I'm Evgenii, an indie iOS dev.
I build apps with SwiftUI and AI agents, and I share the process: build logs, SwiftUI tips from the trenches, and honest numbers.
Lately: making layouts work on iPhone Duo.
Follow along ๐
Cloud agents are great when you're away from your laptop.
I'm on a train right now and didn't want to take MacBook out. So I launched a cloud agent in Cursor from my phone to check the logs of the update I shipped before leaving.
Good use cases for me now:
โข checking logs after a release
โข small fixes and quick investigations
โข tasks that don't need local setup
I think I'm slowly starting to feel how convenient they are.
@dxiaolong The logs I check are from my iOS apps, and unfortunately working on them needs my local MacBook. So I either reassure myself with logs that everything's fine, or wait until I can fix it ๐
@MarioSaputra B. While reworking my apps for iPhone Duo, I first tried to keep close buttons at the top. But then I thought Duo users will quickly learn from system apps and their buttons, so it'll probably feel more natural to see the close button on the side in our indie apps too.
@max_mtwx Written with AI, using the words people there actually search for, not just translated. Not native text from a human, but much better than a plain translator.
Spent today setting up SEO with Grok Bot for my app site.
1. Self-hosted an open-source SEO tool https://t.co/AXH2BpczXX on my own server and connected Grok Bot to it. Also added DataForSeo with 50$ balance.
2. Ran an SEO and GEO audit: what people already search for, where the site is close to the top, and where the gaps are, including visibility in AI search.
3. Built a keyword map for my two main apps, focused on the US and Germany.
4. Made a weekly plan: 1โ2 tasks a day, each done by an agent as a PR, and I review and ship it myself.
The first three improvements to the German pages are already live. From now on it's a loop: weekly analysis, then a new task plan.
Hope this brings results. I'll share how it goes.
I finally decided to clean up and organize my GitHub profile. I'm removing old, unnecessary repos, sorting out the rest, and turning it into a proper public developer profile.
Why it matters to me:
1. It's my tech business card. It's often the first place employers, other devs, or potential partners would look.
2. It can help build my name, so people can recognize me and what I work on.
3. It can open doors: getting noticed by recruiters, new collaborations, finding contributors.
4. It helps my open source. Clear repos make it easier for others to use them and contribute.
5. It shows how I work. Not just what I build, but how I structure, document, and maintain it.
I'll be sharing my open-source projects here, so I want the profile to be ready when people check it out.
I'm flying to Japan next week ๐ฎ
Could you recommend some shops worth visiting for Nintendo games and gaming merch? I'll be spending most of my time in Tokyo and Osaka.
Happy to check out any other recommendations too!
Early notes on my AI coding setup (will update as I learn more).
After the Claude Code ban I went back to ChatGPT. A couple of days in I remembered why I left.
Test: rework my app's navigation architecture, moving sheet and fullScreenCover to the right places. The rules are in my Skill.
ChatGPT broke many of them, didn't give me clean code, and missed things the task logically required. It takes prompts literally: forget something and it won't be done. And of course the constant "You're right, I made a mistake."
Now trying Grok 4.7 in Cursor. Same task: clean code that follows my rules, and it covered cases I never wrote down.
I'm not here to crown a "best/worst model" every time I switch tools. I'm just sharing what works for my workflow right now, a few days in, and I'll keep updating this as I learn more.
Depends on the type of work. Data changes always get their own task. Even when they're part of something bigger, they go into a separate stage. For sensitive areas like data, I know exactly what will change, plan it in advance, and know what to expect. Same for everything else: work is split into stages, and I check the most important ones in git diff. And when the model keeps following my instructions fully, time after time, you can see right away in the code whether there are problems. And of course unit tests, especially for business logic.
Update after 5 days on the $30 ($15) Grok plan with Grok Bot + Cursor:
What I did with Cursor + Grok 4.7 high:
โข several large architecture and UI rewrites in my iOS app, plus smaller tasks โ code quality is good enough that I can hand off a task without checking every line again
โข a SwiftUI icon system (the one I posted about earlier)
With Grok Bot + Cursor cloud agents + Grok 4.7 high:
โข connected the bot to my app landing site: automation so I can review PRs, staging deploys, and SEO work
โข SEO audits of a few landings โ reports are still shallow, but the work took real time
โข Another bot helps with X: stats, post drafts, and nudges to keep posting
That took 91% of weekly limits in 5 days. For the cheapest plan, I think that's a strong result.
Important note: some routine work went to ChatGPT โ icon code, small component layouts, keyword translation into 30 localizations. I was careful with Grok tokens on this plan. On a $100 plan I'd probably stop thinking about limits, like I did with Claude Code.
@aitorapps None. Every task I gave Grok was completed fully โ matching my layout and logic skills, and matching the prompts. With ChatGPT I constantly found problems in the code (yes it worked, but it worked poorly).
I don't like polish work when it's the small awkward ones with no clear solution. They're fine a few at a time during the week. But when they pile up at the end of a big update, my productivity drops hard.
I'm leaving on a trip in a couple of days, so I have to clear all of this today.