Yura

1.4K posts

Yura

@juramoshkov

Founder 💻 Quit a full-time job to build apps 🍏 Working on https://t.co/UVtDyMz8H4 and have built 10 iOS apps 💪 My goal is to reach $10K MRR (now $4k) 💰

Barcelona, Spain Katılım Mayıs 2023

145 Takip Edilen190 Takipçiler

Sabitlenmiş Tweet

Yura@juramoshkov·14 Kas

Friday evening: I'm publishing my first video on X, and I'm a little bit nervous 😬. I want to show you my new service: Trend Watching for the App Store - appstoretrends.xyz As the founder of several iOS apps, I spent a lot of time scrolling top charts on the App Store - sometimes more than 3 hours a day. I wanted to make this process faster. So I started this project in June and vibe coded the first version in 2 days. Immediately this tool helped me to find a tanning routine niche, and my wife and I built the Sunglow iOS app in 3 days, which got more than 45k installs and $18k in 5 months. I realized the tool saved me tons of time and helped me discover niches and trends in minutes. So I continued improving it every day and, more importantly, I became a user of my own service. I can say I built it for myself and kept improving it to make it more useful. I hope guys, some of you find it interesting. I would be happy to hear your feedback.

English

1.5K

Yura@juramoshkov·25 Nis

wow, my app is number 1 in Poland App Store. never thought about this market seriously, but it's time to know deeper 😎

English

Yura@juramoshkov·22 Nis

last month i decided to focus on one existing app and try to make the unit economics work. launched google ads. installs came almost immediately, but conversion was as bad as i expected. did a deep dive into competitors and forums to understand what users actually want. also wrote a quick script to parse app store reviews and analyze feedback, super useful to see what people like and what frustrates them. then spent ~2 weeks on a full redesign (with help from my wife/cofounder): new onboarding, cleaner UX, better positioning. launched the update - and onboarding conversion dropped from ~90% to ~50% 🥲 turned out the interactivity i added was confusing users and making the flow longer. removed it, went back to simple slides - onboarding recovered. but the important part: trial conversion increased 2.5–3x. then came the hardest part, understanding unit economics. with weekly subscriptions and renewals, it’s hard to quickly see if things are working. so i built my own dashboard: funnels (onboarding → paywall → purchase), retention, user-level logs from amplitude, and ad spend. started iterating daily: find drop-offs, fix them, ship updates. also moved from daily to weekly cohorts, much clearer picture. now getting close to break-even. lesson: it’s not about building more apps - it’s about understanding your funnel and improving it.

English

Yura@juramoshkov·31 Mar

I spent all last week working on a big app update. released it on Saturday and started running Google Ads (the main goal of the update was to improve conversion and retention). on Saturday, I saw a sharp DROP 🥲 in onboarding conversion. the main reason was that I added interactivity to the onboarding. It made the user journey longer, and around 30% of users just started tapping around the screen and ended up leaving the app. As a result, I reworked the onboarding again - conversion recovered, and the number of trials increased.

English

Yura@juramoshkov·23 Mar

I was a big fan of Cursor for a long time, but I unsubscribed from it 2 weeks ago 🫣 Why? I ran into a couple of issues: I wanted to build an agent (or a skill) that goes through Amplitude, builds funnels, and generates reports about app problems based on analytics. In the end, Cursor gathered very little data, doesn’t want to browse the internet when it doesn’t know something, and its internal tools aren’t as strong as Codex (and probably Claude, but I haven’t tried it yet). I created the following skills (I want to call them agents, but it’s kind of both true and not): 1. Product analyst - goes through Amplitude, builds funnels, and writes reports. 2. Business analyst - when I come up with a feature, I first write analytics (if the feature is big). It then generates a spec for developers (backend / frontend / iOS apps). 3. Developer skills - backend and mobile for now. Here it’s very important to define development rules, the tools you use, and some architecture guidelines so you don’t have to repeat yourself every time. In my case: using SwiftGen, a coordinator pattern for views, and many other things. But the most interesting part is the integration with Figma. Figma has MCP, Codex connects to it, and when I’m writing business specs or doing development, I just attach links to frames or pages with new designs. As a result, the UI comes out pixel-perfect in ~90% of cases, and with a code style that I actually like.

