Indie Dev Talks: Building and Scaling Mobile (and Web) Apps

Ivan Terekhin

Indie Dev Talks is a weekly conversation about building software as one person. Host Ivan Terekhin runs a portfolio of 10+ mobile apps and is scaling them to $10K/month before moving into game development. Recurring voices include Philipp, Vaclav, and Faruxx. Each episode brings working indie developers together to compare what actually ships revenue: pricing and paywalls, App Store optimization, RevenueCat and subscription mechanics, going from idea to release in a few days, and staying profitable with no team and no funding. Expect real numbers, honest teardowns, and the tradeoffs nobody puts in a launch thread. New episodes every week.

  1. 6d ago

    Indie Dev Talks #16 — LLMs are not the future, GLM over Opus, $50K is the ad floor

    Tom drops in for a rare appearance and the room gets straight into it, with Ivan, Vaclav, Tom, Faruxx, and Philipp debating whether large language models are even the right path forward or whether auto-regressive transformers have already peaked. Tom explains why he moved off Anthropic to a mix of GLM, OpenAI's Codex, and a bare-bones agent he runs with almost no tooling, while Ivan argues the Fable panic was mostly political theatre and the cyber-security danger is overhyped. The group works through whether any of them would actually pay double for a frontier model when Opus 4.8 already clears the bar, why AI is quietly making junior developers look affordable again, and why taste is the one thing these models still cannot fake. In the back half Ivan heads to the dentist and the conversation turns tactical: why paid ads under 50K are mostly wasted, how cross promotion and maintenance beat one-shot launches, Faruxx reviving his Notion widgets around SEO against outdated competitors, Philipp going after boring education incumbents with Lightbulb Learning, and a Reddit marketing playbook for getting visible without getting banned. Timestamps 00:00 — Teeth, dentists, and catching up 03:23 — Tom joins: are LLMs even the future 05:00 — Ditching Anthropic for GLM, Codex, and the PI agent 09:19 — Fable, self-improving AI, and who gets the powerful models 16:01 — Ivan: the Fable panic is political, the danger is hype 20:47 — Pay double for Fable, or is Opus 4.8 enough 31:00 — When junior devs get cheaper than Opus 37:33 — Taste is the last thing LLMs cannot fake 38:33 — Losing track of your own apps when you ship too fast 44:48 — Why paid ads under 50K are a waste 53:41 — Ivan leaves, Faru revives his Notion widgets 59:33 — Beating boring, outdated competitors 01:04:32 — Faru's Reddit marketing playbook 01:06:37 — Lightbulb Learning soft launch and back-to-school timing

    1h 11m
  2. Jun 23

    Indie Dev Talks #15 — ASO Tools Don't Work, 1K MRR on Sober Tracker, GEO Is the New Game

