Built with AI: Day 2 with Brandon Doyle

In this episode of 1 to 100, Hubertus Hofkirchner joins Jack, Rishabh, and Jeremiah to discuss Bitcredit, the credit system he is building on Bitcoin.Hubertus explains why he sees Bitcoin as digital gold. He argues that commerce also needs credit to cover the time between producing goods and receiving payment. He walks through a farmer, miller, and baker exchanging promises to pay as goods move through the supply chain.Bitcredit uses electronic bills of exchange between businesses. Hubertus describes how companies can extend payment terms and use those bills to pay other suppliers. He also explains the role of e-cash in the system.Rishabh asks how businesses could use a currency whose purchasing power changes so much. Hubertus argues that wider use in trade could reduce Bitcoin's volatility over time. He expects early adoption in markets where businesses struggle to obtain credit or make international payments.Jeremiah asks what would persuade US consumers to give up credit card rewards. Hubertus says he is focused first on producers and supply chains. He uses a milk ranch in Latin America to explain how payment terms and purchasing costs affect a business's margins.Jack asks how Hubertus leads an open-source project. Hubertus describes recruiting developers, planning work on Mondays, holding Friday demos, and opening more of the development process to community contributions.Chapters00:00 Bitcoin and other cryptocurrencies04:35 Hubertus’s background05:40 Bitcoin, banks and base money08:24 Why commerce needs credit13:54 Volatility and trade16:14 Central banks and market-based credit20:56 Credit card rewards and early adoption24:19 The first supply-chain transactions28:48 Funding Bitcredit30:41 Leading an open-source team32:48 Skepticism about Bitcoin39:02 Block’s consumer marketing40:16 Hayek on private money

In this episode of Built with AI, Brandon joins Jack to discuss building products with AI, agency creative, and the distribution work that helps a product grow.They review Gridbank and Dan's UGC, two products that give brands access to real-person video clips. The clips can provide hooks and product creative without sending a product to a creator. They also discuss how a landing page should show the product clearly and make the use case easy to understand.Jack shows sites and Shopify themes built with AI. He can recreate a reference site, adapt it to a brand, and replace the products and media. Brandon shows a drawing game he built with Astra for his children. The examples lead to a discussion about how AI changes the cost and speed of product development.Brandon and his brother-in-law bought the recipe app Pichi for $9,000. The app was producing about $2,600 in monthly recurring revenue, with about $900 in monthly operating costs. They plan to add grocery lists and meal plans, then use AI tools to help improve and grow the product.The group discusses how recipe software can combine personalization with recipes from chefs, creators, and people whose food users already follow. They consider a product that turns food photos into recipes and lets users save the dishes they want to cook.The episode also covers distribution. AI can help a small team build websites, games, and software, while trusted audiences, useful partnerships, and a clear product position help those products reach customers.Chapters00:00 Landing pages, Gridbank, and Dan's UGC04:34 Building sites and Shopify themes with AI08:00 A drawing game built with Astra09:17 Buying the recipe app Pichi11:43 Personalization and chef-created recipes14:36 Recipes from creators and public food posts16:11 Designing a faster recipe discovery product24:27 AI, knowledge, and the loss of mystery29:33 AI agencies, brand sites, and distribution

In this episode of Built with AI, Brandon joins Jack to talk about organizing AI work and checking that it was done well.Jack is moving his agents out of a Slack-style workspace. The threads have no titles, so he has to reread them to see what each agent is doing. He wants tasks, projects and recurring systems on a dashboard that shows when his brands last sent email campaigns.Brandon will still want to see progress once he can start agents by voice. At his agency, new clients review every ad, and after a few months they skim a batch in seconds. He wants that trust before AI writes his emails. One of his AI engineers sent a New York restaurant group a bug-fix update that Brandon found too formal and too AI-written.Jack questions the need for specialized agents. He made about 20 sub-agents with different personalities, and one replied that a task was not its job. He wants agents that share skill files and finish the work. Brandon thinks the personalities may be a UI choice for now, one that helps people compartmentalize.ChatGPT built Brandon a day-by-day plan and race-week diet for his first HYROX competition from his training times.Jack shows Codex sending each new product image to Slack for his approval. On his jewelry brand, raising the pop-up opt-in rate from about 1% to 10% helped it past a plateau of $40K to $60K a month.Chapters00:00 Organizing agent work as tasks and projects05:04 Seeing the progress of AI work08:43 Client trust and AI-written emails15:58 Specialized agents or one general agent19:15 A HYROX plan22:39 Marketing as a skill that keeps changing27:54 AI emails and product images for Jack's brands35:44 Image ads, coolness and pop-up opt-ins

