Sora Disappoints, ChatGPT Pro Tested, Inference Time Reasoning & More with Sunny Madra | E2062

Jason and Sunny kick off the show (0s)

• A prediction is made that Open AI will lose their lawsuit and face an injunction, with a potential settlement of billions of dollars, possibly the largest copyright infringement case in history (10s).
• The lawsuit may involve the New York Times and other parties, with content creators potentially joining in if they feel their work has been affected (15s).
• A settlement of over a billion dollars is anticipated, with the possibility of it being a three-comma settlement (23s).
• The podcast "This Week in Startups" is sponsored by various companies, including Lemon, which offers vetted remote developers with a 15% discount for the first four weeks (32s).
• Twist listeners can get an extra 20% off any plan for their first six months at OpenPhone.com (45s).
• Qualifying startups can join Zendesk's startup program and get their products free for six months by visiting zendesk.com (58s).
• The host welcomes back Sunny Madra, who has been absent due to the growth of his company, Gro, and expresses excitement about having him back on the show (1m8s).

Discussing Grok's recent developments and Middle East tech inspiration (1m26s)

• Grok, a company that produces inference chips, has recently raised a significant amount of money, and its parent company, Definitive Intelligence, was acquired by Grok 8-10 years ago (1m26s).
• Post-financing, the company has been busy doing deals and has spent a lot of time in the Middle East, meeting with friends and partners in the region (1m58s).
• The Middle East, particularly countries such as Saudi Arabia, the UAE, Kuwait, Bahrain, Oman, and Israel, is inspiring due to its young population, which brings a lot of energy to the region (2m26s).
• The region's core business, the oil industry, provides a strong foundation for the countries to elevate themselves into new industries, including AI, which they are actively investing in (2m44s).
• The Middle East's strategic location, with 4 billion people within a 1,000-kilometer radius, makes it an attractive hub for business and innovation, with major airports in Dubai and Saudi Arabia serving as key connections to India, Africa, Europe, and other regions (3m14s).
• The region has a positive energy, with a focus on business and innovation, and is less affected by regulatory and cultural issues that are prevalent in other parts of the world (3m37s).

Global business challenges and Jason's 2025 announcements (3m51s)

• The United States has a rival in China, but due to various reasons, including the Chinese government's actions, it's no longer possible to participate in the "poker game" of business with China (4m5s).
• The Chinese government's authoritarian behavior, such as taking control of education startups and rendering investments worthless, has led to a lack of trust in investing in the region (4m38s).
• Europe is in decline, with rare exceptions of successful companies emerging from the Nordics, Berlin, and London, making it challenging to do business there (4m59s).
• A new region has emerged, which is charming and has a group of people who admire the Western democratic approach and have sent their children to the US and Europe for education, resulting in a pool of talented individuals with perfect English and impressive degrees (5m27s).
• This region, which includes countries such as Saudi Arabia, has people who want to do business with the US and are aligned against authoritarian countries like Russia and China (6m3s).
• The region is exciting for business, with many entrepreneurs and investors, including the speaker, participating and building new projects (6m11s).
• The speaker plans to announce two new projects in two different regions in 2025, which will involve spending more time in the region and exposing their family to the local culture (7m13s).
• The region, particularly cities like Dubai and Abu Dhabi, is becoming a destination for smart people from around the world, offering new opportunities for business and growth (7m43s).

Gemini app demo and ChatGPT 4 comparison (7m53s)

• Gemini is an impressive app that has become part of the workflow, rivaling ChatGPT 4 in user experience, with access to data, images, flights, and other information from the Google Suite of services (7m55s).
• The Gemini app has undergone updates, including Gemini 1.5 and Deep research, and a notebook research feature has been released (8m1s).
• A demo of the Gemini app and a comparison with ChatGPT 4 is planned, including a test of ChatGPT Pro, a $200 a month product (8m39s).
• However, there is an issue with upgrading individual accounts in an Enterprise plan to ChatGPT Pro, with no clear upgrade button available (8m48s).
• A comparison test is proposed, where a standard prompt is entered into both ChatGPT 4 and ChatGPT Pro at the same time to compare the results (9m19s).
• The test aims to compare the performance of the regular ChatGPT 4 with the more advanced ChatGPT Pro, which is meant to be at a PhD level (9m50s).

