# 20VC: What I Learned from 100 of the Best CEOs in the World | What I Learned from Staying with Mr Beast for 3 Weeks | How We Will Spend More on Tokens than Salaries with Cliff Weitzman, Speechify

Podcast: The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
Published: May 10, 2026
Reading time: 28 min
Canonical: https://podbrew.app/briefs/the-twenty-minute-vc-20vc-venture-capital-startup-funding-the-pitch-20vc-what-i-

Cliff Weitzman, the founder and CEO of Speechify AI, joins the 20VC podcast. He built Speechify to overcome his own dyslexia and ADHD, creating a Voice AI agent now used by over 60 million people.

The discussion covers an expansive range of topics, from Cliff's insights gained by connecting with 100 top CEOs and observing Mr. Beast's operations, to his innovative strategies for AI-driven advertising. Listeners will hear about Speechify's approach to hyper-growth, talent management, and the projected shift where AI tokens may soon cost more than employee salaries.

Cliff shares actionable wisdom for any entrepreneur or leader navigating the rapidly evolving landscape of technology and business. His journey provides compelling lessons on resilience, the practical applications of AI, and building a company with a strong, unconventional culture, making this a critical listen for strategic thinking and innovation.

## Key takeaways

- Applying a "volume of work" principle, such as taking more shots or reps, can lead to unexpected and favorable outcomes, as seen in Cliff's college application success.

- Implement a "Warrior" rule, ensuring senior leaders remain hands-on practitioners rather than delegating all execution, and hire "barrels" who are full-stack and can complete tasks independently.

- Companies benefit from committing to distinct 'bulking' (growth) or 'cutting' (profitability) cycles for several months at a time, avoiding the inefficiency of trying to achieve both simultaneously.

- During hyper-growth, strategic ad spending should be viewed as an 'investment' in learning new creative formats and testing emerging channels (e.g., OpenAI ads) to unlock future growth, rather than just increasing blended CAC.

- Speechify tests over 1,300 AI-generated ads daily using a custom platform to discover arbitrage opportunities and reduce customer acquisition costs.

- Leverage extreme visuals and iterative testing to create high-performing ads, as demonstrated by the hot tub ad that drove $3 million in revenue after testing nearly 50 variations.

- Ad performance is unpredictable; a creative concept like a rap song about personal experience can unexpectedly generate significant revenue, while anticipated successes might fall flat.

- AI tools are commoditizing the manual 'hammer and chisel' aspects of design, making 'taste' the more valuable human skill.

- Companies are projected to spend more on AI tokens for engineering than on employee salaries, indicating a fundamental shift in operational costs and resource allocation.

- Prioritize a candidate's Adversity Quotient (AQ), which measures persistence and resilience when facing difficult challenges, over traditional IQ or EQ.

- Assess AQ by looking for tangible evidence like published side projects, demonstrating a willingness to overcome obstacles and grapple with hard problems for extended periods.

- Dyslexia can develop a crucial resilience and tolerance for failure in individuals, making them well-suited for the challenges of entrepreneurship and leading startups.

- Speed is considered the number one strategic advantage, enabling companies to outmaneuver competitors, especially within distributed teams.

- Rapid response times, ideally within 60 seconds for engineers, are crucial to prevent blocking teammates and ensure continuous project momentum.

- The 'Titanic Rule' dictates a three-hour window for any message response, even if it's just an acknowledgment of busyness, to provide clarity and maintain workflow in remote environments.

- Engineers must become 'outcome owners,' actively driving features through all stages, including unblocking PRs and QA, to ensure work is deployed and live.

- QA is the most critical skill in an AI-driven software development landscape, as AI struggles with self-QA and many developers miss user-centric edge cases.

- The 'virus strategy' for growth involves a person or company strategically attaching to a larger, established institution to license its reach and accelerate their own growth.

- Uncontrolled AI inference costs can be detrimental, quickly consuming revenue and making consumer AI companies unprofitable, as Speechify experienced when its services became a "charity" at scale.

- Speechify invested heavily in a dedicated 40-person AI engineering team for over two years to drastically reduce inference costs, achieving a 10x reduction to single-digit dollars per million characters, which was critical for their sustained profitability.

