Episode 23: A $100K Hackathon With a $99 Entry Fee

A hackathon with a $100K first prize and a $99 entry fee. The winner is the app that makes the most money in 30 days. Is that a contest or a business model? Marc Lou says the fee is "the ticket to enter the contest". We are not so sure.Aivars also explains why he stopped streaming: you need to be a top player or a very entertaining person, and he says he is neither. We talk about why strong opinions get views, Asmongold and reaction content, and why we still start with native iOS.Also in this episode:- Discord's Rust rewrite and a 1-second launch on macOS- Kotlin Multiplatform for shared business logic- a teleprompter app with huge views, and AI voices vs our accents- Bouncer, Paweł's free Mac app that pauses AI agents when RAM runs low- T3 Code, running agents on a Linux box, and coding from your phone- the $200 Claude plan, the Codex app, and is Fable 5.2 a minor upgrade?Chapters:00:00 Intro00:15 Aivars stops streaming12:23 Why Aivars quit streaming22:56 Asmongold and reaction content25:51 Strong opinions get views31:57 Staying native on iOS34:11 Discord's Rust rewrite39:08 Kotlin Multiplatform41:51 Start with native iOS45:11 A teleprompter app with huge views47:15 AI voices and our accents50:52 Bouncer: pause agents when RAM runs low59:41 The $99 hackathon drama74:50 T3 Code and a Linux box81:04 Running agents on Linux85:28 T3 Code on mobile89:43 The $200 Claude plan and the Codex app93:03 Fable 5.2: a minor upgrade?99:49 Your commentsNew episode every week – subscribe so you don't miss the next one.

We can build almost anything in a weekend now. So what is still worth building? Paweł closed more than 10 apps in two weeks and still can't pick his next product. Aivars is fighting the urge to ship another app nobody will use.We talk about why viral weekend apps are lottery tickets, why tools for other developers rarely sell, and what a copycat can't clone: a community, deep knowledge of a niche, and a direction that works. Paweł saw the maker of one viral app share 25+ copycats with almost the same icon.Also in this episode:- Apple account bans and trademarked keywords- the touch screen MacBook, after Steve Jobs said no in 2010- Touch Bar vs touch screen, and Mac Studio vs a Linux box- will iPadOS and macOS merge?- what comes after the phone, and why there is no third phone OS- think agent firstOmarchy: https://omarchy.orgChapters:00:00 Intro00:15 What we've been up to03:00 Choosing what not to build06:17 Slowing down after a build spree10:14 What should we build now?13:38 Viral apps and tools for developers18:42 The golden era and the copycats23:54 Games you can't clone in a weekend30:13 The life of a copycat32:54 Distribution, ads and UGC37:18 Apple account bans and trademarks45:59 Omarchy and the end of subscriptions48:55 The touch screen MacBook56:42 Mac Studio, Linux box or VPS?59:28 Will macOS and iPadOS merge?63:23 What comes after the phone?69:04 Apple needs to innovate71:46 AI solved software, hardware is next73:49 Why is there no third phone OS?79:31 Think agent firstNew episode every week – subscribe so you don't miss the next one.

Aivars took third place in RevenueCat's Shipathon 2026 – #BuildInPublic, $10,000 – with FeedbackThread, the feedback tool he built live on stream. The grand prize went to Mio, a self-care habit tracker from a team that already had a similar app. Is that cheating, or the best lesson of the whole hackathon?We go through what actually won: apps with proven ideas and distribution behind them, teams beating solo developers, a water tracker and a food journal in the most viral category. Then why building an app now takes 40 minutes, but App Store Connect still slows you down.Also in this episode:- Anthropic's Claude Duet, a proactive agent app, and why we can't try it (or Meta's Muse) from Europe and Thailand- Muse passed 10 million users with weaker models – why the best applied model beats the best model- how I use Grok Bot to prepare our episode outlines- will 99% of compute move to space by 2040?- cheap computer use, and why testing is the new bottleneckShipathon 2026 winners: https://www.revenuecat.com/blog/company/shipaton-2026-winnersClaude Duet: https://claude.com/resources/articles/claude-duetChapters:00:00 Aivars placed at Shipathon00:43 What is RevenueCat's Shipathon?02:13 FeedbackThread and the Build in Public category05:52 What third place wins07:11 26,000 participants and why Aivars expected to lose08:28 Using FeedbackThread in my app09:51 The grand prize: a twist on an app that already worked13:36 Teams beat solo developers14:32 Will a habit tracker last?15:43 What Build in Public winners should share19:43 Game, Peace Prize and design winners22:34 Learn to win hackathons or to make money?25:19 Most viral: water and food trackers28:42 No new ideas, only new twists29:41 An app in 40 minutes, App Store Connect is the bottleneck33:02 Claude Duet and proactive agents35:04 Grok Bot preps our episode outlines37:54 Muse has 10M users with weaker models39:09 Where does this leave Siri?42:59 Keeping up with AI is impossible44:54 The best applied model wins49:09 Agents on a VM in the cloud50:34 99% of compute in space by 2040?52:31 Cheaper computer use for testing54:30 Testing is the new bottleneck56:20 Wrap-upNew episode every week – subscribe so you don't miss the next one.