Lemon.io - Get 15% off your first 4 weeks of developer time (9m57s)

• Lemon.io offers a solution for individuals and businesses that need help with software development but struggle to find and afford great talent, providing access to thousands of on-demand developers who have been vetted for their results-oriented approach and experience (10m17s).
• These developers have competitive rates, and Lemon.io handles the process of finding and integrating them into a team, making it easier to work with great developers (10m40s).
• To ensure quality, Lemon.io only offers handpicked developers with a minimum of 3 years of experience, who are in the top 1% of applicants (10m45s).
• If any issues arise, Lemon.io will find a replacement developer as soon as possible, providing a reliable service for its clients (10m53s).
• Many Launch Founders have had positive experiences working with Lemon.io, and the company offers a call to action to find a perfect developer or tech team in 48 hours or less (11m2s).
• Twist listeners can get 15% off their first four weeks of developer time by visiting Lemon.io/twist, allowing them to hire developers smarter and faster (11m15s).

Evaluating ChatGPT 4.0 Pro limitations and robo-taxi fleet costs (11m24s)

• ChatGPT 4.0 Pro is being tested with a prompt to build a detailed model estimating the cost of creating a robotaxi fleet for all car trips in the USA (11m28s).
• The prompt includes requirements to account for total car trips in the USA, Uber and Lyft trips, fleet efficiency, fleet operations, fleet size, public transit, fleet costs, induced demand, and provide a comprehensive model with data-backed assumptions and linked sources (12m22s).
• The model's output should estimate the total cost of building a robotaxi fleet in the entire US (12m41s).
• ChatGPT 4.0 Pro is seen gathering data, working through a hypothesis, and using logic to assess the problem, including referencing the 2017 National Household Travel Survey (NHTS) for trip estimates (11m53s).
• The AI is calculating robotaxi fleets and other relevant factors to provide a comprehensive model (12m11s).

Differences between ChatGPT models and future improvements (12m46s)

• A comparison was made between two AI models, one costing $200 a month and the other costing $20 a month, to assess their performance in scaling a robot taxi fleet to cover all US trips plus 20% of public transportation trips with a 20% induced demand factor (12m47s).
• The $200 model took 22 seconds to complete the task, while the $20 model completed it in a few seconds, highlighting a significant difference in performance (13m1s).
• The task involved calculating the number of vehicles required to replace the entire US fleet, with assumptions including 20-30 trips per vehicle per day, 5 days off for maintenance, and 360 days on the road per car (13m48s).
• The $200 model estimated that 46.9 billion vehicles would be needed to meet the demand, which would cost $5 trillion at $100,000 per vehicle or $1.5 trillion at $330,000 per vehicle (14m51s).
• The $20 model, on the other hand, took 2 minutes and 42 seconds to complete the task and provided a more organized and detailed response, including references to data sources such as the American Public Transportation Association and SEC filings (15m22s).
• The $20 model's response included a step-by-step illustrative model with clear stated assumptions and calculations, and estimated 365 billion annual car trips, which is similar to the $200 model's estimate (15m48s).
• The $20 model also pulled data from S1 filings from 2018 and 2019 for Uber and Lyft trips, and provided context for the calculations, including a reference to the NITSA data (16m14s).
• A calculation was done to determine the current Riot share, which represents 1 to 1.5% of demands, indicating a more efficient fleet operation (16m34s).
• An assumption was made that each fully autonomous Robo taxi can do 20 to 30 trips a day, with an average trip duration of 15 to 30 minutes and some downtime between rides (16m45s).
• Fleet operation constraints were considered, including 6 hours a day for charging, cleaning, and maintenance, and 5 days a year for major maintenance shutdowns, resulting in 360 days of utilization per year (17m5s).
• The daily utilization breakdown was calculated to be 25 trips per day, 360 days, resulting in 9,000 trips per year (17m19s).
• The required fleet size to handle 100% of US car trips was calculated to be 37.8 million Robo taxis, considering 340 billion trips and 9,000 trips per year (17m41s).
• The calculation took into account induced traffic and public transportation, capturing 20% with FSD cars, resulting in a fleet size of 38.2 million rides (18m7s).
• The results of the calculation were compared to a previous calculation, showing that the results were not significantly different, with a difference of around 10% (18m37s).
• It was suggested that the core language models (LLMs) may not be getting much better due to limited data, and that the value of LLMs is now determined by the interface, instructions, and personalization (19m1s).
• The idea was proposed that the language models themselves may be starting to plateau due to a lack of new data, and that the focus is now on the "fit, finish, and polish" of the models (19m35s).
• Ilia's statement was mentioned, suggesting that the data used to train language models has been largely exhausted, and that further improvements will come from other areas (19m42s).