## 04:00 - 08:01 Overcoming Dyslexia and the Genesis of Speechify

Cliff Weitzman, who moved to the US at age thirteen, faced significant challenges with severe dyslexia, which impacted his ability to read throughout his early education. He applied a "volume of work" strategy to his college applications, writing forty-eight drafts of his main essay and applying to many schools, ultimately gaining admission to Brown University despite his academic struggles.

His self-consciousness about spelling due to dyslexia was highlighted during his freshman year of high school. To avoid misspelling names in yearbooks, he developed a strategy of writing generic messages like "you are a great guy." This led to an embarrassing moment at a sleepover where a friend discovered the repeated message and mistakenly believed Cliff was implying he was gay.

Even upon graduating college, Cliff continued to struggle with spelling, crying after finding three errors in a thirty-page philosophy paper. This lifelong battle with dyslexia, particularly his spelling difficulties, directly inspired him to create Speechify. The deep learning-based text-to-speech model, developed in 2015, helps users read faster and identify spelling mistakes by hearing the words, now reading over ten million books worth of words annually.

> Volume times leverage equals output.

## 08:01 - 12:01 How to Learn from Top CEOs and Implement the 'Warrior' Rule for Leaders

Cliff Weitzman describes his method for extracting insights from top CEOs and experts. He cold emails one hundred potential contacts, often getting responses from high-profile individuals like founders of Instagram or Robinhood, and persists with follow-ups down the organizational chart if needed.

Once connected, he builds rapport by discussing favorite books (having read hundreds himself), often leading to opportunities to observe their work directly, such as how they buy Facebook ads or create content. He views growth as an "arbitrage game" where success depends on superior content or distribution strategies.

A key insight Cliff gained is that many senior leaders become "rusty" as individual practitioners. He developed a "Warrior" rule, insisting that all company members, regardless of seniority, remain hands-on. This ensures leaders are not merely "fat generals" but actively engage with the craft, as exemplified by his expectation for a Head of Growth to source candidates personally.

His hiring philosophy emphasizes "barrels, not ammunition"—seeking full-stack individuals who can execute tasks end-to-end, from video creation to editing and posting. The company values passion for the product, team loyalty, and encourages employees to contribute to different codebases to foster continuous learning and adaptability.

> you wanna hire barrels, not ammunition, you know, which is someone who can go from one to ten on their own and do full stack.

## 12:01 - 16:02 Companies Cycle Between Growth and Profitability with Strategic Ad Spend

Companies, much like bodybuilders, operate in distinct 'bulking' (growth) and 'cutting' (profitability) cycles. It is crucial to commit to one cycle for an extended period, typically around six months, as attempting both simultaneously is ineffective. This focused approach prevents conflicting efforts, such as simultaneously trying to accelerate and decelerate company initiatives.

Speechify's journey exemplifies this, taking four and a half years to achieve product-market fit before entering a hyper-growth phase. Following a period focused on profitability, the company is now back in hyper-growth. During these phases, a 'maniacal focus' on the chosen objective—whether growth or profit—is paramount, as splitting attention, even for tasks like fundraising, can hinder progress.

In its current hyper-growth phase, Speechify doesn't indiscriminately burn cash, even if blended customer acquisition costs (CAC) increase. Instead, they view ad spending as a 'creative investment.' For example, they are willing to spend significant amounts on creating new ad content or experimenting with emerging platforms like OpenAI's ad services. This allows them to learn new formats and channels, positioning them for sustained growth when these opportunities become mainstream.

> I'm okay spending five hundred thousand dollars a month making twenty thousand pieces of creative, because that is a creative investment.

## 16:02 - 22:03 Speechify's AI-Powered Ad Creation and Arbitrage Strategies

Speechify implements an aggressive ad testing strategy, creating and testing over 1,300 AI-generated ads daily using a proprietary platform. This massive scale allows them to uncover arbitrage opportunities by finding ads that perform exceptionally well with a low Cost Per Acquisition (CPA) on platforms like Meta, TikTok, and YouTube.