In this episode of 1 to 100, Liam joins Jack, Rishabh, and Jeremiah to discuss remote hiring, management, and building an organic search business with services and software.Liam explains why his company hires globally. He can recruit from a larger talent pool and serve different markets. Jack describes the cost difference between hiring engineers in the US and South America. Rishabh explains how an office near a strong university can help a company recruit locally.Liam describes a hire who understood marketing deeply and struggled with pace, reliability, and ownership. His team now looks for intelligence, agency, and curiosity. Rishabh argues that AI can handle some of the coordination work that previously required managers. Liam says managers still need enough knowledge of the work to judge its quality.They debate how experience, ego, and family affect career decisions. Rishabh argues that experienced operators may need to return to individual contributor work. Jeremiah explains how having children changed the work he was willing to pursue. Rishabh challenges his broader claim about family and career success, pointing to founders who have families.Liam describes an organic search agency with about 20 people, including five to seven engineers and data scientists. His team builds tools for research and delivery, then uses those tools to produce results for clients. He explains why they moved from low-priced software subscriptions to a service model.The team is considering how client data and software could help the company retain its value as competition grows. Jack raises personal brand as a way to attract customers. Liam wants to build a company that can retain its value independently of his own reputation.Liam recommends selling before building. His team sought commitments, delivered work manually, and developed software after customers paid. He explains how they changed direction when an offer failed. They also discuss outbound sales, marketing budgets, and qualified sales calls that Liam says came through AI search on a new domain.Chapters00:00 Remote hiring and work preferences03:48 Global talent and the cost of an office07:39 Hiring near a strong university10:31 Hiring for a brand and managing performance13:36 AI and layers of management16:54 Hiring for intelligence and agency19:08 Ego and returning to individual work24:33 Family and career decisions28:30 An organic search agency with software32:11 Competition and building a valuable company37:20 Sell before you build41:02 Outbound sales and marketing channels46:09 Qualified leads from AI search

In this episode of Built with AI, Jacob Posel and Cody Plofker join Jack to talk about AI-built landing pages and how they run their agents. Chase Mohseni joins briefly at the start.Cody offers brands a new landing page for $20K, paid only if it beats their current page in an agreed A/B test. If it loses, the brand keeps the page for free. Cody thinks a 5% conversion lift is a fair bar, and Jacob says that lift could be worth $50K to $100K to some brands.Jack shows a skill that copies a reference site or a whole store into a theme. Cody warns that the copied page may perform poorly. His own skill turns a reference page into a wireframe, asks about the product, persona, traffic source and angle, then builds in the brand's design system. He says brand context, such as heat map data and customer surveys, matters most.Jacob explains workers, a feature in the product he builds: specialists inside a general agent like Claude Code or Codex, each with its own skills and knowledge base. He says separate agents help people adopt AI and let teams give each agent different permissions.Cody has models review each other's work because agents grade their own work too easily. Jacob's Codex reviewer checks API values against warehouse data in a separate process with fresh context. He says agents improve when they can read logs, open the site and run tests.Jack keeps losing finished work on unmerged branches. Jacob suggests starting with a /ship command that merges and deploys, then adding checks when something breaks.Chapters00:00 Trying new models02:42 Cody's $20K landing page offer06:13 Copied pages, design systems and context12:12 Workers, skills and separate agents17:21 Models reviewing each other's work20:44 Cody's Hermes agent on a Mac mini23:13 Long agent runs and self-checks26:55 Bringing a small team onto AI29:21 Shipping finished work with /ship

