mid_gen's Launch a Game Studio Thread

Thought I'd try something a bit different on the channel today, so posted some DLSS5 waffle. It's not done much, the algo seems to be struggling to find the audience, as it's clearly not the same as my 'indie dev' viewers.....interesting! Have to try these things though, took me quite a while on my climbing channel to home in on the format that clicks with viewers.

Do have a vlog recorded too, will prob get that posted tomorrow if I get time to capture some gameplay.
 
The DLSS5 stuff has been talked about on every single gaming channel/podcast ever since the Nvidia conference.

Everyone has probably got sick of hearing about it now, I know I am :D So any talk about it now will probably get lost in the noise on Youtube.
 
The DLSS5 stuff has been talked about on every single gaming channel/podcast ever since the Nvidia conference.

Everyone has probably got sick of hearing about it now, I know I am :D So any talk about it now will probably get lost in the noise on Youtube.
There’s various ways content on the tube perform.

The DLSS one has a poor CTR, only 2% or so. And you’re right, with zeitgeist stuff you need to be quick and get in there before there’s thousands of other videos trying to cover it.

BUT

People that do start watching are generally watching all of it, much more so than the vlogs. So the watch hours is relatively good still….and likely to be a bit more on an evergreen, than the vlogs.

I’m going to do a ‘3 months in’ milestone video for the end of March. Probably change up the format a bit.
 
Last edited:
I am just stupid busy at the moment....have a couple days away with the missus next week which will be a much needed bit of downtime.....all good though, just tiring.

Managing to chip away at the game....lots of little things.

Today I added some APIs and extended the game's MCP server so I can now ask claude to stage tests for me. ie. I can just say "create a test with every grid populated on both sides with a mix of ships"....and it just goes through the MCP and python API to set the game up for me, in this case, this glorious scene.



I really need some better explosions.....but there's not really any decent zero-G ship explosions on the unreal marketplace. Might have to set aside a few hours to make my own. Need some reactor meltdown awesomeness...and look into the slicing tools to have the ships break up as they die.
 
Last edited:
Lots of nice little updates in recent days, but I’m going to save it all for the big ‘3 months in’ video I’m going to do next week.

Going to put a bit more effort in, shoot some b-roll, have some backing music, have a bit more in depth run through everything I’ve done so far. Generally up my game on that front.

The DLSS chat as I suspected is doing steady numbers now….difference audience so good to have the extra eyes on it, although predictably has some more spicy comments :)

First proper session with the design agency on Monday to get into the process.

Then a few days off, much needed!
 
Phew, finally an update. Hopefully through the manic phase of house move process and onto the tedious solicitor chasing one.


Nice update. Coming together nicely. Amazing really considering full time job and house move, plus youtube.

Do you intend to go down the steam early access route?
 
Nice update. Coming together nicely. Amazing really considering full time job and house move, plus youtube.
Cheers, has been a stressful few months, but feels like I'm working towards something really positive and it'll all be worth it though!
Do you intend to go down the steam early access route?
I've been thinking a lot about funding. My first thought on Early Access is no, as I'm just not a fan of it personally. I'd rather launch a finished title, do a few maintenance releases, and if successful then perhaps some content updates.

I think my 'Golden Path' funding model for the first 12 months is roughly:

Continue with my freelance work part-time. This is more than enough to pay all my living costs and have enough to pay freelancers for the creative work, and some marketing spend.

Complete this first title under my own steam (hoho), no external funding. Get it launched on Steam with a modestly successful launch. Say I price at £7.99 and shift 20k units, gives me a little over £100k after Valve's cut.

That would set me up to go full-time from next year.

Just kinda thinking out loud. Obviously I could fail miserably with the launch....but this first title I've deliberately chosen something pretty straightforward with mass-market appeal (spaceships, pew pew, can't go too far wrong).

I will talk to some publishers in the coming months....while I'd prefer to not have to give away revenues, there's no doubt a good publisher can bring a lot to the table in terms of knowledge, marketing, audience etc.

As I mentioned in the video, the freelance work I'm doing is for a friend's startup and stopping that work doesn't just impact me (as much as I'd like to go full-time on the dev work, and have the funds to do so), it impacts them, as he wouldn't be getting someone with my skills and experience for the rate he's paying otherwise! It's an arrangement that works well for both of our companies for the short term.
 