    Heads up, the opening few (10-15) minutes have rough audio on Ivan's side. Ivan opens with a political statement on ASO: keyword hunting is cope, the popular tools are cope, what actually moves the needle is real value plus real reviews plus enough volume that the store ranks you for the obvious words. He hit 1K MRR (but more revenue) on Sober Tracker today and is fielding the same question on repeat: how did you get the traffic. Answer is none of the sneaky keyword tactics everyone wants the answer to be. New cast member Valentine joins from France, a freelance SEO consultant pivoting into GEO, building Redchurn (a revenue intelligence app aimed at the gap RevenueCat leaves around churn, LTV by source, and email reengagement workflows) and Be.Console (an SEO/GEO competitor tracking tool he launched yesterday). He'll be a sponsor at the RevenueCat Shipaton in August. Vaclav reveals he bought two of Ivan's premium lifetimes yesterday for $100, which Ivan immediately uses as the perfect signal-vs-noise example: spikes from individual buyers mean nothing at this scale and indie hackers running A/B tests on 100 users are generating pure chaos rather than data. The group dig into Marc Lou's 500-day journey to 20K on data.fast, the bipolar emotional loop of any builder podcast watched honestly, Florian who sold his last company for $17M then struggled for a year and a half before his current podcast took off, and the imposter syndrome epidemic underneath all of it. Ivan announces Discipline Tracker, his new app submitted today, targeted at the dopamine hole that opens when someone drops alcohol via Sober Tracker. The vitamin-app vs painkiller-app framework comes up: sobriety changes lives, smoking is a slower kill so the urgency isn't there, water tracking has zero emotional pull. Closing thirty minutes are a deep GEO SEO masterclass from Valentine, why ChatGPT doesn't need authority signals to pick up your domain (just clean markdown, schema.org JSON-LD, MDX pages), why net linking still matters for Google, and the listicle outreach play Ivan should be running for Sober Tracker against existing best-of lists. Plus a philosophy hour digression on luck, with Ivan defending "I got lucky and that's fine" and Valentine refusing to call it that. Timestamps 00:00 — ASO political statement, keyword hunting is cope 01:00 — Sober Tracker hit 1K MRR today 02:30 — Lifetime over yearly, monthly is wrong for sobriety apps 03:30 — The Australian renewing weekly for a 3 months 04:00 — Valentine joins from France, freelance SEO consultant 07:30 — Find one trusted source, be that one for one person 09:00 — Marc Lou's 500-day journey to 20K MRR 11:00 — Counterproductive advice, launch and kill after two months 13:00 — Tea Timer at zero monetisation, vitamin apps vs painkiller apps 16:00 — Valentine's Redchurn revenue intelligence app 18:00 — The LTV by source view that nobody offers 20:30 — Email reengagement workflows, 30-40% return rate 23:00 — Vaclav bought 2 lifetimes for $100, the perfect noise example (:D). 25:00 — Why 100 users is not data, it's just chaos 27:30 — Indie hackers should build for things they actually care about 32:00 — Sample size matters, even at 10K MAU 35:00 — Supplement Tracker emotional paywall playbook 37:30 — Sober Tracker emotional pull, smoke and water trackers don't have it 40:00 — Audio quality acknowledged, rain in Tbilisi 41:00 — Discipline Tracker submitted today as Sober Tracker upsell 42:30 — Dopamine hole when people drop alcohol 46:30 — Water tracker has no emotional message, just utility 48:30 — Valentine's 5% hit rate, the volume game 52:30 — Florian sold for $17M, struggled 1.5 years before This Trip took off 56:00 — Marc Lou's data.fast video, the bipolar daily mood swings 1:00:00 — Ship or Die from Marc Lou and Jack Friks, community discipline as the product 1:02:30 — Luck debate begins, Ivan vs Valentine 1:11:30 — Vaclav coins "Philosophy Hour with Ivan and Valentine" 1:12:00 — Ivan on singl

    1h 35m
  3. Jun 11

    Indie Dev Talks #14 — Zero Ads to $5K (almost) MRR, Claude Fable Maxing, Is the AI Bubble Popping?

    Ivan opens with a strong month, closing in on almost $5K MRR with traffic flowing from Bing, the App Store, and Google Play, all without ads or social media pushing, and the crew digs into why it works: deep localization paying off in Polish and Dutch markets, reviews that prove real product love including one from a daughter whose dad is sober and back in the family, and the long game of SEO, ASO and GEO versus the paid-ads weekly-subscription churn machine. Vaclav zooms out on the AI economy: subsidized consumer subscriptions versus enterprise pricing, the mad IPO rush, dot-com bubble parallels, and what happens to Lovable, Bolt and the wrapper economy if prices rise. Everyone compares notes on Claude Fable: dramatically faster than Opus 4.8, autonomous enough to run a full daily SEO pipeline with commits, deploys and indexing on its own, the running joke about a remote control app with just buttons 1 and 2, and whether Anthropic will really pull Fable from subscriptions in 10 days. Philipp shares that his 9,000-page education platform is nearly ready to launch with 70 interactive teaching blocks Claude designed, and reveals Mumsnet as a massive untapped channel for family-focused apps. Faruxx reports on Chinese and open-source models falling short on real coding tasks despite the price advantage, and Ivan explains why he cancelled a Reddit-marketing SaaS. Michael waves in briefly from a train to Manchester. Timestamps: 00:00:00 — Catch-up: nearly $5K MRR, traffic from Bing, App Store and Play Store 00:02:13 — AI economics: subsidized subscriptions, enterprise money, wrapper risk (Lovable, Bolt) 00:07:31 — IPO rush and dot-com bubble parallels 00:11:00 — Claude vs Codex vs Gemini: model personalities 00:13:23 — Fable hands-on: faster than Opus 4.8, Philipp's 9,000-page launch prep 00:17:00 — Fable maxing: token burn and fully autonomous SEO runs 00:22:23 — The press-1-press-2 remote control app idea 00:25:15 — 10 days left: will Fable stay in subscriptions? 00:30:23 — Chinese and open-source models: DeepSeek, Kimi, and the benchmark gap 00:40:06 — Growing without ads: localization moat, two business models compared 00:49:10 — Ivan's journey zoomed out: failed markets, PPP pricing, a $5 lifetime first sale 00:56:44 — The review that hit home, and why buying reviews never pays 01:00:04 — Luck vs persistence: keep rolling the dice 01:09:01 — X as community, not distribution channel 01:15:48 — Reddit-marketing SaaS verdict and Mumsnet as an untapped goldmine 01:23:21 — Wrap-up