OpenPhone - Get 20% off your first six months (19m57s)

• Missed calls can be a missed opportunity for customers, as they may call someone else if their call is not picked up immediately (19m57s).
• OpenPhone is a solution that ensures businesses never miss another customer call, and it is super affordable and easy to use (20m8s).
• OpenPhone offers a business phone line and complete control over the business for just $15 a month, a significant reduction from the tens or hundreds of thousands of dollars previously spent on corporate phone systems (20m18s).
• OpenPhone provides features such as syncing with HubSpot and AI-powered call summaries, as well as automated responses to ensure no calls are missed (20m30s).
• OpenPhone allows users to port their existing phone number over at no extra charge (20m39s).
• A 20% discount is available for the first six months of using OpenPhone, which can be redeemed by visiting openphone.com and using the code "Twist" (20m48s).

LLM focus shift and robo taxi cost model experimentation (20m58s)

• Large Language Models (LLMs) seem to be stuck, and the focus has shifted to the inference interface, which interprets instructions, personalizes users, and utilizes proprietary data sources such as Reddit, Twitter, and Google flights data, as well as deeper data and instruction sets in the interface (21m0s).
• An experiment was conducted to compare the performance of two models, one costing $200 and the other costing 10 times more, and the results showed that the output of the $200 model provided a little more background information (21m23s).
• The way the prompt was structured helped direct the less powerful model to achieve similar results to the more powerful model, suggesting that the key to success lies in the prompting technique rather than the model's power (21m44s).
• A new experiment was proposed, where the same prompt would be given to both models to build a detailed cost model for replacing all US rides with Robo taxis (21m54s).
• The results of the new experiment showed that the model came up with an annual operating cost of $150 billion and a total initial deployment cost of $2.8 trillion, with explanations and citations, but with different numbers than expected (22m47s).
• The experiment demonstrated the importance of providing a framework and instructions to the model, as the original prompt included specific details such as the number of rides, public transit rides, and the percentage of induced demand (22m31s).

Importance of prompt engineering in AI tools adoption (23m1s)

• Creating better prompts can significantly improve the performance of AI models, potentially providing 10 times more value than a more expensive model, by investing time in prompt engineering (23m7s).
• A well-crafted prompt can help AI models provide accurate information, such as the daily passenger trips of a Robo taxi, which was correctly answered as 20, citing a paper by Litman in 2019 (23m35s).
• The accuracy of AI models can be influenced by the assumptions and parameters provided in the prompt, such as the number of days off per year and hours off per day for maintenance and charging (23m48s).
• AI models can provide detailed estimates, including the cost of charging infrastructure, maintenance, and operations software, citing sources such as McKenzie and Company and the Rand Corporation (24m34s).
• The cost of AI tools, such as ChatGPT Pro, can be justified by the potential increase in employee efficiency, with an incremental cost of $2,000 per year potentially leading to a 3% increase in productivity (25m1s).
• The adoption of AI tools in businesses depends on employees' willingness to use them, which can be a challenge, as people may not be inclined to adopt new technologies (25m28s).

Workplace AI adoption challenges and generational tech differences (25m31s)

• The world is expected to divide into two groups: those who use AI tools as their default and those who don't, with this division being a result of habit (25m35s).
• A similar division occurred in the mid-1990s when the internet was introduced to the corporate world, with some people eager to use it and others who didn't trust it (25m49s).
• Younger people, particularly those under 20, are more likely to use AI tools, such as Open AI, due to their tech-savviness and energy (26m20s).
• To encourage the adoption of AI tools in the workplace, it's suggested to hire interns or entry-level employees, who are likely to be younger and more open to using these tools (26m14s).
• There are generational differences in the use of AI tools, with younger people being more likely to use them, and older people requiring more encouragement to adopt these tools (27m15s).
• Hiring for entry-level jobs or internships can help bring in younger employees who are already familiar with AI tools, as these positions often attract younger candidates due to the salary and opportunity for college credit (27m3s).