Sean Allen is shutting down his iOS course business because AI wiped out his sales – and I think the boilerplate market I sat on for years is drying up the same way. In episode 20 of Fired by Robots, Aivars and I talk about who still pays for developer courses and templates when an agent can teach you, or build the app from your own projects. I'm now releasing my React Native boilerplate for free.Also in this episode:- iPhone Duo pre-orders sold out in minutes, and Apple is reviewing iPhone Duo apps fast – Aivars' quickest took three minutes- why I'm keeping VidNotes native while still using Expo for some apps, and why Aivars skips cross-platform tools entirely- Amazon, Delta, Yelp, JPMorgan and Schwab blocking outside AI agents- why we're not testing every new model ("we are old people, we have to be pragmatic")- Aivars' simple setup with Codex Dots- the "SaaS is dead" debate: Aivars thinks both sides are wrong, I think agents from the big companies will replace a lot of softwareSean Allen's courses (closing November 24): https://seanallen.teachable.com/Michael Tsai on the shutdown: https://mjtsai.com/blog/2026/10/21/ios-dev-course-going-out-of-business/Sites blocking AI agents: https://officechai.com/ai/amazon-jpmorgan-schwab-and-delta-are-all-blocking-ai-agents-heres-why/VidNotes: https://vidnotes.appChapters:00:00 Back after a two-week break01:08 Apple's fast iPhone Duo reviews02:36 iPhone Duo sold out in minutes05:08 Building for foldables: my teleprompter app07:04 Four apps approved, one in 3 minutes09:00 Moving an app from Swift to React Native10:44 Rebuilding my idle game with Opus 5.511:43 Is React Native dying?14:41 Expo vs native for my apps15:37 Second thoughts on the VidNotes rewrite19:22 Indie devs: start native20:36 Are we testing the new models?24:16 "We are old people"24:43 Grok Bot and my agent systems26:36 How Aivars uses Codex Dots29:03 Too much system building31:03 Websites are blocking AI agents33:12 Is SaaS dead?37:30 Are we outsourcing our thinking?39:50 Why agents will replace SaaS42:11 Sean Allen shuts down his courses44:03 AI tutoring and learning apps48:15 The boilerplate market dried up50:23 Aivars: you don't need a boilerplate52:11 How fast our approach changed this year53:10 Wrap-upNew episode every week – subscribe so you don't miss the next one.