In this episode of Built with AI, Brandon joins Jack to talk about businesses built with always-on agents and the limits of AI-generated creative.Brandon runs an OpenClaw agent on a Mac mini in his Provo, Utah office. In one experiment, the agent found landscapers and HVAC companies that ranked in Google but had no website. It built each one a site, mailed a postcard with a QR code to the preview, and called to follow up. Brandon says it now brings in about $9,500 in monthly recurring revenue.For DAVID AI, Brandon wants ads that look cool and timeless, which he says is rare in B2B. When Claude Code made images for him through Higgsfield, he called the results "AI of two years ago." Jack's designers get better results in Higgsfield itself than he gets by connecting it to Claude or Codex.Within a year, Brandon expects ChatGPT, Claude or Gemini to become the main way consumers browse and shop. Jack argues that selling base intelligence is hard to defend because competitors catch up within months. He sees more value in services that put it to work for businesses.Brandon is also starting a landscape design and installation company with AI. Claude researched the location and industry, then made the name, logo and website. The site serves Utah County, and Brandon plans to outsource installation to crews.Chapters00:00 Desk setups and Mac mini agents04:18 A website experiment that reached $9,500 MRR05:15 Implementation work and custom Shopify builds08:10 Creator outreach and YouTube's partnerships API12:44 Cool B2B ads for DAVID AI15:44 Claude Code's browser and switching models18:37 Browsing and shopping through AI assistants20:27 The economics of selling intelligence23:38 Higgsfield images and taste26:03 Starting a landscaping company with AI

In this episode of Built with AI, we discuss building with AI when the hard part is defining success, checking the output, and improving the system.The discussion covers the shift from code as a technical barrier to prompting through a command line. We look at building data syncs, apps, documents, and workflows with AI while keeping a clear finish line.We also discuss model choice and usage pricing. Faster modes can buy time at a premium, but the newest model is not always the best model for the work.A major theme is self-healing and self-improving automation. We discuss outreach systems, email subject lines, send times, creative blocks, rate limits, and how each step should contribute to the final outcome.The conversation covers hill climbing as a way to optimize. Early signals may be opens, then clicks, add-to-cart events, and conversions. Optimizing one metric in isolation can create misleading results, damage the brand, or produce low-value customers.We discuss connecting data across commerce, marketing, subscription, customer, and finance systems. Programmatic bookkeeping can remove repetitive work, but payment and account-receivable workflows still need reliable checks.The discussion also covers the limits of AI-generated creative work. AI can produce several technically good options, but human taste and domain judgment still matter when choosing website copy, headlines, and user experiences.We close with team adoption. The approach discussed is to lead by example, look for agency and curiosity, automate workflows through natural attrition, and make people responsible for what they send even when AI generated it.Chapters00:00 Live sheet syncs for a brand01:04 Buying time at a premium03:01 Command-line prompting03:45 The outreach machine and its rate limits08:43 Where's the finish line?13:02 Self-healing loops15:10 Open rate is a trap18:01 LTV, profit, and rebills21:35 How the models feel to use26:41 CLI trade-offs and the max plan36:02 Technically good still needs taste38:31 Names people get excited about42:40 Creators and human leverage46:10 Attrition and productivity50:11 Demo: the ad launcher

In this episode of 1 to 100, Finn Brown and Pablo Peniche of Aqua Voice join Jack and Rishabh to talk about voice input, AI agents, and how people will work with software.Finn explains the problem that led to Aqua Voice. He has used voice input since school because dyslexia made typing difficult. Aqua builds on that use case with app context, technical terms, acronyms, edits, commands, and faster transcription.Finn and Pablo describe voice as a way to express ideas before compressing them into text. Speaking lets people think out loud, brainstorm, and give detailed instructions to models. This is useful for programmers who move between coding tools, agents, and terminals.The conversation covers the limits of voice control. GUIs and terminals were designed around keyboards and mice. Voice works well for dictation, editing, and clear commands. It becomes harder when a user must identify windows, objects, or actions with ambiguous references.Pablo explains why programmers use Aqua across their workflow. They move between coding agents, chat applications, and editors. A consistent voice layer can preserve speed, accuracy, and context across each environment.Finn describes a future where voice helps people manage parallel work across humans and agents. A manager could use voice to start loops, review tickets, involve a designer, and monitor the work. The group also discusses meetings that produce clear build instructions.Meetings often contain conflicting suggestions that sound compatible in the moment. An agent needs project context to identify disagreements and create a clear handoff before other agents begin the work.They also discuss Aqua’s API and the role of transcription quality in agent systems. Errors at the input stage can spread through the rest of the workflow. Context-aware transcription can help agents communicate with people through voice, text, and other interfaces.Chapters00:00 Introspection in Tech: A New Era01:54 AI Adoption and Enterprise Readiness04:05 Rising expectations for AI06:34 Context-aware voice input08:13 The Role of Voice in Programming Workflows11:05 Articulating Ideas: The Power of Voice12:53 Interfaces built for voice24:13 Parallelization and Voice Integration25:40 The Virtualized Office Experience26:43 Synthesis and Clarification in Meetings28:03 Agent Ecology and Safety30:10 A voice-driven hackathon38:23 Empowerment Through Voice Technology40:37 Differentiating Factors in Voice Technology50:54 The Future of Voice Solutions in Business