    1h 25m
  4. Jun 3

    Indie Dev Talks #13 — Pricing Experiment Gone Wrong, Bing Beats Google for ChatGPT, Never Kill Apps

    Ivan opens with the pricing experiment he just reversed: a 30 percent raise across Sober Tracker that immediately killed lifetime conversions, reverted yesterday, lifetimes back the next day. Lesson: don't touch anything, just wait. Lifetime plus yearly is 86 percent of his revenue and the volume is what carries the floor. Sober Tracker Android is closing on 600 reviews and growing 10-15 reviews per day, May revenue tripled off the Android price drop. Philipp returns from Marmaris with a six hour Claude 4.8 session done from a sun lounger that turned into the orchestration plan for shipping Lightbulb Learning, walks the group through how the 9,000 pages actually break down (12 subjects, 4 exam boards, topics, subtopics, five year groups), and the Claude Design 75 rendering blocks engine. The episode's most quoted moment is Vaclav calling out Claude 4.8 as "cosplaying as an autistic developer telling me to call it a day after 10 PRs." Faruxx walks through the Codex switch for actual development work while keeping Claude Code for marketing and SEO conversation, plus the wrapper he found called Claude Code CR with 24 specialised agents. Vaclav drops the actual leverage of the episode: 60 percent of his ChatGPT referral traffic comes from being indexed in Bing Webmaster Tools, not Google. The group dig into the GEO era replacing SEO and the "we're the early gold diggers" framing. Faruxx publicly tells Michael he should never have killed Changelog, names his own year-old Widgets for Notion as proof that abandoned apps still earn, and the group dogpile until Michael agrees to spin Changelog back up. Michael unveils HiSEO/HiGEO, an autopilot page generator for competitor alternative pages priced against SEO Matic at $400 a month. Plus the friend Luke's Logo Lattice that came out of a similar quitting moment, the Philipp-was-a-hairdresser-in-Paris reveal, and the hairdressing app idea brainstorm that ends in AI mirrors showing dreadlocks and mohawks. Timestamps: 00:00 — Price raise of 30 percent reverted in one day, lifetimes back the next morning 01:30 — Lifetime plus yearly is 86 percent of revenue, monthly is legacy residue 02:30 — The Australian who renewed weekly 12 times instead of buying lifetime 04:00 — Philipp back from Marmaris, Claude 4.8 from a sun lounger 05:30 — Claude 4.8 refusing the "how much money am I gonna make" conversation 06:30 — Six hour orchestration.md built on the beach 08:00 — Claude 4.8 vs Codex vs Chat GPT pipeline, which finds what 09:30 — Ivan on Claude 4.8 being more critical, "shut up I'm the entrepreneur" 11:00 — Vaclav: Claude cosplaying as an autistic developer telling me to rest 12:30 — Faruxx on Codex for development, Claude Code for marketing and SEO 14:30 — Claude Code CR wrapper with 24 agents 16:00 — Ivan dropping landing page updates, blog posts as motivational morning ritual 18:30 — Vaclav: 60 percent of ChatGPT referrals come from Bing Webmaster Tools 21:00 — Geo vs SEO, the early gold diggers framing 24:00 — Philipp on the 9,000 page roadmap, 12 subjects, 4 boards, topics, subtopics 26:30 — The cached pages with marked exam questions plan 28:00 — Vaclav teasing on the 9,000 page deadline, you said next week 30:00 — Faruxx's Widgets for Notion, year of zero updates, still earning 33:30 — Faruxx publicly telling Michael he shouldn't have killed Changelog 35:00 — The never-kill-apps thesis, Marc Lou's poop analyzer at 10 bucks a month 38:30 — Ivan: sometimes the lower price is the winning condition 40:30 — Sober Tracker May revenue tripled off the Android price cut 41:30 — Approaching 600 Android reviews at 10-15 per day pace 42:30 — Ivan on getting zero customers from X, different audience 44:00 — Michael joins, the SEO Matic alternative pivot reveal 45:30 — HiSEO/HiGEO autopilot page generator for competitor alternatives 48:30 — How it works, brand matching the trickiest part 51:00 — Michael agreeing to spin Changelog back up 53:30 — Release notes as free pSEO surface for solo d