After about three months as "the Codex guy", I went back to Claude this week. I threw everything at Opus 5.5 – demo videos for my mobile app, UI fixes for my teleprompter app, a new version of my beat-making app – and didn't hit the five-hour limit once. Aivars had the same experience after months of rationing Fable, and we talk about why that changes how we work.Then the part that matters if you build apps: OpenAI's DevDay. Sign in with ChatGPT lets Plus and Pro users spend their plan inside partner tools like Notion, and Dots are always-on agents with their own cloud computer that connect to 4,000+ apps. If agents start choosing tools for people, will they pick your app? We talk about MCP, agent catalogs, and what I'm already seeing with VidNotes.Also in this episode:- Aivars had Claude Code cut a moto vlog from 360, DJI and iPhone footage while burning very few tokens- why AI demo videos are a commodity now- I turned my SaaS into a skills package that runs inside Claude, Codex and Hermes- too many token resets and the pressure to burn them- iPhone Duo support: size classes, iPad layouts, and a teleprompter on the rear camera- why we stopped chasing every new model- Aivars is back on Windows after a week on LinuxClaude Opus 5.5: https://www.anthropic.com/claude-opus-5-5Claude Sonnet 5.5: https://www.anthropic.com/claude-sonnet-5-5OpenAI DevDay 2026 recap: https://openai.com/index/devday-2026-recap/VidNotes: https://vidnotes.appChapters:00:00 Bangkok floods02:51 Why I went back to Claude06:34 Aivars on Opus 5.5 and the five-hour limits09:22 Claude Code edits a moto vlog11:06 Demo videos are a commodity now14:01 My SaaS became a skills package17:16 Staying model-agnostic19:59 Too many token resets21:43 iPhone Duo support in practice24:12 A teleprompter for iPhone Duo27:58 Why we stopped chasing new models37:13 Back to Windows after a week on Linux40:54 Codex computer use and dropping a VPS43:56 DevDay: Sign in with ChatGPT48:22 Make your app available to AI agents50:29 VidNotes: MCP, web and credits54:46 OpenAI Dots: always-on agents56:41 One chatbot for everything?58:52 My biggest enemy is focus1:00:17 Wrap-upNew episode every week – subscribe so you don't miss the next one.

One video edit used about 10% of Aivars's AI plan. With GPT-6 Sol and Claude Opus 5.5 newly released, we ask whether cheaper models will let us do more before we hit usage limits.We discuss the reported model improvements, our subscription limits, and the video editing work that uses our allowance. Neither of us had fully tested the new models when we recorded this episode, so this is a discussion of the releases and our existing workflows, not a head-to-head test.We also share how we use AI to monitor flight prices and manage Google Ads. Then we discuss Jev, a model for structured decisions, and why a structured answer can still be wrong. Finally, we cover Expo SDK 58 beta, the VidNotes move to Expo, and Aivars's work to adapt his apps for iPhone Duo.GPT-6 Sol: https://developers.openai.com/api/docs/models/gpt-6-solClaude Opus 5.5: https://www.anthropic.com/claude-opus-5-5Jev: https://typesafe.ai/blog/introducing-system-one-models-and-jevExpo SDK 58 beta: https://expo.dev/changelog/sdk-58-betaiPhone Duo developer resources: https://developer.apple.com/iphone-duo/What task uses the most of your AI allowance? Tell us in the comments.Chapters:00:00 Opening00:15 GPT-6 Sol and Claude Opus 5.502:31 Codex, Xcode, and model access05:21 Model efficiency and automation costs10:07 AI video editing and usage costs13:48 Claude, Codex, and subscription limits19:12 ChatGPT and Claude interface changes25:24 Using AI to monitor flight prices28:59 Launching Google Ads with AI32:28 Jev and structured AI decisions43:48 Expo SDK 58 and the VidNotes rewrite49:07 Adapting apps for iPhone Duo58:13 Your questions and future episodesSubscribe to Fired by Robots for more conversations about AI and app development.

Shopify is moving its mobile apps from React Native to Swift and Kotlin after AI changed the cost of native development.Does that mean indie app developers should follow?Aivars thinks beginners should start with native development. I still think React Native and Expo suit many cross-platform products.We compare native development, React Native, Flutter, and Dart Native. We also discuss performance, complexity, over-the-air updates, and iPhone Duo support.Shopify’s native development announcement:https://shopify.engineering/back-to-nativeShopify’s Shop app migration:https://shopify.engineering/shop-app-migrationApple’s iPhone Duo developer resources:https://developer.apple.com/iphone-duo/Chapters:00:00 Paweł’s new development contract02:11 Apple announces iPhone Duo08:09 Preparing VidNotes for iPhone Duo19:32 Android foldables versus iPhone Duo23:08 Apple’s new Siri and AI29:02 Apple’s localization problem33:24 Does Apple still innovate?39:13 macOS 27 Golden Gate43:02 Shopify drops React Native50:17 Native versus cross-platform for beginners01:06:10 Flutter, Dart Native, and HabitKit01:19:04 Listener feedback and final thoughtsSubscribe for weekly discussions about building, shipping, and growing mobile applications.

