rebuilt my Budapest time travel browser game again, this time with Opus 5.5. It's nicer and nicer. 👉 three js 👉 one shot 👉 opus 5.5 goal is to test the new ai models playable here: bp2.mestertomi.hu/

📣 Értesítés áremelésről! 📣 Szeptember 18-án 23:59:59 után a Data Klub havidíja az új regisztrálóknak 9990 Ft-ról 14 990 Ft-ra nő. Még fontosabb: ha az áremelés ELŐTT regisztrálsz, akkor örökre „befagyasztod magadnak” a 9990 Ft-os havidíjat. Amíg tag maradsz, neked ennyi marad (igen, 3 év múlva is). MI A DATA KLUB? Magyarország talán legnagyobb (több száz fős) data science közössége és képzési platformja. Segítünk tudással, kapcsolati hálóval, portfólióépítési lehetőséggel. KINEK VALÓ? Akik belátható időn belül szeretnének lépegetni a karrierjükben és nincs végtelen idejük. MIT TARTALMAZ? 50+ előadást, 8 tanulási utat, letölthető kódbázist (és pár további extrát, lásd az oldalon), random networköt és más kapcsolatépítési funkciókat, 5 portfólióprojektet és rengeteget szeretetet. Tartalomjegyzék első kommentben. 💬 „Hát egyelőre nem tudom” 🫡 Nem is kell, mert a regisztrációd a „14 nap 499 Ft-ért” próbával indul. Mindenbe belenézhetsz, hogy megalapozottan tudj dönteni, szeretnél-e tag lenni. (Ha korábban tag voltál és egy időre lemondtad: igen, te is „befagyaszthatod” a havidíjat, de csak akkor, ha szept. 18-ig regisztrálsz.) dataklub.hu/

Eh, Claude Code... -.-

The one sentence that's single-handedly eaten up most of my time in 2026 “Just one more prompt and I’m done…”

Will AI ever understand this?

What are you working in right now?

Second best thing in world: Copy-pasting user feedback directly into Claude Code and see how the product practically improves itself.

Btw, how do you make people here on X believe you’re real (assuming you actually are)? There’s so much AI slop that whenever I post, I genuinely wonder whether my very non-AI post somehow sounds AI-generated. And if that’s a valid worry, is it still worth to post here at all?

So now that you can build almost anything, the most undervalued entrepreneurial skill is the ability to say NO to almost every new idea.

I just reserved my Cloudflare Wallet tag: datascience.cloudflare.pay. Reserve yours now at cloudflare.pay

While we’re worrying that everyone is vibe coding now, let’s remember that, based on a rough estimate using public data Of OpenAI’s total user base: >> only ~6.6% are paying users (~93% only have access to the free models) >> only ~0.56% use Codex (~99.4% are not vibe coding) The ratios are probably similar for Claude and Claude Code. While you’re building your next app or sending GPT Sol5.6 off to do 1.5 hours of research, the overwhelming majority of humanity is still wondering why this free AI thing gives such crappy answers. And why companies are spending so much money developing it. (On top of that, less than ~50% of the world’s population has tried at least one AI tool, and around ~25% still does not have internet access.) There’s no big lesson here—just a reminder that if you’re reading this post, you’re probably in the same bubble I am. 🙃

Well, it’s official: my first iOS app is now live and publicly available on the App Store. I don’t have any major business goals for it - it’s a free app. I mainly wanted to see whether, with today’s technology, it’s possible to build something like this without any previous iOS development experience. And yes, it is. The challenges are still the same: marketing, UX, research, pricing, business knowledge, and so on. But the coding part (assuming you already have a solid, broad understanding of IT, software development, and the fundamentals of web development) is much less of a problem now. And for exactly the same reason, it’s also much less of a competitive advantage. That’s interesting because while I was building my app (few weeks), the number of competitors went from ~10 to ~20. So competition grows FAST. And that brings us right back to the point above: from now on, marketing is the real challenge. Anyway, I’ll keep improving the app. Hopefully, some climbers will find it useful. But the bigger opportunity I see is using this knowledge and the next few weeks to build a mobile app for my already working businesses, making them easier for current users and customers to access services on mobile that are already available today.

If I start a new business / build a new product today, should it be b2c or b2b? Why?

Lately, it’s become trendy to have Claude Code with Opus 5 “one-shot” an entire computer game… So I thought I’d try it too and see what it could come up with. I have pretty mixed feelings about the result: ✅ It’s genuinely impressive that a single prompt can produce a working beta in about 1.5 hours. My total time investment: roughly three minutes. ✅ The engine is quite good, and the model understands the game’s basic concept. I tried to create a 3D “game” in which you can explore Budapest’s District V across different historical eras. ✅ With practically zero coding, you get a functional foundation that could be developed further. BUT: ❌ In its current form, it’s the same kind of AI slop as when someone one-shots an X post or image and immediately dumps it onto the internet. The creator is thrilled because it was made so quickly -- and because they’ve finally managed to post something coherent and vaguely meaningful. But from the consumer’s perspective, the experience is much less satisfying. It lacks the care and love that make a product feel human and genuinely enjoyable. ❌ Compared with the “one-shot” games I’ve seen other people share on social media, this is nowhere near that level. It has major flaws, needs a huge number of additional features, and doesn’t look particularly good either. I also realized that the polished, functional “one-shot” games you see online are usually either copies of existing, working games or were created with heavily refined Claude skills and harnesses running behind the scenes. So they aren’t really one-shot projects after all. ❌ Making this game genuinely enjoyable, usable, and potentially commercially viable would still require an enormous amount of work. Even if coding is no longer the hardest part, there’s still product development, research, testing, marketing, and so on. In other words: all the parts that were difficult before are still difficult. MY TAKEAWAY: One-shot game development is the same fairy tale we hear everywhere else that AI will do the work for you, only better and faster. AI is still just a tool—a multiplier applied to your knowledge, persistence, and work ethic. ✌️ Whether that’s good news or bad news is up to you. 😉 P.S. From now on, I’m going to have every new Claude model build the same game. It should be a great benchmark for comparing their progress. P.P.S. You can try the game yourself: bp.mestertomi.hu/

Let’s see what Opus 5 Ultracode can do. I’m going to try to one-shot a computer game. 😁 I’m giving it a 20% chance of success. Perfect entertainment for a summer Sunday evening.

Now that the X newsfeed is better, is it worth coming back here?