We’re Not in Kansas Anymore. Me Neither!

Brought the player into the game world this morning to give the little #ucker a reality check.

Pretty soon, we will be answering to investors, milestones, ROI and all the other criteria that measure success in the real world.

But AI doesn’t live in the real world.

And on most days, it feels like the AI companies don’t live in it either.

The tech is mind-blowing. The blue sky, galactic.

But right now, I’m on earth. This is reality.

And I’m in a good position.

4 machines. All Apple. One brand-new M5. Studio screens. Storage. High-speed internet and a $300 Codex account plus Claude Pro. Studio monitors.

I have a yard which grinds to a standstill when I push the pedal to the metal with any fast character movement.

And I still have 90% of elements outstanding.

And getting AI to forecast anything is impossible because it has no comprehension of time as we know it. It doesn't have to pay the rent.

I was taught to make television by Murphy’s law. What can go wrong will go wrong. I grew up in the 80s under the steel heel of hard, uncompromising men who had a daily TV show to get to air. Failure wasn’t an option when you were a Packer, Murdoch or Wonder World! reporter.

40 years of entrepreneurship also hardens you.

So I test and re-test—way ahead of ever having to.

I plan for the worst.

Load the system up.

Overload it.

Then redline it.

I’d rather do it now on my own money than somebody else’s.

I won’t bore you with all the stuff that needs to go right in a video game for it to work. But if there’s one stray character in the code, it ain’t pretty.

So this morning, we started the Reality Checks.

The first measurement points to drawing the scene: about 31 ms per frame, compared with under 3 ms for the measured collision/security work. That already exceeds the 16.7 ms budget for 60 fps.

The renderer is issuing roughly 4,800 draw calls in this view. Blah. Blah.

For Carny as I’ve described it, tens of thousands to a few hundred thousand lines of custom code is a plausible scale for the finished game. That’s a broad comparison, not an estimate derived from a complete Carny specification.

That excludes the engine, libraries and generated artwork data. Those can add millions of lines without representing work we need to write.

The main systems would be:

- Movement, camera and interaction. - Grave excavation, inventory and body assembly. - Shrift, threats and game rules. - Animation, sound, lighting and weather. - Saves, interfaces and eventually phone/multiplayer connections.

Line count won’t tell us whether Carny is affordable or nearly finished.

A small rendering mistake can stall the whole yard; thousands of additional lines won’t necessarily fix it.

This pass:

- Reduced tunnel draw calls from 4,831 to 2,353, preserving artwork resolution and triangles. - Improved drawing time from roughly 31 ms to 22 ms. - Restored visible Skritch and corrected discarded movement time. - Passed 12 crossing legs across three graves. - Verified the speech-to-music playback sequence.

The yard still misses the performance target. Remaining-art capacity and a sustained 20-minute walkthrough are unproven.

The video is rendered frame by frame. It shows the artwork and movement; it doesn’t prove the game can sustain them in real time. That’s why these reality checks matter.

So, now I’ve established the Scratch’s size relationship to the yard, I am returning to decorating it fully and then starting a new LAB to work on the camera angles of the Scratch digging a grave from above ground.

I’m not flogging AI. I’m saying use it—but use it with real-world wisdom and experience.

If left unchecked, AI will give you the world. As an atlas.