A key tactic involves 'reskinning' successful ad creatives using AI to diversify demographics and settings. For instance, a proven ad featuring the founder's sister, Gaffin, can be altered to show her as a 60-year-old woman, a 40-year-old man, or different ethnicities, placed in various backgrounds like coffee shops or libraries. This broad testing helps identify underserved audience segments, such as users aged 60 and above who benefit from Speechify for reading glasses.

They also employ whitelisting on Instagram, where creators produce multiple videos for Speechify to run as paid ads without posting them organically. The top-performing ads are then posted organically from the creator's account. This method enables extensive testing and optimization of creator content before public release, enhancing performance.

The daily testing process funnels winning ads into main campaigns, allowing them to compete against Speechify's best-ever performing ads. Only the highest-converting ads receive increased spend, continuously optimizing their ad portfolio and maintaining a competitive edge in a saturated market by finding unique ad plays.

> Because it's an arbitrage game, if you use the tool that everybody else uses, typically you're not gonna win. So we just built our own platform.

## 22:03 - 26:04 Cliff Weitzman on How a Hot Tub Ad Generated $3 Million and the Power of Repurposing Content

Cliff Weitzman, founder of Speechify, shared how a single ad drove $3 million in revenue. He analyzed the top 100 best-performing ads in history, rewriting scripts for Speechify. His key insight was the importance of the first image in a video. For Speechify, he created a striking visual: himself in a hot tub, wearing a suit and bright red headphones. The ad then featured a direct and confident script, challenging user expectations about the voice quality.

He created nearly fifty variations of this video, continuously testing them. One particular YouTube ad was highly successful, performing for three years and garnering hundreds of millions of impressions on Instagram alone. Cliff emphasizes that ad success can be random; for instance, an Old Spice parody ad didn't perform well, but a rap song he wrote about his experience with dyslexia and ADHD unexpectedly generated another $3 million in revenue.

Weitzman also highlighted the effectiveness of repurposing content, citing a viral video comparing American and French founders that his team transformed into ten different pieces. He draws inspiration from creators like Mr. Beast, who utilize formats such as "ladders" (e.g., $1 hotel to $1 million hotel) or competitions across different age groups. These proven formats can be adapted to new content.

In performance marketing, he advocates for extensive testing across platforms like TikTok, Instagram, and YouTube, with different hooks. Since non-performing content in paid ads won't be served organically, there's no brand damage from testing hundreds of variations to find what resonates best and drives conversions.

> Are the voices any good? No, our voices are f*cking great.

## 26:03 - 30:04 Speechify's Multi-Channel Growth Strategy and a Strong Case for Investing in Meta

Speechify serves over 50 million users, maintaining a philosophy of offering a quality experience even to those who use the product for free, viewing it as a form of charity. Their growth is driven by a highly diversified strategy that taps into various user acquisition channels.

Word-of-mouth remains Speechify's most significant user acquisition channel, supplemented by strong organic growth through Apple features, including a design award and multiple features. Search Engine Optimization (SEO) also contributes substantially, and a notable 15% of users now come organically from ChatGPT.

For performance marketing, Meta is considered the

## 30:03 - 34:04 AI Commoditizes Design and Drives Engineering Token Spending

Cliff Weitzman argues that the manual, 'hammer and chisel' aspects of design, such as pixel-perfect work in tools like Figma or Photoshop, are being commoditized by AI. At Speechify, new product launches initially rely on default AI-generated designs, with human designers brought in later to refine the user experience and apply their taste. The value in design is shifting away from execution and toward subjective judgment and refinement.

Speechify has embraced an AI-first engineering culture, with leadership using cloud code terminals and widely adopting tools like Claude Code within their IDEs. Weitzman actively pushes his team to maximize their use of AI, sometimes needing to 'drag' them into fully leveraging these new technologies, even for tasks like creating marketing clips.

A significant shift in resource allocation is occurring: Weitzman projects that Speechify's engineering team will soon spend more on AI tokens than on employee salaries. He believes this financial trend, where token spending outpaces salary costs, will become a typical characteristic for AI-native companies, highlighting the increasing integration and cost of AI operations.