New productivity hacks and Gemini app's growth (27m22s)

• A new productivity hack called "New Tab Override" by Soran Hensell allows users to set a default URL to open in new tabs, and also changes the focus from the address bar to the webpage, making it easier to start typing without having to click or hit tab (27m27s).
• This hack is useful for quickly opening new tabs and searching for information, and can be especially helpful for people who frequently open new tabs during meetings or while working (28m1s).
• The discussion also touches on the Gemini app's growth and its ability to perform inference time reasoning, which allows it to do additional reasoning at the time of query, rather than just relying on pre-built models (28m56s).
• Inference time reasoning is a feature that enables the model to iterate and refine its responses in real-time, rather than just predicting tokens based on a pre-built model (29m6s).
• This feature is different from the original iterations of ChatGPT, which relied on standard prediction of tokens, and can produce more accurate results, especially when used with sophisticated prompts (30m2s).
• The discussion also highlights the importance of prompt engineering in getting the most out of models like Gemini, and how it can reduce the need for inference time reasoning (30m12s).
• The Gemini model was able to come up with the same results as a more expensive model, using inference time reasoning to refine its responses and produce more accurate results (30m21s).

Zendesk - Get six months free (30m27s)

• To deliver exceptional customer experiences, a platform like the Zenes Suite is necessary, providing tools to build stronger relationships without growing headcount, and is used by companies like Shopify, Squarespace, Uber, and Instacart (30m38s).
• Unity, a famous company, saved $1.3 million with Zenes' automations and self-service, and saw an 83% increase in their first response time (31m5s).
• The Zenes for Startups program offers unlimited access to all Zenes products, expert insights, best practices, and entry into their community of Founders, all at no cost for the first six months (31m31s).
• Zenes provides metrics for easy reporting, keeping businesses agile and investor-ready (31m22s).
• The Zenes Suite is easy to set up and scales with growing businesses (31m16s).
• The Gemini API is gaining steam, with huge growth, although the exact percentage is unclear, with one report suggesting it may be over 50% (33m21s).
• Gemini Advanced 1.5 Pro with deep research is a better product for certain tasks, providing more detailed analysis and showing its work (32m45s).
• The Gemini app does not currently have deep research, but has other features and is comparable to ChatGPT's app (33m2s).
• Developers can use the Gemini API, and Google offers $150,000 in credits through the startupcreditrepairbusiness program (33m55s).

Startup support programs and Google Gemini's research capabilities (34m6s)

• Digital Ocean provides credits to startups, while AWS has a "rack rate thing" but is not very supportive of startups outside of Y Combinator, with whom they have a close relationship (34m27s).
• Y Combinator is described as being "sharp-elbowed" and prioritizing their own interests over others, but the speaker is confident in their own abilities and expects to match Y Combinator's application numbers (34m51s).
• Google Gemini's research capabilities were demonstrated by generating a report on total rides in the US annually, including bus trips, and estimating the average cost per ride and total cost of manufacturing and deploying robot taxis (35m0s).
• The report also included a table with estimated costs of manufacturing, deploying, and maintaining robot taxis, as well as projected average distance traveled and total cost per robot taxi ride (35m20s).
• Google Gemini's capabilities also include creating a document in Google Docs based on the report, which can be saved and referenced in the future, potentially even incorporating data from emails and conversations (36m10s).
• The speaker believes that Google is a "sleeping giant" in the AI space, and that their capabilities will become increasingly interesting and powerful in the future (36m56s).
• Other companies, such as Grok, are also doing a good job in the AI space, and will be discussed further (37m0s).

AI model performance evaluation and simplification (37m8s)