Well happy with that video, it's doing gangbusters on the algorithm. Although I did have my first troll that started posting abuse as fast as I removed it (which one of you was it :P). Tempted to write a Claude agent that'll just filter stuff like that out and remove it without me having to read it. Although I did think 'washed up midlife crisis loser role-playing at game dev' was kinda funny :)
 
Finally, first week I have two full working days to work on the game :)

I could do with finding a small handful of testers....not for gameplay testing just yet (although you're welcome to mess around with what's there), but I want to sort out a pipeline for packaging builds and making them available, also be useful if it's people with different hardware (my desktop is 265k + 5080), so AMD/Intel GPU just so I get an early heads up if there's any major compatibility issues. Let me know if you'd be interested.

On the subject of testing....I've just created a bridge module and enhanced the game's MCP server. This gives the MCP tools to fetch the current gamestate and map into the inputs.....basically lets anything connecting to the MCP play the game when it's in the editor.

So right now, Claude Code is playing the game for me in the background. These are the sorts of things AI can do that save so much time and effort that would normally go into either manual testing or arduous orchestration efforts. Now I just say "Claude, play the game for half an hour". Next I'll add a frametime dump to the MCP so it can watch for frame hitches, identify the causes and propose fixes.

Code:
Full loop complete! We're back in MapPhase at point 23 (was 30). Visited indices now show [30, 23]. Adjacent points are 22, 24, 25, and 30
  (visited).

  That was a complete game loop:
  1. MainMenu -> start_new_game -> MapPhase
  2. MapPhase -> select_destination(23) -> start_jump -> JumpPhase/Planning
  3. Planning -> advance_step -> Initiating (auto) -> InWarp
  4. InWarp -> advance_step -> Alert
  5. Alert -> advance_step -> Combat (paused, assessed, unpaused, fought)
  6. Combat -> advance_step -> Warp2 -> advance_step -> Arrived
  7. Arrived -> complete_jump -> MapPhase at new position

  Lost 3 of 4 ships to 10 missile threats, but the Defense Frigate with PDCs survived untouched. All tools working end-to-end!

✻ Worked for 2m 40s

❯ super cool. keep playing!

● Current position: point 23. Adjacent points: 22, 24, 25, 30(visited). Let me pick point 22 — it connects via an Asteroid route with 3
  waves. Tougher fight but let's see how the lone Defense Frigate handles it.

  Actually, with only 1 ship and PDCs (anti-missile only), I might struggle against heavy threats. Let me pick point 25 instead — let me
  check the edge.

  Looking at the edges: 23->25 is Hostile with 3 waves, 22->23 is Asteroid with 3 waves, 23->24 is Hostile with 3 waves. All 3-wave routes
  except 23->30 which is visited. Let me go with point 25 — Hostile, 3 waves. Fortune favors the bold.

● subliminal-mcp - select_destination (MCP)(point_index: 25)
  ⎿  {
       "result": "Success.\n\nOutput:\nselected\r\n\n\nResult:\nNone"
     }

● subliminal-mcp - start_jump (MCP)
  ⎿  {
       "result": "Success.\n\nOutput:\nstarted\r\n\n\nResult:\nNone"
     }

● subliminal-mcp - get_game_state (MCP)
  ⎿  {"result":"Success.\n\nOutput:\n{\"phase\": \"JumpPhase\", \"step\": \"Planning\", \"combat_paused\": false, \"combat_loop\": 0, \"
     total_loops\": 3, \"player_ships\": [{\"name\": \"Heavy Frigate\", \"ship_class\": \"Medium\", \"cell\": [2, 2], \"health\": 100.0,
      \"max_health\": 100.0, \"shield\": 100.0, \"max_shield\": 100.0, \"weapons\": [{\"name\": \"PDC Left\", \"weapon_type\": \"AntiMis
     … +168 lines (ctrl+o to expand)

● Ooh, 3 combat waves on a Hostile route with just 1 surviving Defense Frigate (PDCs only). This is going to be rough — PDCs are AntiMissile
  weapons, not great against ships. But let's see what happens. Into the breach!
 
Last edited:
I could do with finding a small handful of testers....not for gameplay testing just yet (although you're welcome to mess around with what's there), but I want to sort out a pipeline for packaging builds and making them available, also be useful if it's people with different hardware (my desktop is 265k + 5080), so AMD/Intel GPU just so I get an early heads up if there's any major compatibility issues. Let me know if you'd be interested.
I've got an AMD GPU if that helps.
 
I've got an AMD GPU if that helps.
Yeah, I'll be up for some testing if you want. 7800x3D and a 7900xtx.
Same - happy to help.

