Thursday, June 5, 2025

Marek's Dev Diary: June 5, 2025

What is this

Every Thursday, I will share a dev diary about what we've been working on over the past few weeks. I'll focus on the interesting challenges and solutions that I encountered. I won't be able to cover everything, but I'll share what caught my interest.

Why am I doing it

I want to bring our community along on this journey, and I simply love writing about things I'm passionate about! This is my unfiltered dev journal, so please keep in mind that what I write here are my thoughts and will be outdated by the time you read this, as so many things change quickly. Any plans I mention aren't set in stone and everything is subject to change. Also, if you don't like spoilers, then don't read this.

A few months ago, I restructured my schedule to alternate weekly between Space Engineers 2 and AI People. This approach lets me maintain a deep focus on each project. I'm really someone who needs to dig deep into something and work on it until it's done, rather than switching context every few hours.

AI People

This week's focus was AI People, specifically exploring AI-assisted programming (what we call "vibe coding") to accelerate development of our AI NPCs. However, instead of diving into planned features, we spent the week refining our methodology.

Our experience with Cursor + Opus 4 and Gemini Pro 2.5 revealed a frustrating pattern. Initial progress feels promising, but then you hit a wall: request a change, the AI modifies code, you skip review and test directly, discover new bugs or no improvement, repeat. Hours later, you realize you're going nowhere.

The core issue? Current AI agents approach software engineering fundamentally differently than experienced developers.

How AI agents work today: You describe a feature → AI reasons briefly → finds files → implements changes → declares completion.

How expert developers actually work:
  1. Fully understand requirements and context
  2. Study relevant code thoroughly (nothing more, nothing less)
  3. Break complex changes into testable chunks
  4. Implement with constant awareness of ripple effects
  5. Review for edge cases and unintended consequences
  6. Update all affected elements - comments, references, documentation, architecture diagrams
  7. Write comprehensive tests and iterate based on results
  8. Access running systems for real-time debugging and observation
Current tools can't replicate this workflow. They also lack game runtime access, can't insert diagnostic traces, and miss the holistic view that makes great code.

This realization led us to explore building our own SWE agent. We're studying Claude Code, which implements some of these concepts plus additional capabilities like sub-agents.

Key insights from this exploration:

LLMs feel superhuman in their domains. Yes, they still have gaps and can only handle minute-long tasks rather than day-long projects, but within their scope? Opus 4 writes a complete Tetris game in seconds - a task that would take me days. The bottleneck isn't intelligence; it's the scaffolding.

When AI programming fails, it's rarely the model's fault. It's inadequate tooling around it. I'm convinced 2025 will bring revolutionary improvements: correct SWE loop, specialized agents for exploration, coding, testing, reviewing, evaluating, validating, etc; sophisticated code search and indexing, intelligent test automation, multimodal feedback loops. Imagine Gemini analyzing gameplay video to fix bugs autonomously.

We're validating our approach on smaller codebases and design documents. Design docs are particularly revealing - text changes are far easier to track than code modifications, exposing flawed agent behavior immediately.

Case in point: I asked Cursor to reformat log specifications in our design doc. It updated one section, missed another, left duplicates, never reviewed its work. In a text document, these mistakes jump out instantly. In code, they'd hide among thousands of lines. Classic junior developer behavior - making changes without verifying impact.

How about costs? Sure, spending $100 on tokens in a day might seem expensive. But if the AI delivers in one day what would take me two weeks? That $100 is cheap. Plus you're iterating in hours instead of weeks. Clear win.

We're not quite there yet, but I'm confident this year will bring the breakthrough. Once we crack this, we'll accelerate AI People development dramatically, running parallel experiments and iterating at unprecedented speed.

Space Engineers 2

Given my AI agent focus this week, my SE2 time went into writing the SE2 Vision document - a comprehensive guide defining requirements, constraints, and KPIs for the team.

Our north star: Make SE2 mainstream while delivering 10x improvements across every dimension - art, code, design, quality, performance.


The key insight: SE2 will match or exceed SE1's complexity, but we're wrapping it in an accessibility layer. New players start with intuitive, manageable systems. Complexity reveals itself progressively as skills develop.

We're also prioritizing engaging gameplay loops and meaningful progression. The complexity remains - it just becomes fun to master rather than overwhelming to encounter.

Thursday, May 29, 2025

Marek's Dev Diary: May 29, 2025

What is this

Every Thursday, I will share a dev diary about what we've been working on over the past few weeks. I'll focus on the interesting challenges and solutions that I encountered. I won't be able to cover everything, but I'll share what caught my interest.

Why am I doing it

I want to bring our community along on this journey, and I simply love writing about things I'm passionate about! This is my unfiltered dev journal, so please keep in mind that what I write here are my thoughts and will be outdated by the time you read this, as so many things change quickly. Any plans I mention aren't set in stone and everything is subject to change. Also, if you don't like spoilers, then don't read this.