In this episode of 1 to 100, Evan Lee explains his move from Shoelace, which evolved into a growth marketing agency, to Motion. He built Motion’s marketing, sales, customer success, and support functions. Motion now has about 85 people and has raised $65 million in total capital.AI changed Motion’s hiring priorities. The company shifted toward engineering and spent about a year and a half building internal AI infrastructure. Motion did not reduce headcount because of AI. Employees use the company’s internal product in their daily work.Evan’s hiring approach starts with conviction in the business. Motion develops people internally and looks for candidates who are undervalued by the market but have high potential. Its interview process covers instinct, admiration, team impact, intellectual honesty, communication, storytelling, and judgment.Evan describes Runneth as a virtual machine for each brand. It stays active, works across the team, and provides a shared source of truth. Runneth pulls in reviews, competitor information, and news to build a context engine, persona matrix, and messaging matrix.Teams can use Runneth through Slack to create ads, generate creator briefs, review creative, and apply naming conventions. Evan says users can also store media-buying logic in the system. Motion is focusing the product on marketing.Evan argues that marketers should build strong marketing judgment as AI makes more execution available. He describes a durable take as an opinion rooted in the brand, the customer, culture, and the market. Developing those opinions requires absorbing more information and connecting it to the work.The conversation covers outcome-based software, AI budgets, templates, and the growing importance of strategy. Evan also explains the name Runneth and discusses how trust builds through repeated results.Chapters00:00 Evan and Motion02:48 Motion’s growth and AI shift05:33 Hiring for potential11:03 Runneth and a computer for each brand14:49 Brand context and work in Slack19:24 Expert agents and templates21:39 Outside expertise and marketing skills26:36 Marketing judgment and durable takes33:07 Paying for outcomes36:25 The singularity38:49 The name Runneth

In this episode of 1 to 100, Alex Beller from Postscript joins us to talk about rebuilding a scaled SaaS company from scratch with AI coding agents, moving from SMS into a broader marketing platform, and managing the organizational change that follows.Postscript has about 20,000 customers, an application shaped by eight years of work, and roughly 50 engineers. Alex explains why Adam started rebuilding the product alone as an experiment, then turned it into the company’s new direction after seeing what AI-native development could do.The new product, first called Acorn, competed with Oak, the existing Postscript platform. The company eventually decided to move all customers to Acorn and call it Postscript 2.0. Alex says the new codebase created a 10x to 50x improvement in speed and capacity compared with the old system.That rebuild also changed the product strategy. Email is no longer an add-on to an SMS platform. Email, SMS, and other channels are part of an AI-native CRM. Alex explains why starting from scratch made this multi-channel product easier to build.The internal transition was difficult. Some employees were excited by the new direction. Others were skeptical because the original system contained years of their work, including mistakes made by the founding team. Postscript gave people time to adjust, but some employees left and the company made proactive leadership changes.We also talk about the difference between adding AI to an old workflow and building around AI from the start. Alex argues that old CDPs are built around passive, structured data. The new system uses brand context, customer behavior, reviews, surveys, and results from ongoing experiments to create more proactive and personalized messages.Alex expects this to create revenue gains as well as cost savings. He describes a possible future where a small retention team can manage more work without an agency, while the software handles more of the campaign creation and orchestration. Adoption will take time. The product still needs to connect with familiar tools such as a brand’s marketing calendar and help teams move toward this new operating model.The lesson for SaaS founders is that some AI transformations may require a founder-led rebuild from scratch. One high-agency founder can build the new system while the rest of the company runs the existing business. It is painful and demanding, but it may be the fastest way to escape the constraints of an old product and old operating model.Chapters00:00 Cold open: tech at home, staying offline01:53 The co-CEO experiment and Adam going quiet04:10 Rebuilding Postscript from scratch08:03 Migration, the email launch, and multi-channel08:59 Telling the team: skeptics and change management12:30 Managing believers and leavers14:47 A customer churns for email and SMS17:35 Identity, dilution, and the advisory board19:54 Investors, category outcomes, business health23:02 Bolt-on AI versus AI-native26:00 Retention teams, agencies of one, revenue alpha28:48 Adoption stories: FERMAT and Foreplay34:16 Advice for founders: the rebuild playbook39:34 Hiring after the rebuild