Wow that is an elegant way to do it. This could create a renaissance of drag n drop interactions for the web

I use

- opencode 2 as the harness it works well for pretty much any specific task

- openrouter to use deepseek flash 4.1 … insanely cheap, works amazingly well and fast

- orca as the orchestrator to manage all the work (self hosted). There’s a mobile app to orchestrate afk too

And yep, vscode to browse repos etc

Sorry mate, a lot of this is vibe coded. Thanks for taking the time to check it out though

Nice! Here’s the permalink for “Rock Music” (it’s funny how this is implemented)

https://benwinding.github.io/encarta_95_web/#/article/Rock%2...

Well since nobody is suggesting it, then I will.

Go full native for each platform. React Native and Flutter are great (I’ve used them both), but there’s several reasons not to choose them anymore

1. Complexity - you’re adding another layer between your app and the device, debugging, compilation, device quirks all become much more annoying

2. AI coding makes it easier - This works both ways, flutter and react native become easier to develop, but so does native development

3. Support is always better on native - consider any issue you have building an iOS app, now add in React issues and then React native issues and then a bunch of other framework version issues… it’s generally better support if you use the tools the platform expect you to use

4. Code reuse is not as big of a deal as you think - trying to share UI code across android, iOS, web ends up making all platforms you build for worse, as bugs and performance issues go to all platforms… if you have a simply designed app, then a native experience is always preferable by users

it's not about reusing ui code. it's about already having Node.JS as backend and Next.JS as web frontend and proper layered architecture already there alright! now for mobile going native look a bit too much :(

I gave it a shot 4 months ago. 150k GitHub stars, it must be worth trying at least right?!… turned out to be some of the buggiest and clunkiest software I’ve ever used…

The idea of “your own agent instance that you can link to anything” has merit, but it was just so annoying to configure and actually quite expensive when linked to OpenAI (a few dollars a day to do nothing) also seemed to be a never ending stream of happy-baked vibe-coded features being merged… the whole project just gave me the ick

Is "happy-baked" a typo or is that an actual term? If it was an accident, I kind of like it.

That is a typo sorry (half-baked) reminds me of a bakery!

I'm going to think of unreviewed AI written code that the AI swears works, as "happybaked" from now on.

My initial thoughts are Cursor is already generalised to build all software, and Games are just a subset of all software. So it’s probably good enough.

Also point 3 “No money in gaming” is probably the main factor. There is orders of magnitude less money in gaming, than in SAAS. Think of how much money businesses pay for even basic software.

I’d also say performance is so critical and specific to every game, that getting AI to manage & architect it is fairly difficult (slop kills performance). This is nowhere near as big of a concern for most SAAS (slop friendly)

Even with all these reasons, I still think it’s worth trying to build!

If you have an enterprise contract with them, the legal protections for the consumer is much higher… is what I’ve been told anyway

It isn't, in both cases you're protected by exactly the same legal system

It's also the same planet but finding a commonality somewhere in the chain is not the same thing as finding a lack of difference in the matter.

Same legal system, but an OpenAI enterprise contract is not the same as a ChatGPT pro subscription... The terms of service agreement would be quite different

Not sure what you are saying exactly, but

> the legal protections for the consumer is much higher

You know you give away the right to file class action law suits against Anthropic when you accept their Terms Of Use, right? (at least the Americans ones)

Is that the case with enterprise contracts as well? Can’t imagine any decent procurement / legal team accepting this.

If you have an enterprise contract with them you're still unlikely to ever know that you've been lied to unless some whistleblower at the AI company comes forward and even if you do somehow find out, it's too late. Once they have your data and have trained on it there's no taking it back. At most they'll pay out some tiny settlement that's a fraction of how much money they make in a week and they'll continue to profit from your data forever.

Pure vibe-coding isn’t that satisfying (as you’ve discovered) the problem is you’re constantly reviewing its huge output, and it probably goes wild in directions making assumptions.

It’s much more fulfilling to give the LLM guardrails by engineering the solution.

This means coming up with the high level architecture, dependencies, breaking down the system into deliverable chunks, and implementing those chunks. The LLM can help in all those steps of the process. And it will be much easier to review and keep an understanding of the code.

Personally, I only have success with coding agents if it's for filling in some design I already engineered and half-implemented (like added interfaces and method declarations for example)

I feel like this is the best way to condition the agent, I don't have much success trying to do it via the prompt only

> The difference doesn’t matter until it does, and then it’s infuriating to work out what’s going on.

I love this quote, reminds me how annoying it can be when something isn’t reproducible locally… only in production…