    1h 20m
  5. May 20

    Indie Dev Talks #12 — Faruxx Pivots to Job Hunting, 3K MRR Hit, Vibe Coding Is the Golden Era

    Ivan opens with a confirmed 3K monthly revenue across the portfolio and pushes back against friends telling him it's not stable, framing the lifetime line as a normal distribution issue rather than instability. He walks through the in-app upsell he launched for users past the 30 day streak threshold, three lifetime conversions already, his updated screenshots now showing his own 200 day streak, the sober garden feature with the growing tree generated through Gemini MCP graphics, and the Apple Search Ads Germany campaign finally hitting profitability while Google Ads Canada produced zero conversions. Philipp joins from his pre-Turkey trip frenzy, eight thousand pages ready to render with Claude Design pulling 75 rendering blocks, going up against Medley.ai at half their price, and a feedback-form-before-paywall plan that Ivan calls out as Apple-bannable while Vaclav defends it as farming signal from committed users. Faruxx returns after the wedding break, walks through ripping Pocket Claw off the broken upstream and pivoting to a job-hunting niche seeded by an engineering SaaS client who's offering Telegram group access for distribution, with the group flagging Greg Isenberg's recent video covering the same niche-agent thesis. Michael surfaces mid-call to share his Changelog journey has paused and his web design agency is taking off, with two clients finding him via ChatGPT recommending "local web designer" through some path he can't reproduce. The episode closes on the vibe coding debate, Ivan dismissing Codex for context-window reasons and confessing he runs Claude on bypass permissions, Vaclav defending that companies were shipping insecure code long before AI, and Philipp running a multi-step chat-prompts-Claude-Claude-writes-Chat-reviews commit pipeline. Timestamps 00:00 — 3K revenue confirmed, the not-stable-but-actually-stable normal distribution argument 02:00 — In-app upsell launched on 30 day streak, three lifetime conversions 03:30 — Updated screenshots with Ivan's own 200 day streak 04:30 — Sober garden feature, Gemini MCP graphics, the tree and water drops 06:00 — Ported to Smoke Tracker but Smoke Tracker isn't working 07:30 — Apple Search Ads Germany hits profitability, increasing budget 08:30 — Google Ads Canada wasted, zero conversions 09:10 — Philipp joins, going to Turkey tomorrow, eight thousand pages ready 10:30 — Lightbulb Learning state, Claude Design pulling 75 rendering blocks 13:00 — Pricing at 12.99 against Medley at 25 14:30 — Three free renders, feedback form before paywall plan 16:30 — Ivan flagging Apple-bannable dark pattern from his own history 18:00 — Vaclav defends Philipp's intent, farming signal from committed users 20:00 — Ivan only asks for review on day three milestone 22:14 — Faruxx joins, post-wedding, getting back to work 23:00 — Pocket Claw pivot off the broken upstream onto custom harness 24:30 — The job-hunting agent niche idea 26:30 — Philipp suggests CV cover letter auto-generation 28:00 — Engineering niche, paying customers, Telegram group distribution 29:30 — Greg Isenberg's video covering the same niche-agent thesis 31:00 — Blood work, loneliness, old people app ideas, the decade-old idea problem 32:30 — Find your angle, not your unique idea 35:00 — The engineering SaaS client backstory and Telegram distribution 38:00 — LinkedIn pays at 35 per month, the willing-to-pay engineer market 39:00 — Niche within niche, international job hunting 41:30 — Cursor Composer mentioned, Ivan considering it 44:00 — Ivan dismissing Codex, hates context window compacting 45:00 — The vibe coding debate, can you ship a real app vibe coded 46:30 — Ivan totally vibe coded Sober Tracker, security and GDPR handled 48:30 — Vaclav on real companies shipping insecure code long before AI 50:00 — Philipp's chat-prompts-Claude-Claude-Chat-reviews commit pipeline 52:30 — Ivan running Claude on bypass permissions, the golden era 54:06 — Michael joins, web design agency taking off 56:00 — Two clients