> Next year, our XPEC will spend more in tokens than we'll spend on salaries.

## 34:04 - 40:05 Prioritizing Adversity Quotient (AQ) and Identifying Red Flags in Hiring

Adversity Quotient (AQ) is more crucial than IQ or EQ, measuring a person's resilience and persistence when faced with extremely tough situations. It's about the willingness to stick with a challenge for extended periods, like an engineer grappling with a difficult problem for eight hours straight, rather than giving up after thirty minutes. The most impactful company advancements often occur after prolonged effort on hard problems.

To identify high AQ in candidates, look for tangible evidence of projects they've built and published. This demonstrates an ability to overcome minor obstacles, like registering an Apple developer account, to achieve a larger goal. Ideally, candidates should showcase something novel they created, proving their willingness to do the work rather than just possessing raw intellect. A key insight is that 'a little bit of slope makes up for a lot of y-intercept,' meaning growth rate is more important than starting point.

Be aware of red flags during the hiring process. These include any sign of dishonesty, such as manipulating statistics or changing definitions to make data look better. Another red flag is excessive talking that derails focus, indicating someone who might annoy others and pull the team off track. Lastly, be wary of candidates who are too comfortable or solely motivated by money.

A strong candidate will express genuine interest in the company or product, such as having a personal connection to its mission, like a child with dyslexia, or demonstrating thorough research into the company and its founders. Such answers signal loyalty and commitment, unlike those primarily driven by salary, who are likely to leave when a better offer comes along.

> A little bit of slope makes up for a lot of y-intercept.

## 40:04 - 44:05 Dyslexia and ADHD can be CEO strengths, and AI coaches are an emerging critical role

Dyslexia can foster resilience and a high tolerance for failure, which are essential qualities for entrepreneurs and CEOs. Individuals with dyslexia often face initial challenges, like struggling with reading, which can lead to a dark path for some, but for others, it cultivates a persistent mindset where trying repeatedly and failing is accepted as part of the process. This mirrors the nature of startups, where overcoming stacked odds and embracing what seems impossible requires an unwavering belief in continuous effort.

People with ADHD are not slow, lazy, or unintelligent; their minds simply operate differently. If an individual's brain processes information at 600 words per minute but they only read at 200 words per minute, distraction is inevitable. This fast-paced, multi-stimulated way of thinking can be highly beneficial for a CEO leading a large, established company (the 10 to 100 stage), where managing diverse functions is key. However, this same trait can pose difficulties during the early stages (zero to ten), which require intense focus on individual contributor work.

The role of an "AI Coach" is predicted to become incredibly commonplace within the next five years. This position involves guiding an organization on how to effectively integrate and utilize artificial intelligence tools. Companies are increasingly recognizing the necessity of dedicated personnel to ensure their teams are proficient in leveraging AI, similar to the work done by Austin Ray at Ramp, whose sole responsibility is teaching the organization to use AI tools. Prioritizing coaching internal teams on AI use is a vital investment for future success, despite initial perceptions that it might be a time sink for leaders.

> Being ADHD is actually very good to be a CEO.

## 44:05 - 48:05 Speechify's challenges with hiring from Google and extreme talent retention strategies

Cliff Weitzman explains Speechify's reluctance to hire from large organizations like Google, citing a perceived lack of "hunger" and loyalty among candidates from such environments. He notes that while exceptions exist, these hires often require coaching out of habits that can negatively impact company culture. This issue is particularly prevalent in the Bay Area, where employee loyalty is generally lower than in regions like Europe.

Speechify takes extraordinary measures to retain its talent, believing that employees typically leave not for financial reasons, but due to dissatisfaction in other areas of their lives. The company focuses on identifying and solving these underlying personal issues to ensure long-term commitment, often very quickly.

Weitzman recounts two extreme examples of retention. In one case, he helped an early teammate, who was considering leaving due to a lack of friends, by organizing social events to help them build a local network, which resulted in the teammate staying. In another instance, to prevent a key engineer based in Ukraine from leaving right before the war, Weitzman personally flew to Ukraine, ignoring travel restrictions, to work alongside the engineer and reinforce their commitment to the company.