Is GPT-6 Astra truly the best AI model for indie app builders? We do not agree.Paweł sees a major improvement in reasoning, automation, coding, and subscription value.Aivars sees a better model, but not a revolution. Claude still produced better iOS designs during his direct comparison.We discuss real work from the past week:• Astra fixed an iOS keyboard bug that other models could not solve.• Astra created an interactive 3D avatar using Blender.• Claude defeated Astra during an app design comparison.• Our apps reached approximately $4,600 and $2,600 in MRR.• Paweł spent about $200 sending more than 40 Upwork proposals.• One unexpected Upwork message produced a $500 weekend project.• Hermes agents could become a reusable team of digital workers.Fired by Robots follows two indie developers building products without a corporate playbook.Website:https://firedbyrobots.comSpotify:https://open.spotify.com/show/033k2vKCq4wYBIdzX2whcZApple Podcasts:https://podcasts.apple.com/us/podcast/fired-by-robots/id1896779874CHAPTERS00:00 Introducing GPT-6 Astra02:52 Is Astra really better?07:03 Astra fixes a week-long iOS bug08:10 Claude still wins at app design09:13 Building a 3D avatar with Astra13:05 Astra and Claude subscription value20:04 Where Astra still needs supervision23:06 Weekly shipping commitments28:00 August app revenue and hard paywalls30:46 Upwork is now pay-to-compete33:11 A $500 weekend automation project41:21 Freelancing, pressure, and focus46:40 Switching from Windows to Omarchy49:37 Building a team of Hermes agents59:39 Apple event and iOS 27#IndieDev #AICoding #GPT6Astra

The conversation covers a range of topics including personal commitments, software development, AI models, and the future of software and technology. It delves into the challenges and opportunities in the software development industry, the impact of AI models on coding and design, and the evolving landscape of software and technology. The conversation covers a range of topics including UI development, streaming, Linux usage, cloud-based development, Grogbot, marketing strategies, Apple's surprise event, and the behavior of software developers. The discussion also delves into the challenges of AI projects and the dynamics of the developer community.TakeawaysPersonal commitments and focus are essential for project completion.AI models like Fable and Codex are changing the landscape of coding and design.The future of software development may involve a shift towards AI-powered tools and a reimagining of the software industry. UI development and streamingLinux usage and cloud-based developmentGrogbot and marketing strategiesApple's surprise event and developer community dynamicsChapters00:00 Personal Commitments and Project Completion03:21 The Impact of AI Models on Coding and Design10:17 The Future of Software Development and Technology21:00 Challenges and Opportunities in the Software Development Industry27:55 The Role of AI Models in Software Development37:02 UI Development and Streaming38:00 Linux Usage and Cloud-Based Development39:29 Grogbot and Marketing Strategies01:00:07 Apple's Surprise Event and Developer Community Dynamics

The conversation delves into the exploration of Omachi and Linux, discussing the features, integration, and considerations for hardware and operating systems. It also covers the use of Grogbot and AI agents for various tasks and applications. The conversation covers a range of topics including email marketing, AI model comparison, Grogbot use case, AI subscriptions, gaming and AI automation, reflections on the Apple ecosystem and development, focus on specific app development and marketing, and accountability and future plans.TakeawaysExploring the features and integration of Omachi on Linux.Considerations for hardware and operating systems when evaluating new machines.The use of Grogbot and AI agents for various tasks and applications. AI model comparison is essential for understanding the quality of AI-generated content.Grogbot's use case and functionality provide valuable insights into AI automation and content analysis.Chapters00:00 Exploring Omachi and Linux02:47 Omachi Features and Integration07:00 Considerations for Hardware and OS26:13 Grogbot and AI Agents43:55 Email Marketing and AI Writing45:12 Comparison of AI Models46:33 Grogbot Use Case and Functionality47:18 Discussion on AI Subscriptions and Usage01:01:16 Gaming and AI Automation01:11:16 Reflections on Apple Ecosystem and Development01:16:34 Focus on Specific App Development and Marketing01:20:09 Accountability and Future Plans