English

113

Yura@juramoshkov·19 Mar

I dropped off social media for a couple of weeks - kind of accidentally 🫣 During that time, I was carefully working on a new project, which is now waiting for Apple review 🔥 I also spent a lot of time creating content for TikTok across several accounts: my personal one and one for the new app. So far, the max is 2,500 views on my personal account and 1,000 on the app account. I also gained a lot of insights on paid ads and analytics - I’ll share more about that soon ✅

English

111

Yura@juramoshkov·8 Mar

the great concept

Andrej Karpathy@karpathy

nanochat now trains GPT-2 capability model in just 2 hours on a single 8XH100 node (down from ~3 hours 1 month ago). Getting a lot closer to ~interactive! A bunch of tuning and features (fp8) went in but the biggest difference was a switch of the dataset from FineWeb-edu to NVIDIA ClimbMix (nice work NVIDIA!). I had tried Olmo, FineWeb, DCLM which all led to regressions, ClimbMix worked really well out of the box (to the point that I am slightly suspicious about about goodharting, though reading the paper it seems ~ok). In other news, after trying a few approaches for how to set things up, I now have AI Agents iterating on nanochat automatically, so I'll just leave this running for a while, go relax a bit and enjoy the feeling of post-agi :). Visualized here as an example: 110 changes made over the last ~12 hours, bringing the validation loss so far from 0.862415 down to 0.858039 for a d12 model, at no cost to wall clock time. The agent works on a feature branch, tries out ideas, merges them when they work and iterates. Amusingly, over the last ~2 weeks I almost feel like I've iterated more on the "meta-setup" where I optimize and tune the agent flows even more than the nanochat repo directly.

English

Yura@juramoshkov·24 Şub

It's a very big pain to make an integration between Superwall and Firebase. I have one app, where I see revenue, but don't have "purchase" events on Google Analytics. It's frustrated, because you can't launch ads or if you run ads, you just can't true data. And communication with support is so slow

English

Yura@juramoshkov·22 Şub

@justddev Cool, but what kind of work did it automate ?

English

David@justddev·22 Şub

@juramoshkov I'm still experimenting with it a lot but it automates quite a lot of the manual work now

English

David@justddev·21 Şub

really curious to see how this evolves as a builder myself i added a ton of custom skills to my openclaw in the last week and already see how powerful it can be this could honestly be a game changer for people without technical knowledge (at least as long as it doesn’t start writing random stuff to disk or leaking your keys)

Oliver Henry@oliverhenry

x.com/i/article/2024…

English

1.7K

Yura@juramoshkov·20 Şub

That's why I love vibecoding. I build this tool for myself for 2 hours. Nothing fancy, just one task: collect prices for in-apps and subscriptions for different countries in apps. Why? If you have users from multiple countries, it's really important to adapt prices depends on country. It can increase conversion rate and improve loyalty. Before that I look at my competitors manually and it was so long and boring, so I decided to automate this process

English

251

Yura@juramoshkov·19 Şub

So, a couple of hours spent figuring out how install attribution from ads works on iOS. For now, I’m looking only at Google Ads + Adjust (or pure Firebase without Adjust). ‼️Add Google ODM ‼️ This is a Google tool that helps attribute users even if they have disabled IDFA. Basically, it’s Google’s own mechanism so they don’t have to rely only on Apple’s SKAN. How to add it: If you’re using Adjust, add the adjustgoogleodm framework to your Xcode project. If you’re using only Firebase, the SPM library is already downloaded - you need to import GoogleAdsOnDeviceConversion, and most importantly, add -ObjC to Other Linker Flags. Without this flag, the library won’t start. To verify in both cases, run the project and check the logs. You should see: Conversion service GoogleAdsOnDeviceConversion framework is linked. If the log says the service is not linked, then the framework isn’t working and you need to dig deeper. After that, you’ll need to release an app update for the changes to take effect.

English

117

Yura@juramoshkov·19 Şub

So yesterday I launched the campaign and immediately got 10–15 additional installs (I calculated it as current installs minus the average number over the past few days). In Adjust everything is zero. On the Google Ads side also zero. But considering I spent around $7, a $0.70 cost per install seems profitable. Though there wasn’t a single purchase. I’ve already started digging into event forwarding from Adjust to Google, and then I want to analyze sessions in Amplitude to understand why users dropped off.