5800x3d & 4070 super
Thanks! Anyone that wants to help test stuff can hop on the discord : https://discord.gg/MeVrnCZz

Nowt on there at the moment, but will start hooking my build pipelines up over the next week and setting up the channels.

Should start things properly so will sort out an NDA before I distribute anything.
 
First full day I'm actually able to focus 100% on the game, happy days.

Got my CI system hooked up to a Discord webhook so if you're interested you can see what changes I'm committing on the build channel.

Tidying up the UI so it works for all aspect ratios and fixing up the targeting functionality.

Juiced up the combat a bit last week with camera shakes and some 'reactor meltdown' type explosions for the ships and some more vfx for missiles.
 
Have you seen this?

Had a read through it....not a lot of substance to it tbh, someone just knocked that out in a few hours with Claude. Various assumptions in there that don't work in practice. All the 'art direction' stuff is fantasy.

I have to wade through reams of AI slop documentation that the CEO of my fintech client produces.....I can't stand it personally. I'm constantly pushing back to a workflow where the the skilled programmer (me) is at the center of the process, and Claude is just a tool to assist. The AI models are not good enough yet to run things....they can't make rational architectural decisions...and the danger with handing the reigns over to a massive framework like that is that it produces so much verbose output than you can't possibly as the human involved, actually process everything it's producing (because all of it *sounds* plausible)....to weed out the things that aren't right and need changing.

I find going lighter with the orchestration and stuff is best with Claude. I am still making all the architectural and design decisions about the game....Claude is just faster at writing the code than I am so I let it do the grunt work.
 
I find going lighter with the orchestration and stuff is best with Claude. I am still making all the architectural and design decisions about the game....Claude is just faster at writing the code than I am so I let it do the grunt work.

That's very cool. I'm not a developer myself, I work in a Senior Technical Manager role for a US security vendor and I'm pivoting hard into Claude at the moment. CoWork is an amazing productivity boost for me. As a test, this past weekend I told it I wanted to build an NDR project to teach my son about network monitoring tools - he's doing GCSE CompSci at the moment - it built the project scope and documentation in 5 mins for me to review. Then over the next six hours built the whole thing from scratch - commissioning a virtual appliance on my esxi server, building a packet capture/processing engine, a protocol parsing engine that does raw TCP, UDP, DNS, HTTP, TLS, SSH and SMB, a database to store extracted metadata and metrics, a threat intel engine to download IoCs from providers and process them across the database, a full interactive UI and a machine learning engine that is currently baselining my network to perform predictive analysis, peer group analysis and importance and privilege inference based detections.

I didn't write a single bit of code or design any UI elements myself, I just told it what I wanted and it built it. The only issue I had was it creating regressions when I asked it to add new features - we fixed that by me telling it to use my gitlab-ce server to store the project and use proper code version control going forward.

Finally it documented the entire repo in word and the document referenced the GCSE Syllabus and where each bit of code it created was relevant to a particular area of study!

It's nuts what you can build today without being an enterprise developer. Next stage is to use other systems to check it's work.
 
Last edited:
I find going lighter with the orchestration and stuff is best with Claude. I am still making all the architectural and design decisions about the game....Claude is just faster at writing the code than I am so I let it do the grunt work.

I'm not a programmer so I'm struggling to imagine this, but I would love to know what you mean. Can yo give an example?
 
I'm not a programmer so I'm struggling to imagine this, but I would love to know what you mean. Can yo give an example?
So, I wanted to have the targetting UI show an icon rather than debug text....I just say to claude...

"I want to change the weapon targetting UI to render an icon rather than text, but use the existing text string as a tooltip".

There's a whole bunch of code that needs to be written to do this, adding the new UTexture2D property, initialising it properly, modifying the UI orchestration code to make sure the property is fed through into the UMG widget to render etc.....

Back in the stone age before Claude :) I'd have to navigate to the weapon component, manually type in the property declaration, add the appropriate includes. Then find the UI code, type that in manually...etc. Just takes time to navigate around and enter code manually.

Now, once I've give claude that command, it takes maybe a minute to write the plan, I quickly review it, then I approve it and the changs are done in a few seconds. I can be still working in the Unreal Editor while this is happening. When it's done Claude can use my MCP to trigger Live Coding and hot reload the changes into the editor...I get notification when that happens, and I can get on with the editor work to wire up the assets.

I haven't written a line of code by hand for months. Claude is just faster and better at writing it. It's all nice clean performant code, exactly as if I'd written it by hand myself, but it's done in a tenth of the time.
 
Back
Top Bottom