> people typically leave not for money, they leave for another thing in their life that's not satisfying them.

## 48:05 - 52:06 Embracing Speed as a Strategic Advantage Through Rapid Communication and the Titanic Rule

Speed is identified as the paramount strategic advantage for any company, particularly in today's environment. For a remote-first organization like Speechify, with 200 employees across 36 countries, efficient communication is critical to prevent bottlenecks and maintain momentum.

The company emphasizes a goal of 60-second response times for engineers. If an individual is blocking another, it is their responsibility to inform the blocker. While learning to unblock oneself is encouraged, a lack of response leaves the blocked person unable to decide whether to wait or invest time in self-unblocking.

A core communication principle, dubbed the 'Titanic Rule,' mandates a response to messages within three hours. This response can be as simple as 'busy now, will respond by 4 PM' or 'unavailable for three days,' providing clarity and a timeline. If no response is received, a direct phone call is encouraged to get the necessary information.

This culture of rapid response is linked to the flexibility of remote work; the privilege of working from anywhere necessitates a commitment to speed. It cultivates a sense of urgency, ensuring the company operates efficiently with the mindset that tasks should have been completed yesterday to help the 'living organism' win.

> Speed is the number one strategic advantage for any company, especially today.

## 52:05 - 54:06 Increase Company Speed Through Clear Deadlines and Proactive Communication

To accelerate company speed, focus on concrete actions rather than vague goals. Implement daily deadlines with specific times, such as "3 PM on Tuesday," and include all relevant team members on calendar invites, explicitly labeling them "This is not a meeting, it's a deadline." This creates immediate accountability.

Should a deadline become unachievable, individuals are responsible for promptly updating everyone involved with an explanation. Furthermore, demonstrate progress rapidly by sending screen recordings of completed work for feedback. Leading by example is crucial: respond quickly to messages and make calls when urgency is required. Your visible hunger and fast pace will encourage others to adapt.

A past regret involved hiring two experienced individuals from Apple and Snapchat. While highly qualified, they disrupted the company's culture by adhering strictly to a 5:30 PM departure, indicating they didn't fully immerse themselves in the company's mission. Both quit after three months, which was fortunate, but the mistake was not acting sooner within the first thirty days, likely due to a prior desperation for hires.

> They would leave at five thirty PM. They like didn't live and breathe the company.

## 54:06 - 1:00:06 Rethinking Performance Reviews and the Imperative of QA

Traditional performance reviews are considered a total waste of time, as they address issues that should have been communicated much earlier by leadership. The core philosophy advocated is 'outcome ownership,' emphasizing that no credit is given unless work is actively shipped and in production. The analogy of a farmer with the best milk that spoils because it wasn't delivered to the doorstep illustrates that effort without a delivered outcome is useless.

To embody outcome ownership, employees must relentlessly drive their features to production. This means proactively unblocking pull requests by directly calling reviewers and immediately engaging QA teams to resolve issues. The goal is to accelerate timelines and ensure features are live quickly, transforming engineers into true outcome owners focused on tangible, deployed results.

Quality Assurance (QA) is highlighted as the most valuable skill, especially in a world where software engineering and design are becoming commoditized by AI. AI-generated code struggles to perfectly QA itself across different devices and edge cases. Many engineers and designers often fail to adequately test from a user's perspective. The ability to swiftly identify and fix bugs post-production, creating 'indelible' software, is what differentiates a good product from a great one. Engineers who might not be strong outcome owners could still excel in QA, a skill worth nurturing.

> If you made something amazing, but it's not in production, it's a waste of time.

## 1:00:06 - 1:04:06 Cultivating a Doer-Driven Culture Through Leadership, Not Management

The company operates without traditional managers, instead having 'leaders' who are encouraged to be hands-on with coding, designing, and user interaction. This structure rejects the idea of a 'fat general' overseeing from afar and significantly reduces time spent on formal performance reviews, cutting them from eight hours weekly per leader to just fifteen minutes every other week.

Traditional HR functions are minimized, particularly those dealing with internal politics; talent acquisition is treated as a separate, crucial function. The philosophy is that great people don't need constant management or vacation monitoring. Instead, employees are expected to manage upwards, clearly communicating what they need to be unblocked by leadership.

