Peter Šutarík

I delegated invoicing to AI this morning. Built an MCP server connecting Claude Code to @iDoklad (local invoicing system). Now I just say "invoice client X for 5 hours of consulting" and it finds the client, creates the invoice with correct VAT, generates a PDF, and emails it.

@forgebitz Because majority of people lack taste as well

designers are absolutely not cooked llms don't have any taste at all but i can see how so people think they do

Check out Sonar - ASO tool for keyword research for both App Store and Google Play. trysonar.app

I made (my first) $121.60 in April 2026. 📡 Sonar - $114.60 🍳 ReelMeal - $7 Late March, I launched both Sonar (ASO tool for indie devs with multiple apps) and ReelMeal (iOS app to save recipes from TikTok/Instagram). In April, I got my first paying customers on both. Small numbers, but it's the first time strangers have paid me, so it feels special. Goal for May: Double the numbers!

Sonar is live on OpenLaunch! 🎉 Check it out!

Every streak has to end At the start of 2026, I set myself a resolution: ship something (however small) every day of the year. It went pretty well, until it didn't 😅 My streak ended on a trip to Amsterdam for Measurecamp. So what have I shipped so far in 2026? - Finally finished my personal site (petersutarik.com) - Built, scrapped, and rebuilt my personal AI assistant Arthur, who helps me with my day-to-day workflows - Built internal Claude Code skills for invoicing & SEO research - Built and launched a small SaaS that earned its first dollar: trysonar.app - Built and launched an iOS app that got its first paying user this month No regrets though. Back to shipping :)

@MarekLecian Stále to vyzerá ako AI slop :D

Nový ChatGPT Images 2.0 ... prostě WOW. openai.com/index/introduc… Tohle je opravdu next level.

@alexcooldev Even worse then

@PeterSutarik No, that’s a bucket for seafood.

I just bought this, hope OpenClaw is still solid compared to Claude Cowork 🙄

@anulagarwal Get ranked. trysonar.app

pitch me your SaaS in 2 words let's see how creative everybody is

@thekitze Just open the bottle man 😭

using gpt 5.4 for openclaw twitter.com/FalconryFinanc…

@anulagarwal I mean, you might have edge cases where your app won't be localized to that language, but then just default to English imo (or show this screen)

@PeterSutarik ok i actually did not think of that lol 🫡

Pro tip for building mobile apps to increase conversion: Your first onboarding screen should always be a language selection screen. You don't want the user to follow the onboarding in a language they may not be comfortable in.

@robj3d3 These are so obvious for non-native speakers (at least they were for me 😅)

Just now I learned “afternoon” is literally “after” “noon” i.e. after 12pm Nothing will ever be the same again.

@anulagarwal trysonar.app

what are some good ASO tools for Android/Play Store?

@marclou Just support proper data layer maybe?

DataFast Goals now shows the custom parameters you send. If you track an "initiate_checkout" goal, you can pass "planType" as a custom parameter. In your dashboard, you'll now see how many people initiated a checkout, and which plan they chose.

@romankoch_ Both popularity and difficulty are just estimates for the most part, so it really depends what variables the tool uses to compute that

Yesterday i've compared ASO keywords on two different platforms, and in 50% the cases the difference huge in popularity and difficulty. The entire ASO feels like a gambling now, need to compare one or two more to get more clear picture.

@arvidkahl These charts not starting at 0 should be criminal

Still impressive, but sure looks different when zoomed out.

In my Claude Code's CLAUDE .md file, I put stuff like "every change needs tests, don't let me skip them", or "update the docs as part of the work, not after". Feels much better when adding/editing features.

@AlexBelogubov Aren’t you afraid using Sonnet for email stuff? Might be prone to prompt injection, or?

I run 15+ AI cron jobs daily across 3 models. Here's real cost data from 15 days of production: 𝗧𝗶𝗲𝗿 𝟭: Claude Opus 4.6 → ~$125/mo Main AI coordinator + content writing. Complex multi-step reasoning, creative decisions, agent orchestration. 𝗧𝗶𝗲𝗿 𝟮: Claude Sonnet 4.5 → ~$55/mo X digest, Gmail digest, CEO standup, ops audits, triage. Analytical tasks — needs context understanding but not creativity. 𝗧𝗶𝗲𝗿 𝟯: Codex 5.3 → ~$18/mo Support inbox sweep, watchdog, budget guard. Routine tasks with clear instructions. 𝗧𝗶𝗲𝗿 𝟰: MiniMax M2.1 → ~$0.50/mo Infra health check every 5 min. One command, one check. Cheapest job. Total: ~$200/mo If everything ran on Opus: ~$800+ 3 lessons from 15 days of real usage: 𝟭. Cheap models break on complex prompts silently. Codex generated reports but skipped the API call to post them. 24h of missing data before I caught it. Fix: explicit curl examples + CRITICAL warnings in the prompt. 𝟮. Sonnet is the sweet spot for analytics. Switched 6 crons from Codex → Sonnet. CEO standup went from a useless one-liner to a full 6-section report. Cost +$40/mo, quality 10x. 𝟯. Model routing isn't set-and-forget. Changed models 4 times in 5 days. Each switch needs prompt adjustments. The cheaper the model — the more explicit your instructions must be. The rule: match model intelligence to task complexity. Don't pay for creativity when you need a checkbox.