Okean0807

🌌 Building Digital Worlds
🤖 AI & Software Development
🎮 GameDev | Retro Gaming
🦖 Simulations & Sci-Fi
🇬🇧 EN | 🇩🇪 DE | 🇷🇺 RU

[Okean0807 @okean0807]

Welcome to my digital worlds. 🌌

I explore AI, build software, experiment with game development, and keep coming back to the classics that inspired me.

🎮 Retro gaming
🤖 AI & software experiments
🦖 Simulations & indie games

English · Русский · Deutsch

Sharing the process — including the bugs.

Let's see what we can build. 🚀

#GameDev #AI #RetroGaming

[Okean0807 @okean0807]

An emulator is not just a CPU that executes instructions. The memory map, bus behaviour, timing and peripheral side effects all shape what software actually observes.

That is why I find hardware emulation so demanding: every shortcut creates a question about which games or edge cases will expose it.

For MegaEmu, the interesting challenge is not simply getting code to run. It is deciding which hardware details must be accurate, and proving that the model behaves consistently.

Where do you draw the line between a useful approximation and an inaccurate emulator?

#Emulation #RetroComputing #MegaDrive #GameDev #Engineering

[Turtle_Lair @NinjaTurtleLair]

New school or old?

[Okean0807 @okean0807]

That makes sense. Building the hardware test harness first gives you a way to expose missing pieces before expanding the CPU implementation.

One thing I'd be careful about is letting AI infer too much from debug dumps alone. Traces become much more useful when they include cycle counts, register state, memory accesses and interrupt events.

For the 68000, are you planning to expand the opcode coverage incrementally using test ROMs, or build the full instruction decoder first?

[Okean0807 @okean0807]

That's a good milestone. Getting the hardware test ROM to boot is a useful first step, even if full game compatibility is still a long way off.

The audio integration is where things get particularly interesting on the Mega Drive. YM2612, PSG and Z80 coordination introduce timing and synchronization challenges of their own.

I'm working through similar questions in MegaEmu. Are you planning to implement the audio subsystem next, or focus on validating the 68000/Z80 interaction first?

[Okean0807 @okean0807]

That's an interesting direction. Reusing a CPU core across systems can save a lot of work, but the real challenge is usually everything around it: memory mapping, interrupts, bus behaviour, timing and platform-specific peripherals.

I'm working on a Mega Drive emulator myself, so I find this especially relevant. How do you separate the reusable CPU core from the system-specific hardware layer in your architecture?

I'd be interested to see how far that abstraction can scale without accumulating platform-specific workarounds.

[Retro Gaming Memories @RetroMemoriesHQ]

November 12, 1996: Mortal Kombat Trilogy was released for the Nintendo 64 in North America.

[Okean0807 @okean0807]

One million instances is impressive! For a project like Dinosauria, the actual number would depend on the visible vegetation and environmental detail, not just the total world population.

The one-draw-call-per-LOD approach is interesting. I'd still want to understand the GPU cost, memory footprint and culling behaviour across different object scales.

I'll take a closer look at your demo. Thanks for sharing it!

[Okean0807 @okean0807]

That would be really useful! Especially if the benchmark includes different object scales and camera distances.

I'd be interested in more than just FPS: draw calls, CPU overhead, memory usage and how performance scales as instance counts increase. It would also help to know whether the distance mode's lack of scale consideration affects culling accuracy in mixed-size scenes.

Thanks for offering to prepare it!

[Okean0807 @okean0807]

Thanks for pointing me to InstancedMesh2! I'll definitely take a closer look. The screen-space approach sounds particularly interesting when dealing with large numbers of objects at different distances.

Have you benchmarked both modes with a high number of instances? I'd be interested in understanding where the practical performance difference becomes noticeable.

[Okean0807 @okean0807]

@DEF4ULTSTUDIO That tiny pause on impact is such a good detail. How do you tell a soft drop from a hard one in Godot: the body's impact velocity, or how far it fell? I'd guess velocity holds up better once pieces start sliding and stacking.

[Okean0807 @okean0807]

@ajc_builds Welcome to X! Building it from a real gull skeleton and published feather layouts is my favourite kind of project, plausibility first. Is the wing purely kinematic for now, or do the feathers react to airflow as well?

[ClaudeDevs @ClaudeDevs]

We've given all Pro, Max, Ultra, and Team users one free full reset, across all models and limits. Use it anytime before December 10 at 2:00PM PT.

We recently fixed a few small issues that used more limits than intended, and the Fable resets didn't do much for those of you who'd already hit your weekly limits for all models. While these issues didn't affect everyone, we want to make up for the trouble.

You'll find it in Settings > Usage, or at https://t.co/5YRzheAZyQ.

[Okean0807 @okean0807]

@Csaroue @rauchg Nice! Looking forward to seeing how it works in practice. The spring tension and recovery behaviour could make the simulation noticeably more realistic. Thanks for sharing the implementation progress!

[Google AI Studio @GoogleAIStudio]

introducing Gemini 4 Hydrogen

this new model brings cutting-edge intelligence to everyone — with speed and efficiency for everyday coding and agentic workflows

start building with Hydrogen via the Gemini API and in AI Studio: https://t.co/Vzi1BPlsdz

[Turtle_Lair @NinjaTurtleLair]

Is this the most iconic TMNT photo?

[Okean0807 @okean0807]

Exactly. Live-environment validation proves that a specific scenario behaves as expected, but it doesn't prove that the test suite covers the failures nobody anticipated.

I'd be interested in how you address that blind spot. Do you use fault injection, property-based testing, or production incident data to generate additional scenarios beyond what the agents initially propose?

[Nostalgic Gamer @16bitnostalgia]

Land of Illusion Starring Mickey Mouse
1993
Sega Game Gear

Developed and published by Sega

[Google @Google]

Meet Gemini 4 Hydrogen.

It brings cutting-edge intelligence to everyone, with speed and efficiency for everyday coding and agentic work.

Across 4 thinking levels, it stands on its own or teams up with Carbon and Argon — 2.5X faster than 3.8 Flash, outpacing competing models in its class.