    1h 24m
  6. May 13

    Indie Dev Talks #11 — Approaching 3K Monthly, Stop Thinking and Just Build, German Ads Working

    Ivan checks in from Tbilisi running a 4am wake-up, 5am gym, big-headphones monk mode and announces he's two to three days from shipping Mirror Brief, a looksmaxxing app he built as an experiment copying another builder's playbook and consciously tying back to Sober Tracker via the puffy-face-from-drinking angle. He walks Vaclav through the new Apple Search Ads campaign in Germany, the proper attribution tags he set up this run, and confirms ad-driven users are actually starting trials now with install-to-trial at 2.3% and trial-to-paid at 50%. Vaclav presses him on Lukasz's recent suggestion to run 100 apps and pull 10 bucks from each, and Ivan pushes back with a soulmate metaphor, the same reason he's about to delete WinFast and WinGym from his life entirely. They dig into the Sober Tracker organic growth theory of reviews driving store rank rather than any active tweaking, with Android approaching 500 reviews. Ivan shares the February 2025 decision to stop thinking and just flow, his old post-grad professor's "stop reading and start writing" line that took fifteen years to land, and why he deleted TikTok and Instagram from his phone as an addictive personality move. The pair close on the MRR-vs-lifetime debate with Claude regularly bullying Ivan about a stuck MRR while ignoring the lifetime line, the fact that 1% of 10 million struggling US drinkers is a lifetime business, and Ivan disclosing he's now approaching 3K in monthly revenue across the portfolio. Timestamps 00:00 — 4am wake-ups, 5am gym, big-headphones monk mode in Tbilisi 01:30 — Sober party means 9pm bedtime 03:00 — Mirror Brief teased, the Sober Tracker tie-in via puffy face 06:00 — Pre-launch anxiety despite 14 apps, what the looksmaxxing app does 08:00 — Apple Search Ads Germany, attribution tags set up this run 09:30 — Install-to-trial at 2.3%, trial-to-paid at 50% 10:30 — The bipolar ads-on, ads-off, ads-on journey 11:30 — Deleted TikTok and Instagram, the addictive personality reasoning 13:30 — Tbilisi's real fat dogs vs virtual feeds 14:30 — Monk mode and gold-digging, low life standards as an advantage 16:00 — Michael's burnout vs Ivan's nine months without validation 18:30 — Mirror Brief origin, copying a TikTok-ads looksmaxxing playbook as experiment 20:30 — Checkers launched, Supplement Tracker watch companion shipped in an hour 22:30 — Mirror Brief pricing, weekly plus lifetime, no trials 24:30 — Lukasz's 100-app suggestion, Ivan's soulmate metaphor pushback 27:00 — WinFast and WinGym headed for deletion, the cortisol-to-beer pipeline 32:30 — Sober Tracker organic theory, reviews drive rank 33:30 — Android approaching 500 reviews, iOS at 150 37:00 — Stack debates as overthinking, Flutter vs React Native vs native doesn't matter 40:00 — Stop reading and start writing, the post-grad professor line landing 15 years late 41:30 — The February 2025 decision to stop thinking and just flow 44:30 — When overthinking is actually justified 46:30 — Hitting bottom as the prerequisite to natural-mode 49:30 — Working alone is faster, can't collaborate at this pace anymore 51:30 — Next-gen indie hacker residency idea 53:00 — Claude bullying about stuck MRR, lifetime is real revenue 54:30 — Approaching 3K monthly total across the portfolio 56:00 — 1% of 10 million US drinkers is infinite market 58:00 — Behave like indie, not enterprise, scale your standards down 1:00:00 — Carve a niche inside someone else's big market 1:02:00 — Steep at 200 to 300 monthly, the caffeine tracking user request 1:05:00 — Code talker as profession, indie game dev still the endpoint 1:06:00 — Roll call, hoping for Michael, Adria, Lukasz next time