The core of the culture revolves around 'doers' who are deeply interested in the product and constantly working towards its improvement, sending demos and making progress. This energy and continuous thinking about the product, exemplified by a team member launching a feature after a shower thought, is what the company hires for.

The speaker believes everyone can be a 'doer' and emphasizes 'awakening' people's potential. Leaders are tasked with believing in individuals more than they believe in themselves, providing more responsibility and opportunities to help them rise to the occasion, echoing Steve Jobs' sentiment that the world is built by people no smarter than us.

> I'm not gonna be your manager. You need to manage me.

## 1:04:06 - 1:10:08 Speechify's Secretive Funding and Logan Paul's 'Virus Strategy' for Growth

Cliff Weitzman explains Speechify's choice to keep its funding and valuation private. With 1.1 million five-star reviews and 94% market share in B2C voice agents, the company does not need to publicize for user acquisition or hiring, having received 178,000 engineering applications last year. They also intentionally avoid news coverage to deter competitors and ensure users are attracted purely by the product's merit.

The conversation then shifts to Logan Paul's career strategy, which Cliff describes as a 'virus strategy.' This involves a creator or company attaching themselves to a larger, established organization to leverage its existing infrastructure and audience for exponential growth. Logan transitioned from being a top creator on platforms like Vine and YouTube into boxing, and then joined the WWE, a major entertainment entity that significantly broadened his reach.

This 'virus strategy' can be applied by joining an institution with a strong name but potentially weaker leadership, then becoming the new leadership. Alternatively, one could participate in established platforms like the 'All In' podcast or shows like 'Shark Tank' to leverage their massive existing viewership. Another method is partnering with a major publisher like Penguin Random House for a book, using their distribution network to reach a wider audience.

> If you find lightning in a bottle, you don't wanna tell other people about it.

## 1:10:08 - 1:14:09 Mr. Beast's Data-Driven Approach to Universal Content Creation

Cliff Weitzman spent three weeks living with Mr. Beast, observing his content creation process. He noted that Mr. Beast's strategy is intensely data-driven, a skill he has honed since age 13. This approach is rooted in an deep understanding of human psychology, which informs how he optimizes video formats and constantly improves his content.

A key element of Mr. Beast's success is his focus on creating universally understandable content that transcends language barriers. By simplifying concepts to their core, like a bucket, hole, and duct tape in an ad example, he ensures his videos resonate across every continent. Cliff applies this principle to his own business, using 'books' instead of 'PDFs' in ads because books are a simpler concept.

Beyond content, Mr. Beast demonstrates an extraordinary level of ambition as an operator, business person, and creator. He consistently takes on large-scale projects like Beast Games and Feastables, and has expanded into streaming to leverage clipped content on platforms like TikTok and Instagram Reels. This strategic diversification and relentless pursuit of new ventures show his forward-thinking approach to growth.

Despite some fluctuations in main YouTube views, Mr. Beast continues to make big moves, including creating the largest unscripted show in history by participant numbers. He is in a period of strategic recalibration, similar to 'bulking and cutting,' to identify and grow the next phase of his extensive business empire.

> he has spent between the age of thirteen and today just studying what converts, and he's just like, his dialed in on human psychology is insane.

## 1:14:09 - 1:18:09 High-Leverage Investing and AI's Energy Future

Cliff Weitzman shares how his brother achieved a 36-times return on Nvidia using a 3x levered option, while Cliff himself bought stock directly. Their high conviction stemmed from understanding the technology behind AI and believing Nvidia was mispriced. Cliff similarly invested a third of his money in Tesla in 2015, leveraging his background in renewable energy engineering.

He advocates for an investment philosophy of 'diversify to concentrate,' meaning investors should explore broadly to identify high-conviction opportunities, and then commit significant resources to those specific plays. This strategy allows for responsible leveraging when a truly promising opportunity is identified.

The discussion also addresses AI's rapidly increasing energy demands, which currently account for 1% of global energy utilization and are projected to increase 10 to 12 times in the coming years. Cliff identifies various energy investment opportunities, noting that while hydropower is efficient, new dammable sites are scarce, and fusion nuclear technology is promising but still in the future.

