Reflection and Plan for a Demo
Yesterday, I took my final exam fro the spring semester—so I’m finally back.
In a way, the break was useful. It gave me some distance from the work and a chance to reflect.
Right now, we're building Type 2. But honestly, it feels like we've been bouncing around—exploring every direction without landing on the one that matters the most.
The goal has always been the same:
create something that feels alive.
Not just a system that works, but a demo that hits people emotionally. Something that makes them pause. Something that stays with them.
Like the feeling you get from Diana in Pragmata—not because it's realistic, but because it lands. It creates a sense of longing, or nostalgia, or curiosity. It makes you feel like there's something more there.
That's what Sprited needs.
Not somthing explained through words—but something people instantly feel when they see it. Something distant. Something hard to replicate. Something with its own gravity.
What we've tried so far
(-5) Pixel-being concept: growing an organism from a single pixel
(-4) Built Machi: a 2D side-scrolling pixel world
(-3) Added material layers for physical properties
(-2) Introduced foreground/background for depth
(-1) Simulated tree growth within Machi
(0) Defined: “Sprited is a Digital Being company”
(1) Updated homepage to reflect that vision
(2) Explored limb growth from a single cell
(3) Pivoted to capsule-based body modeling
(4) Dropped capsules → returned to SpriteDX sprites
(5) Designed a 5-organ system (hunger, fatigue, etc.)
(6) Needed an environment for simulation
(7) Tried reusing Machi → lacked locomotion
(8) Switched to Celeste-style grid + locomotion (OSS)
(9) Implemented movement + pathfinding
(10) Questioned positioning:
“Are we building digital beings, or virtual worlds for AI agents?”(11) Added food + HP decay
(12) Built an autonomous agent that seeks food
Where we are now
We have:
A Celeste-style environment
A working locomotion system
A basic agent loop
But the "being" is still just a red rectangle.
It works—but it doesn't feel alive.
Our plan
Our plan is to continue form where we left off.
Right now, we have an agent that moves in response to hunger. It feels very mechanical, but it does at least demonstrate that the physical-needs and satiation loop is functioning.
That's an important milestone.
The next step is to introduce more than one drive.
We need a second internal need besides hunger—something like fatigue/sleep, or perhaps thirst/drinking if that is easier to implement first. A one-drive system is too simple. There is no real tension, no tradeoff, and no meaningful choice. The agent just keeps seeking food whenever the threshold is crossed.
At the moment, there is no decision-making system. It is simply a threshold check that pushes the character into "get food" routine.
What I want to do next is integrate a lightweight decision-making model—something like low-power Gemma 4 E2B—and use it to choose between possible actions based on the current state.
For example:
get food
go to sleep
That would give us the beginnings of an actual feedback loop.
Instead of the character following single hardcoded survival reflex, it would have to choose among competing needs. That is where more lifelike behavior can start to emerge.
Alternatives
Let's lay out the options.
(A) Stick to the plan
Continue building on what we have. Evolve the system step by step—add another drive, introduce decision-making, and push toward emergent behavior.(B) Create a fake demo video
Instead of building the system, create a mocked-up demo that looks alive. If the goal is to evoke a feeling, a well-crafted video could achieve that. It could also double as an investor pitch—something polished, intentional, and emotionally compelling.
My take
Option B is tempting.
A strong demo video could communicate the vision quickly and effectively. It might even resonate more immediately than a rough, real prototype.
But there are two problems:
Execution risk — I'm not a video production expert. Creating a high-quality, convincing reel would be difficult and time-consuming.
Misalignment with strengths — I'm a builder. A tinkerer. I move fastest when I'm iterating on real systems, not crafting polished illusions.
More importantly, we're not actually blocked right now.
We have a working foundation. It's primitive, but it's real. There is sill clear forward momentum in the system itself—especially around adding competing drives and decision-making.
So for now, we continue building.
If we later reach a plateau—where progress slows or the system isn't communicating the vision clearly—then creating a demo reel becomes a strategic move.
But we're not there yet.
Second Perspective
There's another constraint we need to factor in.
A sub-goal is to have something we can showcase at the June 18 Tech Weekend event. This will be the first time Sprited is in front of a potential investor audience, so we need something we can present—along with a compelling story around it.
We have less than two months.
Realistically, even less than that—given upcoming kid-related travel and limited working time.
So whatever we choose to do next can't be "interesting" or "correct" in isolation. It needs to converge toward something demo-able within that timeframe.
What's a "A Compelling Demo"?
A compelling demo is not just something that works.
It's something that makes people feel like they just saw something new—even if they can't fully explain why.
It should be understandable in seconds, but leave a lasting impression.
Requirements
Born into a World
The being is not spawned arbitrarily—it is born into a virtual, physical environment.There is a sense of place
The world has rules (gravity, obstacles, resources)
The being exists within that world, not separate from it
Driven by Internal Needs
The being is governed by multiple internal drives, such as hunger, fatigue, and curiosity. These are not just numbers—they create tension. The being cannot satisfy all needs at once.Makes Decisions
The bing must choose. Not just react. Even simple choice like "eat vs. sleep" is enough, as long as it is visible and understandable.Feels Alive
This is the hardest requirement.
"Alive" is not realism—it's perceived intention.
The observer should feel:"It wants something"
"It hesitated"
It changed its mind"
It is doing this for a reason"
Comes with a Pitch
The demo alone is not enough. It must be paired with a clear, simple narrative.
Litmus Test
Someone can watch it for 20-30 seconds
Without explanation
And say: "wait… is that thing actually doing that?"
The moment of curiosity—that pause
Next Step
...is to get back to working. I will add second drive mechanic. After that, we will add in Gemma (or something) for decision making. Wish me luck!
Pixel also suggests looking into:
Stochastic failures
Hesitation
Reversal of thoughts
Visualization of decision making process.
Q: Isn't this just like The Sims?
Valid question. I think this may come up. We are building "digital beings" in a virtual world which is a similar in concept with Sims.
v0.1: Imagine Sims but with emergent behaviors in a semantic virtual world. The environment grows with the agent and the agents form the world, communities and stories within.
v0.2: It looks like Sims, but instead of controlling characters, you interest with autonomous beings that make their own decisions and shape the world.
v0.3: Imagine Sims but with true embodiment. Instead of shells moving, it fully embeds itself into the virtual world and shape the world.
v0.4: It looks like Sims, but instead of controlling characters, you interact with autonomous beings that make their own decisions—and the world evolves from those decisions.
v0.5: It looks like Sims, but instead of controlling characters, you interact with autonomous beings that make their own decisions.
The interesting part is—you don't always get what you want.
— Sprited Dev 🐛