    1h 7m
  7. May 6

    Indie Dev Talks #10 — There Is No Perfect Launch, 7 Agents Running, OpenClaw Dead

    Vaclav opens with Philipp running seven Claude Code agents in parallel after dropping ChatGPT as the strict controller and just telling the model directly what he wants, a complete reversal from where he was a week ago. Philipp walks through the Lightbulb Learning roadmap, twelve subjects across four exam boards heading toward 3,000 pages, his decision to miss this exam season and warm up for August through Pinterest and Instagram, and the pricing fight against Medley.ai sitting on 1.7M in funding. Ivan shares that he's gone analytics-maxing on PostHog MCP, found 40% of Math Zen users never solving a problem after onboarding, redesigned the home page, and now believes the store boost he's seeing is the algorithm rewarding the 60% trial-to-paid conversion rather than anything he's actively doing. Faruxx pops in from a train platform to share he's getting married and walks through ripping Pocket Claw off OpenClaw and onto a custom agent harness because the upstream founder ships breaking changes every other day. The group dig into the OpenClaw hype cycle dying in two months flat, why Marc Lou doubts himself at 80K MRR, why Jack Friks won't push Post Bridge despite it being his money maker, and the long debate about launching imperfect, with Ivan offering his $5 lifetime for an entire year as the cautionary tale and Philipp pushing back that he wants 25 complete subjects on day one. 00:00 — Philipp running 5 to 7 Claude Code agents in parallel 02:00 — Dropping ChatGPT as the strict controller, taking direct command of Claude 04:00 — Lightbulb Learning roadmap, missing exam season, warming up for August 06:00 — Going up against Medley.ai with 1.7M in funding 07:30 — Google Search Console plus read-only service accounts for Claude 09:30 — Ivan analytics-maxing on PostHog MCP 10:30 — 40% of Math Zen users never solving a problem after onboarding 12:00 — Store boost theory, 60% trial-to-paid as the trigger 13:30 — Pricing debate, Philipp's 15 vs Medley's 25, Max Blade's double-it advice 16:00 — Faruxx joins from a train, getting married, Azerbaijani weddings 18:00 — Pocket Claw pivot off OpenClaw onto a custom agent harness 21:00 — The OpenClaw hype cycle, dead in two months flat 22:30 — Ivan's overconfidence this run, life-minimizing and indie-maxing 24:30 — Smoke Tracker stuck at 1K MAU, the passion theory vs sibling app 27:30 — A/B testing and staged rollouts in indie hacking 29:30 — Why TikTok is not organic for Ivan, Reddit and ASO are 31:00 — Mo Barron's prayer lock 100M shorts play 33:30 — Marc Lou doubting himself when 80K dropped to 78K 36:00 — Jack Friks not pushing Post Bridge despite the money 38:00 — 37signals founder who couldn't imagine scale beyond 3M 40:00 — Revenue vs profit, Rob Hallam splits with Tibo 41:30 — Indie hacking as the hedge for the AI unemployment era 42:30 — Philipp drops in on Ivan's stream and watches him make a cup of tea 44:00 — Jack Friks on why everyone can build but distribution still decides 45:30 — Philipp's developer-apps-fail theory, the greasy-hair lecture hall 48:00 — Vaclav calling out Philipp's perfectionist rabbit hole 50:00 — Launch imperfect, no perfect launch ever happens 52:30 — Ivan's $5 lifetime for nearly a year cautionary tale 1:00:00 — TikTok and Pinterest warm-up time, posting before launch 1:01:30 — Supplement Tracker onboarding inheriting accumulated reps 1:05:00 — Six weeks isn't long, the OCR Camera apps Philipp shelved 1:06:30 — Wrap, introvert energy and lunch

