Fatih İlhan

6 months ago I started Nokta Studio from Ankara 🇹🇷 Shipped: → LoopSignal (SaaS feedback boards) → ConvertCenter (utility site) → 18 Apify APIs → 3 mobile apps Total MRR: $0. The product was never the problem. Distribution was. This time: audience first, product second.

Just shipped comparely.world 🌍 Tax-aware country comparison for digital nomads & expats. 35 jurisdictions, special regimes auto-picked. Built because "cost of living" comparisons skip the part that actually matters: net after tax. Feedback welcome.

@marclou Trust is the killer. Advisors with no skin in the game will pump bad listings. Fix: tie advisor fee to acquisition outcome, not listing improvement. Pay only when a deal closes. Self-selects for serious people.

I’m exploring a weird idea for this: Turning the affiliate program into something closer to a network of startup sales advisors. Instead of only referring buyers, advisors could help founders improve their listing, suggest a better asking price, and find strategic buyers who might actually acquire the startup. There are obvious concerns around seller privacy, spam, trust, and the fact that these advisors would not be endorsed by TrustMRR because I want to work solo. But I wonder if this could help more small startups get acquired faster. Curious what you think?

@levelsio I was not expecting this to find in my city but not surprised

Check your city on my site hoodmaps.com

🇪🇸 Barcelona is particularly interesting The tourist center used to be just pickpockets targeting tourists, but now you have the highest crime rate in El Raval with drug dens/squats (narco piso), violent robberies and stabbings The problem as a tourist in a new city you'd never know that because Google Maps doesn't tell you But Hoodmaps does 😊

My astrologer wife needed a tool no one else in Turkey had built. Shipped it in a day with Claude Code. Already live, already in her Instagram story.

@thejustinwelsh Consistency is always the key, both for improving your life and body. As long as you keep showing up, results will come.

Improving your life is a lot like improving your body. It's not easy, and it's not fast. It's difficult and slow. But if you commit to it, a few years later, you'll be amazed at your progress.

This is gamechanger! I've been using Claude Code for some time remotely through Google's Remote Desktop application but now seriously consider switching back to Codex for the remote SSH connection.

8/ Wrote up the full build with code + lessons. The "scraping a scraper" detour and how I found the hidden JSON endpoint might be the most useful parts. → [dev.to/fatihbuilds/i-…] If you're job-hunting in any structured market, the scoring layer ports cleanly.

7/ First production email landed yesterday. One job. Exactly the kind I'd actually apply to. Twice a week from now on, Monday + Thursday mornings. May add LLM scoring on top (Claude API, ~$1/mo) once I have more pattern data.

1/ Built my own job filter this week. Openings in my industry scattered across 5+ sites, none know my profile. Pipeline emails me 5-10 fitting roles, twice a week. Here's how it went 👇

Monday morning routine: 7:30 — toddler wakes up 8:45 — drop-off 9:00 — office, real data work 17:30 — return to home 21:30 — Cursor, finally The 21:30 hour is the only one I own. Building for the day I own more of them.

@elie2222 @FredKSchott @mattpocockuk @DmytroKrasun 12 = atomic features. e.g. "URL routing", "currency converter", "income mode toggle". No looping, sequential. Same Claude Code session, prompts back to back. Restart only when context starts drifting.

@fatihbuilds @FredKSchott @mattpocockuk @DmytroKrasun When you say 12 feature what’s that mean? It loops itself? Also do you run it from Claude session or you keep starting a new session?

New video on Flue, Sandcastle, and doing it yourself! It covers: 1️⃣ Flue by @FredKSchott. An agent harness framework 2️⃣ Sandcastle by @mattpocockuk. A TypeScript library for orchestrating coding agents in sandboxes Thanks to @DmytroKrasun's ScreenshotOne for sponsoring!

@elie2222 @FredKSchott @mattpocockuk @DmytroKrasun 4 setup, 12 feature, 5 integration, 3 polish. Each small enough to one-shot, big enough to ship. Re: Claude Max — same dealbreaker for me. Already paying for it.

@fatihbuilds @FredKSchott @mattpocockuk @DmytroKrasun i basically go and do that at the end of the video. it's totally fine. you have full control. how do you put 24 together? that's a lot!

@tdinh_me The parallel chat is the part I keep underestimating. For Comparely I run Claude + GPT side-by-side on the same prompt — different models catch different bugs. Built it myself with raw APIs. If TypingMind does this with 12 providers at once, my "side-by-side script" just died.

TypingMind now supports 12 LLM providers natively 🥰 (you can chat with all of them in parallel at once!)

@DmytroKrasun The "screenshot looks like real browsing" bar is harder than it sounds. I built a country comparison site (Comparely) and even fetching Numbeo/cost-of-living data clean breaks for the same reasons — cookie banners, lazy loads, anti-bot.

The hardest websites to screenshot are commerce websites. Animations, sales and localization pop-ups, cookie banners, chat widgets, lazy loading... I created Shops.Gallery for myself to improve ScreenshotOne’s rendering. It also helps potential customers assess the rendering quality quickly. All screenshots are generated automatically from publicly accessible websites, without manual cleanup or intervention. I will keep improving banner blocking, rendering consistency, and rendering heuristics until screenshots look as close to the real browsing experience as possible.