• The grades given to the AI models are: B+ to 01 Pro, A to 01 with deep research, and A+ on a value basis to Gemini 1.5 with deep research, considering the $200 a month pricing does not matter in corporate America as it's a valuable tool for employees and team members (37m20s).
• The value of the AI models is more important than the pricing, and in a corporate setting, the cost is negligible compared to other expenses, such as employee parking (38m17s).
• The reasoning process provided by Open AI and Gemini's deep research is a powerful feature for users in a work context, as it shares the thought process behind the output (38m49s).
• The naming and versioning of the AI models, such as Gemini 1.5 Pro with deep research, can be confusing for consumers and should be simplified to just "Gemini" with options for deep research or other features (39m12s).
• The current menu options for Gemini, including 1.5 Pro, 1.5 flash, and 2.0 experimental, are too confusing and should be streamlined to provide a better user experience (39m36s).
• If only one AI model could be used, the choice would be between Open AI Pro and Gemini deep research, with the latter being preferred due to Google's access to data and the potential for the gap between the two models to grow (40m20s).
• The use of AI models in a work context can be beneficial, but it's also important to consider the time and resources required to manage and utilize the tools effectively, such as the 10 weeks required to manage high school interns (40m42s).
• The preference is to hire people straight from school, specifically from the University of Texas, due to their intelligence and blue-collar work ethic, with the top 1 or 2% being comparable to Ivy League students (41m2s).
• The University of Texas is a large, well-funded public school, with lower tuition fees for in-state residents, making it a more affordable option compared to private schools in the Bay Area or New York (41m57s).
• The University of Austin, on the other hand, is a startup school with a focus on free will, libertarian, and Republican values, aiming to teach from first principles and avoid woke culture, with its first class consisting of around 50 students (42m40s).
• The idea of creating a Founder University is being considered, with plans to establish a physical space in Austin, Texas, and offer a unique approach to teaching entrepreneurship, potentially with a model where students receive $25,000 to start their company instead of paying tuition (43m44s).
• The Founder University would focus on teaching how to be a founder, with the goal of providing a comprehensive education in entrepreneurship, and the details of the program, such as the format and structure, are currently being figured out (44m3s).
• A company received a million-dollar valuation for 2.5%, which is considered a good deal, with the expectation that only one out of three companies will secure another round of funding, making it a high-risk investment (44m16s).
• The investment is comparable to the Y combinator accelerator, but with high risk, as two out of three companies may not make it to the next round of funding, effectively making the investment worth 75k at 3 million for 2.5% (44m36s).
• The job involves seeing a lot of good projects and taking high-risk bets, which is an enjoyable part of the job (44m48s).
• A lightning round is initiated to quickly discuss several topics, including some non-demo items, such as Meta's recent launch (44m59s).

Meta's Llama 3.370b launch and AI industry impact (45m2s)

• Llama 3.370b is a model that has shown significant improvement and is making good inroads against its competitors in benchmark tests, despite being a relatively small model (45m3s).
• Meta is open-sourcing Llama 3.370b, allowing others to use it without trying to make money from it, which is seen as a positive move for the industry (45m39s).
• The open-source model has some limitations, such as a cap on the number of users, but it is still a significant step forward (45m46s).
• The move is seen as a way for Meta to provide an alternative to closed AI systems, such as those developed by Sam Altman, and to establish itself as a backstop in the industry (46m0s).
• A comparison of Llama 3.370b with other models, such as Gemini Pro 1.5 and GPT 4, shows that it is becoming increasingly competitive in terms of pricing, with costs ranging from $0.10 to $10 per million tokens (46m15s).
• Meta's pricing strategy is seen as a key factor in its ability to compete with other models, with the company having a history of being able to drive down costs and make its products more affordable (46m34s).

Infrastructure costs, competitive landscape, and AI-generated content evolution (46m37s)