Cliff emphasizes solar power as a critical immediate solution. Solar technology is nearing its theoretical efficiency limits (47-49%), and while China has significantly reduced manufacturing costs, installation remains the primary expense in the US. He points out that vast amounts of public land could be used for solar farms, and the United States needs to greatly increase its solar installations to meet future energy needs.

> The biggest lesson I have in ten years of investing is you have to diversify to concentrate.

## 1:18:09 - 1:22:09 Achieving Product-Market Fit Through Relentless Iteration and Direct User Feedback

The path to product-market fit often requires continuous iteration and a willingness to refine the user experience. Early in Speechify's development, the founder used 'duct tape testing' on Stanford's campus, physically covering confusing buttons and making daily fixes to improve the product based on immediate user observations. This hands-on approach allowed for rapid problem identification and resolution.

A crucial strategy involved establishing hyper-personal feedback channels. The app featured a 'Help/Message Us' button that directly linked to the founder's personal iMessage, enabling real-time support and problem-solving. This direct line of communication ensured that user issues were addressed immediately and personally.

Another key method was proactively contacting churned users. The founder would export a daily list of users who had churned after 30 days and call them to understand their reasons for leaving. This process, though emotionally challenging—likened to 'getting kicked in the teeth'—provided invaluable insights into the product's shortcomings and enabled targeted improvements.

Ultimately, achieving product-market fit is about persistence. If users aren't adopting the product, it often signifies an implementation issue rather than a flawed core idea. Founders must stay in the game, iterating for potentially years, until the product 'fits like a glove' and truly meets user needs.

> It's like getting kicked in the teeth every single time someone tells you your baby is ugly, but it allowed me to solve the problems.

## 1:22:09 - 1:26:10 Speechify's Hacker Hire and Strategy to Reduce AI Inference Costs

Speechify's CEO, Cliff Weitzman, recounted finding and hiring a hacker who was exploiting their AI models to use them for free. After confronting the individual, a young student, Weitzman offered him a role. The hacker then lived with Weitzman for a year, working as a personal engineer on early voice projects and optimizing various personal apps, demonstrating a philosophy of collaborating with talented individuals even after conflict.

Weitzman highlights that high AI inference costs are "burning AI companies alive." In its early days, Speechify faced a similar challenge where user growth quickly made the business unsustainable, turning it into a "charity" due to escalating operational expenses.

To counter this, Speechify dedicated a 40-person AI engineering team and significant resources, including GPUs and a data center, for two and a half years solely to reduce inference costs. This focused effort brought the cost down to single-digit dollars per million characters, a substantial reduction compared to the industry standard of 30 to 100 dollars.

The company's experience allowed for long-term profitability, serving as a cautionary tale for new AI ventures. Weitzman uses the example of Cursor, a popular AI coding tool, to illustrate how a company can generate significant revenue for its product while effectively passing most of that money directly to underlying AI model providers like Anthropic due to unoptimized inference expenses.

> The main thing to look for is inference costs burning those companies alive.

## 1:26:10 - 1:30:11 Apple's On-Device AI Limitations and Strategic Business Growth

Apple's strong philosophy of on-device security creates significant challenges for its AI capabilities. This approach largely prevents the use of cloud-based AI models and restricts app access to full RAM, making it difficult to implement advanced features like text-to-speech or speech-to-text functionalities that are common in other AI-first products.

The speaker distinguishes between companies that focus on reducing costs and those that prioritize increasing revenue. While cost-cutting is a common business strategy, truly exceptional leadership is required to generate new revenue streams. Companies with strong growth levers, differentiated designs, or robust communities are generally better long-term investments.

Customer churn rates are highly dependent on a product's utility. Products that serve essential needs, such as a wheelchair or an accessibility tool like Speechify, experience very low churn because users depend on them. Conversely, frivolous applications or non-essential tools face much higher churn rates as they are not critical for users.