    1h 7m
  8. Apr 29

    Indie Dev Talks #9 — Apps on Autopilot, Claude 4.7 Therapy Mode, Identity Crisis

    Ivan checks in from Tbilisi after two weeks of 8-hour streams and a social life that's a stark contrast to monk-mode Bangkok, and admits he hates the state his apps are in: everything is on autopilot, MRR is around 1.5K, and he can no longer tell which input causes which output. He walks Vaclav through his next bet, a coaching tier for Sober Tracker inspired by a Sunflower CEO interview he ran through Claude, and explains why he's chasing the same thing on Math Zen with localization plays into Korea and a possible Huawei Play Store push for China. Vaclav shares that his Health 3 SEO had a big jump a month ago and has flatlined since, no matter what he changes, and Ivan brings up a Kobe Bryant clip about trusting the process and not over-tweaking. They dig into Claude 4.7's new emotional streak, with Philipp later sharing that Claude has been telling him to "have a rest, we'll come back to this tomorrow" mid-build, and Ivan offering a system prompt fix that killed the behavior for him. Philipp also walks through his rebuild of Lightbulb Learning using Claude design output and asks whether to drop Cursor for Claude Code. The episode closes on Ivan's coder identity crisis now that he never writes code by hand, Vaclav reframing it as the next compiler abstraction layer, and a roll call of who burned out this week. Timestamps 00:00 — Tbilisi life, 8-hour streams, Diablo 4 with friends 02:00 — Apps on autopilot, losing the sense of cause and effect at 1.5K MRR 04:00 — Coaching tier plan for Sober Tracker, running the Sunflower CEO interview through Claude 06:30 — Vaclav's Health 3 SEO bump and the four-week plateau since 09:00 — Kobe Bryant on trusting the process instead of tweaking 11:00 — Claude producing artificial keyword-stuffed App Store copy, dry January, banned phrases 14:30 — Korean Math Zen lifetime sales, China and the Huawei Play Store question 17:00 — Checkers game project, why Ivan skipped the Vibe Jam, mortality and doing work you respect 21:00 — Counter-Strike, Metallica, gameplay over graphics 24:00 — Why Ivan won't do TikTok dance content for Sober Tracker 27:00 — Don't bet everything on the dream, the guy doing 50K/mo on paid mobile ads 30:00 — Marketing has to feel natural, writers should write not dance 32:30 — Philipp joins with bad news from a friend's accident, mortality framing for product choices 35:00 — Lightbulb Learning rebuild using Claude design output 41:00 — Math Zen seasonal traffic spike during exam season 43:30 — Claude 4.7 telling Philipp to take a rest mid-build 47:00 — The anecdotal 4.6 quality downgrade since 4.7 launched 51:00 — Ivan's system prompt fix for the emotional output 54:00 — Memory limits, Claude forgetting which tools it has access to 58:00 — Should Philipp move from Cursor to Claude Code 1:00:30 — Vibe coding debate, security, data migrations, Vaclav's tier 1 bank perspective 1:07:00 — Ivan's coder identity crisis, who am I if I don't write code 1:09:00 — Punch cards to natural language, just another compiler abstraction 1:13:00 — Acceptance, short goals, life is unpredictable 1:15:30 — Roll call, who burned out this week, Garmin sleep score wisdom

    1h 19m

About

Indie Dev Talks is a weekly conversation about building software as one person. Host Ivan Terekhin runs a portfolio of 10+ mobile apps and is scaling them to $10K/month before moving into game development. Recurring voices include Philipp, Vaclav, and Faruxx. Each episode brings working indie developers together to compare what actually ships revenue: pricing and paywalls, App Store optimization, RevenueCat and subscription mechanics, going from idea to release in a few days, and staying profitable with no team and no funding. Expect real numbers, honest teardowns, and the tradeoffs nobody puts in a launch thread. New episodes every week.