Space Engineers 2

This week's work was mainly about survival gameplay - discussion with the designers and playtesting what already works.

Resource Gathering


Drilling and Ore Detection - The basics work! Now we need to polish it: better animations, particle effects, and voxel deformation (how terrain changes when you drill).


Building and Repairing


Projection Building - This is different from SE1. You place blocks as semi-transparent holograms first - just projections without physics that you can walk through. Then you weld them block by block to make them real.

Why we went this direction: You see what you're building before spending resources, you can place entire Workshop blueprints as projections, and we can bring creative mode features like undo/redo into survival - since you're just manipulating projections, not actual blocks!

Build Planner Queue - Every projection you place gets added to a build queue. This queue can be sent to production blocks to craft the components you need. It's all connected!

Repair HUD - A new concept we're working on. The game compares your current grid against the original blueprint (or a saved snapshot), shows missing blocks as projections, and displays them on your HUD (similar to GPS markers). Should make repairs much more intuitive.


Testing the Loop


I've been playtesting this myself. Placed cockpit projections around me - some on ground, some in air (supporting truss blocks are added to keep the block while welding it). Then I mined iron and silicon and welded them using backpack building. Very satisfying to see your build take shape!

Note: Backpack building (direct crafting) is only for early game basic blocks. Later stages require manufactured components and/or block wrappers (which I may talk about next time).

That's it for this week!

Thursday, May 22, 2025

Marek's Dev Diary: May 22, 2025

What is this

Every Thursday, I will share a dev diary about what we've been working on over the past few weeks. I'll focus on the interesting challenges and solutions that I encountered. I won't be able to cover everything, but I'll share what caught my interest.

Why am I doing it

I want to bring our community along on this journey, and I simply love writing about things I'm passionate about! This is my unfiltered dev journal, so please keep in mind that what I write here are my thoughts and will be outdated by the time you read this, as so many things change quickly. Any plans I mention aren't set in stone and everything is subject to change. Also, if you don't like spoilers, then don't read this.

AI People

This week, I’m focusing on the AI People project. We’re returning to the drawing boards, redesigning both the gameplay and the technology behind AI NPCs. Our goal is to make the NPCs more goal-oriented, with distinct personalities, capable of solving environmental and social challenges. Their interactions—with each other and the player—should lead to emergent stories and butterfly-effect-like chain reactions spanning from hours to weeks.

I decided to work on the prototype personally, so a colleague and I have been vibe coding all day. Vibe coding is when you instruct a coding agent (like Cursor, OpenAI Codex, or Jules) to autonomously make changes to the codebase. It’s going well—it feels like we can achieve in a single day what used to take weeks. Iterations and feedback loops are much faster, which means we can prototype at an incredible pace.

Today, we also implemented automated testing. After the agent completes code changes, it runs the tests and, based on the results (e.g., if some tests fail), it starts fixing the issues itself. This new workflow is a game-changer because we no longer have to tediously copy-paste errors from logs. The tests were written by the agent as well, saving us even more time.

I’m already looking forward to the next five years when this workflow will be even more simplified and faster. Imagine just asking for something like "add grounded NPC actions like goto(), attack(), harvest()—planned and executed in sequence," and the agent makes the changes, tests them, and delivers an interactive game in a second.

Just see how fast the new diffusion LLMs are: https://x.com/InceptionAILabs/status/1894847919624462794

Space Engineers

Every two weeks, we hold an internal team biweekly presentation where individual teams share their progress. This week’s session was particularly exciting!

SE1 Team


The SE1 team shared their progress on the new survival mode and introduced some new blocks, which are shaping up nicely.

SE2 Team


Most of the SE2 team is working on VS2, focusing on planets and survival mechanics. Here’s some of what they showcased:
  • Procedural World Generator: This now spawns asteroid fields and rings with a realistic distribution.
  • Clouds: Already in-game and visible in many screenshots (though they’re still programmer-art clouds; artists will refine them later).
  • Planet Flora Generator: We now have grass and trees on planets! Currently, it’s just one type of tree everywhere, but artists are working on a flora library that includes dozens of trees, bushes, fynbos, plants, and even underwater flora.
  • Backpack Building and Welding: This feature is nearly complete, which is exciting because we’ll soon be able to playtest the full survival game loop and start iterating on improvements.

Water Team



The water team gave what might be their best presentation yet. They showcased several impressive improvements:
  • Enhanced performance.
  • New marching cubes for the water surface.
  • New particle effects for dynamic water (think waterfalls and splashes from your ship).
  • Improved resolution of the water simulation, plus many other upgrades.

Art and Visuals


The artists made another pass on improving voxel materials for the planets, and the results are stunning—screenshots are starting to look photorealistic. On top of that, our planets are fully volumetric and destructible, allowing players to dig tunnels and add materials.


The new particle effects on the Hydrogen Thrusters look fierce, and new character models are in development, which are shaping up to be very cool.


It’s been an exciting week, and I’m looking forward to seeing how these features evolve in the coming sprints!