Firefly Is Back for Halloween — This Time, It Wants to Know What You’ll Like to Eat 👻🔥
There are ghosts in technology.
Some are forgotten products.Some are ideas that arrived twenty years too early.Some refuse to die.
And some come back for Halloween.
At Ajinomatrix, we still haven’t delivered iSmell 2.0 to Apple.
Yet.
But while Cupertino somehow keeps surviving without a smell port, we did manage to resurrect another delicious piece of digital history:
the spirit of Firefly AI.
Only this time, it doesn’t want to know which song you’ll like.
It wants to know:
Will U Like It?
And yes, we actually built it.
Welcome to:
Back from the ashes
If you were online early enough to remember the first wave of recommendation systems, Firefly was one of those magical ideas that seemed to come from the future.
Tell the machine what you liked.
Rate a few things.
Let it understand your preferences.
Then ask:
“What else might I like?”
For music, that was revolutionary.
For Ajinomatrix, the obvious question became:
Why stop at songs?
Why not ask the same question about food?
And because this is Ajinomatrix, we didn’t stop at:
“People who liked this also liked that.”
We connected it to something much stranger.
A recipe can now become a structured JRF process.
That process can become an Expected MP6 sensory prediction.
Your own taste preferences can be expressed on compatible sensory dimensions.
And then WillULikeIT compares:
you × the predicted experience
to produce a transparent:
Personal Sensory Fit
Not a prophecy.
Not a magic probability.
Not “AI knows your soul.”
A measurable, inspectable estimate of how closely the predicted sensory experience matches what you say you like.
Then comes the important part.
You taste it.
And reality gets a vote.
Firefly for food, but with evidence
The original recommendation dream was:
tell us what you like → we predict what else you might like
WillULikeIT adds a sensory layer:
tell us what you like → describe what the food is expected to feel like → compare the two → taste it → tell us whether we were right
That produces something extremely useful:
prediction + outcome
Every time.
And because Ajinomatrix has become mildly obsessed with provenance, the original prediction stays frozen.
No retroactive cheating.
No rewriting history because dinner disagreed with the algorithm.
The machine makes its guess.
You eat.
Reality grades it.
Halloween appropriate, really.
It starts with you
WillULikeIT now opens with a rapid preference sweep.
A few taps establish an experimental taste profile.
You can refine it further if you want.
You can also build profiles for the people you cook for.
Your partner.
Your children.
Your impossible uncle who hates coriander but claims he is “easy.”
Then load:
an Expected MP6,
a JRF recipe,
or one of the demonstration cases,
and WillULikeIT calculates the fit.
You can see:
where the prediction matches your preferences;
where it diverges;
which attributes matter;
how much of your preference profile was actually covered.
And because missing information is not magically turned into zero, the system also tells you when it does not know enough.
We consider this a feature.
And then FEELLLM enters the séance
Of course, once you know that something is likely to fit you, the next question is:
Why?
Or:
What should I change?
Or:
Can I make this less sweet without losing the roasted character?
This is where FEELLLM.org comes in.
WillULikeIT can generate a structured context brief containing:
your sensory preference profile;
the recipe;
the predicted MP6;
the fit result;
the strongest matches;
the mismatches.
Then your preferred AI can reason over the problem without confusing:
what was predicted
with:
what was actually tasted.
That distinction matters.
A lot.
So, did Ajinomatrix finally make iSmell 2.0 for Apple?
No.
Not yet.
Tim Cook can relax.
But we did something else instead.
We took one of the most charming ideas of early AI recommendation systems…
dragged it out of the digital graveyard…
gave it JRF…
gave it MP6…
gave it FEELLLM…
gave it ABED…
and pointed it at dinner.
Firefly is back.
For taste.
For Halloween.
Nostalgie quand tu nous tiens. :3
Try it:
Related tools:
And remember:
The prediction freezes first. Then reality gets to haunt it.
Happy Halloween from Ajinomatrix. 🎃👅🔥
Comments