioning.
• Ownership of automatable assets future-proofs against labor disruptions more than employment does.
• Niche software proliferation will fragment markets, favoring acquirers of tailored solutions.
• API-driven diligence streamlines transactions, aligning buyer precision with seller convenience.
• Organizational inertia slows AI diffusion, amplifying advantages for lean, founder-led entities.
• Indie building empowers juniors in an AI-constrained job market, inverting traditional career paths.
• Manifestos catalyze private alignment but risk public polarization in efficiency discourse.
• Permanent compounding portfolios thrive on AI's ability to operate infinite small-scale businesses.
• Cultural shifts prioritize revenue density over expansion, redefining startup maturity.

QUOTES

• "We’re obsessed with the idea that someone is going to build a business doing $100M in revenue with less than 10 employees."
• "AI is writing 90% of the code... a billion dollar software company with one employee."
• "That had to with just blown your mind... the quality of the answers the completeness of the answers was just incredible."
• "The ground truth actually the code that drives the app and the data that is user specific."
• "I'm going to be my own customer... I don't want to sell software so I'm going to buy software."
• "If one guy in his home office can destroy the economy then we've calibrated our safety settings on these models all wrong."
• "The role models in the startup business today are the companies that are growing to 10 million 100 million AR on very small teams."
• "There's a very significant undercurrent of how do you squeeze the most efficiency out of anything these days."
• "The genie is out of the bottle... we're entering this completely unknown unknown."
• "As soon as that can possibly happen sign me up... I'm going to go ride my mountain bike and go trail running."
• "Constellation Software... they've acquired more than 500 businesses and only sold one of them."
• "Small Niche products... are going to steal market share from Big monolithic products."
• "You want to be an owner of something... if you can own this thing and it gets automated that's great for me."

HABITS

• Regularly input entire source codebases into LLMs to query and understand app functionality.
• Maintain a supervised human-in-the-loop mode for AI tools to capture and iterate on failures.
• Analyze customer support emails and cancellation data daily to prioritize product features.
• Generate multiple script variations for AI avatars in marketing videos and test probabilistically.
• Read long-form essays like "Situational Awareness" to build conviction on AI trends.
• Spend spare time listening to AI podcasts while driving to deepen rabbit-hole exploration.
• Export and rebuild P&Ls from API data for accurate underwriting of acquisitions.
• Write manifestos or essays to clarify and articulate automation experiences.
• Tinker with APIs and custom workflows during off-hours to prototype automations.
• Record personal video footage on golf courses to train realistic AI avatars for content.

FACTS

• Self-driving cars at Uber took 10 years from demo to commercial viability.
• LLMs improved dramatically every week or month, unlike incremental self-driving progress.
• Constellation Software has acquired over 500 businesses, selling only one.
• There are 20,000 to 30,000 SaaS companies with $500K-$5M ARR today.
• Intercom's AI agent requires knowledge base articles, leading to errors without them.
• Google's Gemini 1M token model became available about three months after Alan bought his app.
• YC Demo Day features pitches like one-person billion-dollar AI companies.
• Software automation could impact 0.5% of US GDP via SaaS efficiency.
• Small SaaS owners often lack liquidity options below $5M ARR.
• TikTok outperforms other platforms for B2C mobile app user acquisition.
• AI video generation costs a few bucks per short-form video versus $10,000 for traditional production.
• Situational Awareness essay by Leopold Aschenbrenner was published in June but read by Alan in October.

REFERENCES

• Rocketable: AI-maximalist software holding company for acquiring and automating SaaS.
• Indie VC: Investment firm partnering with lean, efficient founders.
• Golf app: Acquired SaaS business for cash flow, automated with AI.
• Uber self-driving truck: 2017 project Alan worked on.
• Cruise: Autonomous vehicle company where Alan was an alum.
• Situational Awareness essay series: By Leopold Aschenbrenner, 120 pages on AI trends.
• Intercom AI customer support agent: Off-the-shelf tool that underperformed.
• Google Gemini 1M token model: Used for source code analysis.
• Viva Labs: YC startup for AI-generated short-form video content.
• Craft Full: Tool for extracting customer insights from support emails.
• Overcast podcast app: Handcrafted by Marco Arment as a premium example.
• Constellation Software: Model for permanent-hold acquisitions, 100-150 per year.
• X (Twitter): Platform where Alan posted manifesto and saw reactions.
• YC Demo Day: Event where Alan pitched.
• David Pell's class: Where Alan wrote essays on automation.
• O1 reasoning model: Released around time Alan read Situational Awareness.
• Stripe API: For read-only financial access in underwriting.
• Apple App Store API: For revenue data verification.
• Facebook Advertising API: For cost analysis in acquisitions.

HOW TO APPLY

• Identify bottlenecks in your SaaS operations, like customer support, and assess AI automation potential using source code as context.
• Export your app's full codebase into a large-context LLM like Gemini to query features and data explanations instantly.
• Implement supervised mode for AI tools, reviewing outputs to log failures and build improvement datasets.
• Analyze high-volume customer interactions via tools like Craft Full to extract feature requests and prioritize development.
• Generate probabilistic product specs by feeding LLM with support data, creating wireframes, and simulating A/B tests.
• For marketing, record base footage for AI avatars and use platforms like Viva Labs to produce and test video variations on TikTok.
• Target acquisitions in low-ACV, high-customer SaaS niches where automation generalizes across common workflows.
• Underwrite deals with API access to Stripe and app stores, rebuilding P&Ls to offer quick, data-driven valuations.
• Structure as a holding company: build custom AI platform first, then integrate bought businesses for compounding efficiency.