The speaker expresses optimism about a hypothetical future Apple CEO who is more engineering-focused, anticipating better support for developers to build innovative tools. This contrasts with Tim Cook's widely recognized genius in supply chain management, which has ensured product availability and contributed to successes like AirPods, generating massive revenue.

> Any Harvard MBA can figure out how to cut costs, just requires being smart. It takes a genius to increase revenue.

## 1:30:11 - 1:32:11 Cliff Weitzman makes a contrarian bet on OpenAI over Anthropic and advises never to bet against Elon Musk and Grok.

Cliff Weitzman acknowledges that Anthropic is currently running circles around OpenAI, especially in coding, and he personally uses Anthropic for almost everything while avoiding ChatGPT due to its tendency for 'click bait'.

Despite Anthropic's technical lead, Weitzman places a contrarian bet on OpenAI, citing Sam Altman's exceptional skill in securing enterprise deals, which he describes as 'black magic'. He also notes OpenAI's strong brand recognition outside the tech community, backed by significant early investment from Microsoft since its 2015 founding.

Regarding Grok, Weitzman believes it will be a meaningful competitor in the AI space. He emphasizes a personal rule when considering Elon Musk's ventures: 'never bet against Elon,' suggesting Musk's drive makes his projects consistently impactful.

> never bet against Elon.

## 1:32:11 - 1:38:11 Gratitude, Addressing AI-Driven Wealth Inequality, and the Importance of Lifelong Learning

Cliff Weitzman shares a profound lesson on gratitude learned during a challenging Mr. Beast project, where he spent eleven days sleeping in a Walmart parking lot. Initially, he felt resentment over production issues, but realized he had forgotten to be grateful for the unique opportunity and its potential to inspire millions. He emphasizes that finding goodness and maintaining optimism, even in tough situations, is a core input for resilience.

His experience highlighted the immense difficulty of starting from zero, especially without basic technology like a phone, which is crucial for navigation and employment. Weitzman expresses deep concern over escalating wealth inequality driven by AI, where value accrues to a select few. He advocates for universal tech access, suggesting governments should provide free Android phones as a basic human right to ensure everyone can participate in society.

To counter this imbalance and empower individuals, Weitzman stresses the importance of continuous self-education. While he values his university experience for personal growth and connections, he asserts that individuals are the "quarterback of their own education." He exemplifies this by listening to a hundred audiobooks annually, now leveraging AI to generate and consume detailed research outputs across diverse subjects, encouraging everyone to adopt a similar lifelong learning mindset.

> The most important thing is even when things are shitty, you gotta find something to be grateful for.

## 1:38:11 - 1:42:12 Personal Habits for Peak Performance and AI-Driven Human Evolution

Cliff Weitzman maintains a strict dietary regimen focused on protein. He consumes eighty grams of pea-based protein via shakes every morning and evening, ensuring consistent intake. His overall nutritional strategy aims for one gram of protein per pound of body weight daily, alongside fifty to eighty grams of fat, with carbohydrate intake adjusted (130g for cutting, 260-270g for bulking) based on his body's needs, treating it like a science equation. DEXA scans are also mentioned as a useful tool.

His workout routine centers on tricking, parkour, and gymnastics. When traveling, he seeks out local gyms, like Over Gravity in East London, and trains with the best athletes available. He also advocates for hiring coaches or experts to observe and provide tips, whether for physical sports or learning an instrument like guitar, to accelerate improvement. He also suggests using AI to generate personalized workout routines based on specific improvement goals.

Looking ahead, Cliff envisions humanity undergoing a new form of evolution, driven not by physiology but by the brain's ability to process and understand information faster. He likens modern technology, including his computer, phone, and even his team at Speechify, to "familiars" from fantasy, enabling individuals to achieve more by augmenting their capabilities. This means easier access to learning and building, like fixing an AC or even setting up a fusion reactor.

This evolving landscape is expected to instill a higher level of agency in people, empowering them to accomplish tasks more readily than ever before. However, he also cautions about the potential downside where individuals might abdicate decision-making to AI, emphasizing the importance of staying in the "driver's seat."

> I feel the same way about my computer and my phone and all the tools that I get to use... You can just achieve more.

---

Get podcast briefs for shows you follow: https://podbrew.app/