Yura@juramoshkov

so here we go launched 2 campaigns for my 2 apps on Google Ads the goal is pretty simple: make ltv > cpi but can I achieve it? good question also I launch 2 different approaches: attribution through adjust for the one app and direct connection between google ads and firebase analytics

English

Yura@juramoshkov·18 Şub

English

143

Yura@juramoshkov·16 Şub

How I failed with building Openclaw Agents team. So I had a clear task: I wanted to automate reddit posts creation for appstoretrends with 3 agents. 1st agent - a Reddit Researcher. Its task is to download top-day posts from selected subreddits every day and write research notes about them 2nd agent - a Reddit Copywriter. its task is to write reddit post about appstoretrends based on research notes. 3rd agent - a Team Leader. Its task is to coordinate tasks between agents and solve problem. I used openai-codex with oauth, I have a $20 OpenAI subscription. Every agent has a cron job to launch every 15 minutes. If it hasn't active tasks(or mentions), it does nothing. They can mention each other, if they get stuck. For example when reddit researcher finishes it's job, it mention team leader and team leader creates a new task for copywriter and attach research documents to this task. Copywriter takes the task and write a post. That's it. I tried different setup, databases and different prompts. For the last time I used Convex to share context between different agents. The main problems are: 1. They burnt a lot of tokens. I started my last version on Friday(I mean I changed agent prompts in their workspaces, setup convex and scripts to work with it) and day after my codex usage quota was dead. 2. a researcher created many tasks, but finished only few of them. maybe it's because of prompt or some another reason 3. to change agent workspace files I used openclaw chat, so I sent it instruction how agent should work and it changed agent files. The result was not clear instruction for working with active tasks, convex and other ones. Sometimes it deleted working instructions after my new requests

English

Yura@juramoshkov·13 Şub

I found the solution! You need to add an app store id and your team id on Firebase side (Project -> Settings -> iOS App). Otherwise, google ads can not sync your events from Google Analytics(and firebase). 4 hours of research, guys! Setting up an attribution is most challenging thing.

Yura@juramoshkov

I'm setting up Google ads for my iOS app through firebase + google analytics(without using adjust or other MMP) What I don't understand right now: Google ads haven't received any install events from my app yet So it means I need to setup a conversion on google analytics or google ads. I have an event(that I marked as key event) that is sent only once after the install, but I can't create a conversion from the event 😕 The question is how events could be connected to conversions? I'm digging deeper...

English

Yura@juramoshkov·13 Şub

English

166

Yura@juramoshkov·13 Şub

These are installs and revenue for 5 my different apps that I launched in January over the last 7 days. You never know which niche is overcrowded or without money until you ship the app And the good thing that some of the apps have a lot of weekly subscriptions, so in the next months mrr will be higher.

English

Yura@juramoshkov·12 Şub

So I’m continuing to work on coordinating my OpenClaw multi-agent setup. Right now the agents are running and writing to Supabase, but I realized Supabase isn’t ideal for my use case since it doesn’t support array types out of the box. So I’m going to try Convex, as @pbteja1998 recommended. It’s not that easy or obvious (at least for me) how to properly coordinate multiple agents. The best tip I’ve found so far: open a session with the agent that isn’t behaving as expected and give it clear instructions to rewrite its core files (like agents or soul). After that, it usually starts working correctly.

English

Yura@juramoshkov·12 Şub

I’ve been waiting for approval for 7 days, and it’s still in review. Is that normal? Apple, sorry for hating on you, you have an amazing review process 😂

Yura@juramoshkov

My first android app was sent to the review 🔥

English

Yura@juramoshkov·12 Şub

@mariosaputra how long did you wait for the approve? I've been waiting for 7 days and no results

English

Mario Saputra@mariosaputra·11 Şub

Android is doing pretty well this month, so I might do it a little bit. I also started getting comfortable with the console, which I used to find confusing at first as an Apple Store Connect user.

English

2.6K

Keşfet

@justddev @pbteja1998 @elonmusk @BarackObama @taylorswift13 @cristiano @BillGates @NASA