ONE-SENTENCE TAKEAWAY

Embrace AI automation to build a lean, scalable software holding company owning thousands of niche SaaS assets.

RECOMMENDATIONS

• Prioritize SaaS acquisitions under $5M ARR with high customer counts for AI data leverage.
• Build custom workflows integrating source code and user data before relying on off-the-shelf tools.
• Adopt a J-curve mindset: invest heavily upfront in automation despite initial productivity dips.
• Use supervised human-in-the-loop for all AI deployments to safely iterate toward superhuman results.
• Shift career focus to owning automatable assets, avoiding junior roles vulnerable to AI displacement.
• Generate and test multiple AI content variations probabilistically to optimize marketing ROI.
• Read trend essays like Situational Awareness to gain conviction on AI timelines and capabilities.
• Automate diligence with API access for faster, more accurate deal processing.
• Target niches where workflows generalize, avoiding domain-specific operations like actuarial services.
• Publicly manifest bold efficiency visions while anticipating private endorsements from peers.
• Rethink funding: aim for $10M-$100M ARR on minimal teams, ignoring headcount vanities.
• Prepare for fragmented markets by acquiring tailored niche products that erode monoliths.
• Tinker daily with LLMs in personal projects to stay at the bleeding edge of tooling.
• Foster YC-like ambition: evaluate opportunities by "what if it works" potential, not certainty.
• Position as a top buyer for small SaaS sellers by simplifying liquidity through tech-enabled processes.
• Monitor labor diffusion: accelerate personal upskilling in AI orchestration over traditional skills.
• Balance efficiency discourse to mitigate political baggage, emphasizing builder benefits.

MEMO

In the sunlit buzz of San Francisco's startup scene, Alan Wells, a former Uber and Cruise engineer, stepped into a pivotal meeting with Bryce Roberts of Indie VC. What began as a casual DM sparked by Wells' bold tweet—envisioning a one-person, billion-dollar AI software empire—evolved into a live-recorded pitch for Rocketable. Wells, fresh from Y Combinator, outlined his vision: a holding company acquiring and automating small SaaS businesses, leveraging AI to slash operational needs. His journey started modestly with a golf app purchase, but a eureka moment came when he fed its 100,000-line codebase into Google's Gemini model, decoding complexities in 90 seconds that once devoured hours.

Skepticism defined Wells' early AI stance, scarred by a decade grinding on self-driving cars that inched forward without breakthroughs. LLMs shattered that doubt; unlike autonomous vehicles' plodding evolution, models improved weekly, fueled by massive pre-built investments. "Somebody else has already spent the $50 to $1 billion in capex," Wells noted, allowing solo founders like him to tap superpowers for pennies. This shift "pilled" him on AGI, reinforced by Leopold Aschenbrenner's "Situational Awareness" essay, which mirrored his lived acceleration toward superhuman capabilities in single-digit years.

The golf app's spaghetti code epitomized pre-AI drudgery: tracing nested logic for customer queries. Off-the-shelf fixes like Intercom's AI agent flopped, vomiting errors from absent knowledge bases. Wells' unlock? Treating source code as ground truth, blending it with user data for founder-like responses. Relief mingled with excitement—this wasn't just relief; it was a builder's mandate. Custom APIs bridged the gap where products faltered, echoing his Cruise days optimizing data loops for self-driving maneuvers.

Automation's J-curve loomed large: early efforts tank productivity as flawed bots demand supervision. Yet Wells, battle-tested, pushed through, climbing toward exponential gains. Customer support conquered, he eyed product management next—parsing emails for pain points, auto-generating specs, wireframes, and A/B variants. High-customer SaaS, even low-ACV, shines here, providing data rivers for quantitative loops. "The line of sight is so clear," he said, though tools lag in stitching it end-to-end.

Marketing beckons with AI flair: Viva Labs turns app screenshots and Wells' golf-course footage into avatar-driven TikTok videos, testing dozens cheaply versus traditional $10,000 shoots. Probabilistic wins emerge from volume, aligning with his manifesto’s roar—a deliberate provocation declaring high automation already viable, not marginal. Reactions split: skeptics decried models' unreadiness, doomsayers feared economic ruin, yet practitioners whispered approval, posting anonymously. "If one guy in his home office can destroy the economy," Wells quipped, "we've calibrated wrong."

Labor tremors ripple outward. Software's AI diffusion outpaces others, but organizational politics brake it. Hiring stalls; lean teams hitting $100M ARR redefine success, ditching headcount vanities. Juniors must build indie, not chase vanishing roles. Wells' holding company, inspired by Constellation Software's 500+ acquisitions (one sold, regretted), targets micro-SaaS untapped by private equity. AI drops viability to $500K ARR, eyeing 10,000 niches stealing share from monoliths.

Liquidity beckons sellers starved of options below $5M. Wells' pitch: API access to Stripe and app stores for instant, rebuilt P&Ls and offers—no spreadsheets. This vertically integrated beast builds its platform first, automating sourcing to operations. "What if it works?" echoes YC's ethos, pulling Wells from passive income dreams to bleeding-edge ambition. Owning automated assets trumps employment; he'd bike trails post-automation.

As cameras rolled, Roberts committed, sealing a week's whirlwind from pitch to partnership. Wells' artifact captures a zeitgeist: hyper-efficiency's undercurrent, genie unbound. Handcrafted software niches linger like vinyl, but mass markets tilt AI. For builders sharing his worldview, the call rings clear—reach out, automate boldly. In this unknown, one engineer's swing redefines flourishing.