• The current pricing of AI services suggests that companies are not making a significant profit, with a large chunk of the margin going to Nvidia, which has an 80% margin on its products (46m55s).
• Nvidia's high margin could be a challenge for the industry, potentially leading to a compression of their margin and lower pricing to compete with other companies like Amazon and Apple (47m22s).
• Companies are providing inference chips and other AI services at increasingly competitive prices, which could lead to a decrease in Nvidia's margin (47m42s).
• Custom chips are not typically made for individual clients, but rather companies develop their own chips that can run a wide range of models (47m48s).
• The pricing and capabilities of AI services are becoming more interesting, with some companies offering more competitive pricing and others focusing on developing their own infrastructure (48m2s).
• Sora, an AI service, has been tried and reviewed, with the conclusion that it is not yet industry-leading, with its generated content looking like it was created using game engines and not realistic enough for prime-time use (48m14s).
• The training data used for Sora and other AI services is often generated using game engines, which can result in content that looks professional but not industry-leading (49m7s).
• The generated content from Sora and other AI services is not yet good enough for use in high-end productions, such as those by Disney, George Lucas, or JJ Abrams, but could be useful for storyboarding and other pre-production purposes (49m51s).
• The current state of AI-generated content is about 60-70% of what would be acceptable for prime-time use, and it may take a few years or more for the technology to improve to the point where it is industry-leading (50m1s).
• Marvel should release an AI model in partnership with a company, allowing Disney Plus subscribers to create and share Disney character models and short videos within the Disney app, which would be a killer feature (50m17s).
• A Chinese AI model has been demonstrated to create a two-minute video that looks like a professionally shot film, similar to a George Clooney movie, by stealing from Hollywood to achieve the effect (50m46s).
• The video created by the Chinese AI model looks distinctly closer to a Hollywood film than a stock photography library (51m19s).
• The Open AI whistleblower who allegedly committed suicide or was killed had a lot at stake, and it is predicted that Open AI will lose their lawsuit and have to settle for billions, potentially the largest copyright infringement case in history (51m30s).
• The settlement could be a three-comma settlement, and content creators who feel their work has been infringed upon may join the lawsuit (52m2s).
• The death of the 26-year-old Open AI whistleblower is considered a tragedy, and the circumstances surrounding it are unclear and suspicious (52m25s).
• The Sora AI model is considered disappointing, with a grade of B, and its use case is unclear (53m17s).
• The Chinese AI model has demonstrated a significant jump in reasoning and inference time, similar to Gmail's ability to guess the fifth word, but now able to guess the third or fourth word (53m42s).
• The development of AI technology, such as Gemini, is advancing rapidly, with significant improvements happening in just a few years, potentially leading to major breakthroughs in the near future (54m1s).
• In the future, AI could be used to generate creative content, such as a lost episode of a TV show like The Sopranos or a film directed by a character from the Star Wars series (54m10s).
• AI could also be used to create personalized content, such as a talent show featuring a person's family members or a film directed by a person's character from a favorite TV show (54m42s).
• ChatGPT can be used to generate responses in the style of famous historical figures or celebrities, such as Stalin, Einstein, or Bob Dylan (55m20s).
• A person can use ChatGPT to communicate in a way that is not typical of their personality, such as a person who is bad with women using ChatGPT to generate responses in a more confident or charming way (55m2s).
• The use of AI to generate creative content and responses raises interesting questions and possibilities, and is a topic of ongoing discussion and exploration (55m38s).

Trust and bias in language models, news analysis startup ideas (55m44s)

• Perplexity was used to understand the current situation in Syria, and it provided a good explanation of the events unfolding in real-time (55m44s).
• The same task was performed using GROK, which also has access to real-time information, giving it an advantage (56m4s).
• The use of real-time information sources like Perplexity and GROK can be beneficial, but there are concerns about anonymous accounts and their potential biases (56m24s).
• Anonymous accounts with a large following, such as those on X, can be problematic as they may be controlled by foreign actors or spammers, and their biases can influence language models (56m46s).
• The potential for foreign actors or spammers to create multiple paid accounts to influence people and language models in real-time is a concern (58m10s).
• This could be done by hiring a team of people to manage multiple accounts, each with a different pseudonym, and feeding language models biased information (59m2s).
• The use of real-time information from platforms like Reddit, which has a deal with OpenAI, can be exploited by foreign actors or spammers to influence language models (59m17s).
• Trusting data on Reddit has become challenging due to the existence of large language models (LLMs), which can potentially manipulate information, but LLMs can also be used to analyze a large amount of data from multiple accounts to get closer to the truth (59m25s).
• The advantage of the current era is the ability to have a large number of parallel agents looking at data and aggregating it to find the ground source of truth (59m42s).
• There are a limited number of influential accounts on platforms, likely in the low hundreds, making it possible to shape reality and the message (59m56s).
• Wikipedia has a small number of super editors, under 100, who are highly influential and some of whom get paid covertly by PR firms, highlighting the need for a system to identify bad actors and great content creators (1h0m11s).
• The existence of LLMs presents an opportunity for a startup to analyze data, identify bad actors, and promote great content creators (1h0m33s).
• The idea of creating a startup to do covert intelligence operations for corporations and individuals is proposed, with the potential to start a new CIA-like organization (1h0m52s).
• The importance of having deep discussions and insights on topics like AI and its applications is emphasized, with the show aiming to provide a platform for insiders to share their knowledge (1h1m21s).
• The show's host invites listeners to search the AI archive powered by podcast AI and to reach out to him for potential investment opportunities through founder.university (1h1m31s).