Glean and Foretell
FEATURES + PREDICTION — *the clues you pick are the only ground the guess can leap from: choose the clues, and you shape — and can secretly sabotage — every prediction that follows.*
Loading audio…
Press play to listen along. The line being read lights up as you go.
Show full transcript
Loading transcript…
Glean the magpie and Foretell the tree frog work the two ends of one job. Glean stands at the near end, over a heap of clues, keeping a few and dropping the rest. Foretell sits at the far end, on the tip of his branch, taking whatever clues she hands him and leaping — guessing about something brand new. What passes between them is a small bundle of chosen clues; what comes out the far end is a guess. And here is the thing they both know in their bones: the guess can only ever be as true as the clues, and only as general as the clues let it be.
"I pick what the machine gets to look at," Glean says, patting her satchel. "And I leap from only that," Foretell says, tapping his empty spyglass. "Whatever you don't hand me, I never even know was there." "So if I hand you junk," Glean says. "Then my very best leap," Foretell finishes, "lands face-first in the mud — confidently."
Glean does the choosing, and she does it thinking one step ahead, about the leap to come. She spreads a heap of photo-clues across her end of the bench: the animal's ears, its tail, its size, the background snow, the time of day, the file name. "Watch me sort — but watch me sort for Foretell," she says. "Ears — keep, that's really about the animal, it'll still be true on a photo he's never seen. Tail — keep. Background snow — drop it; it only rode along with the wolves in my pile, and the first snowless wolf will break his leap. File name — drop it, that's about the photographer, not the beast." She sweeps the traps off the bench and ties the survivors — ears, tail, size — into a tidy bundle. "These carry the real signal. Not what sits near the answer. What causes it." She slides the bundle down the bench. "Now it's his leap — but notice, he can only ever leap on what I chose to keep. My choosing is the whole width of his world."
Foretell does the leaping, and he shows exactly how a clue's quality lives or dies at the far end. He takes Glean's clean bundle and points his empty spyglass at a photo the machine has truly never met. "From ears, tail, size — my guess: a fox. Here's my reason: pointed ears, bushy tail, small. You can check it." He beams. "A guess, with the clues showing, that'll hold up on new photos, because it leans on things that are really about the animal." Then, to show the danger, he sets the good bundle down and picks up a sabotaged one — someone slipped the snow-background clue back in. He leaps again. "From snow-in-the-picture… my guess: wolf." He shakes his head. "Same me. Same leap. But fed the trap clue, my best guess is nonsense — I'm guessing the weather, not the animal, and I'll be wrong the instant a wolf stands on grass. I didn't get dumber. I got fed worse, and the badness only showed up out here, at the leap, on a photo the pile never had."
Then a real test came, and it showed the sneakiest way the two of them can be broken together. A tool was meant to guess whether a plant was sick. Someone had let it grab every clue lying around — leaf colour, leaf spots, and the little plastic label reading "greenhouse 4," where every sick plant in the pile had happened to come from. Glean caught it first. "That label isn't about the plant," she said. "It's about where it stood — and worse, in my pile it basically is the answer, because all the sick ones lived there. Drop it." Foretell nodded hard. "Because if you leave it in and I leap on it, I don't learn plants at all — I learn 'greenhouse 4 means sick.' I'll look brilliant on your pile and be useless the first healthy plant that comes from greenhouse 4, or the first sick one that doesn't." So they did it right, as a pair: Glean stripped the bundle down to clues truly about the plant — the yellowing, the spots, the droop — and only then did Foretell leap: "sick, because of the yellowing and the spots; here's my reason, you can check it." The guess came out sharp and honest and it would survive a new greenhouse — but only because her choosing and his leaping had happened in the right order, each trusting the other. "A guess is a leap," they said, almost together, "and a leap is only as true as the ground you leap from — and the ground," Glean added, "is the clues I hand him, chosen for the plant and never for the label."
"So a good guess starts way before the guess," the kid said slowly. "It starts with what someone chose to notice — and whether that thing will still be true somewhere new?"
"That's the secret of it," Glean said, and Foretell nearly fell off his branch nodding. She tied one more clean bundle — nothing that merely rode along, nothing that smuggled in the answer — and he made one more honest leap, and together they set the finished guess, clues and reasons showing, down in the middle of the bench. Under the quiet, the two of them felt the same calm settle — not the sinking of a clever leap wasted on trap clues, not the hollow flattery of a machine that had secretly leaned on the answer, but a light, paired, well-built gladness. It was the ease of a magpie who kept only what was truly about the thing, and a frog who leapt only from that, so a guess came out between them both bold and honest and ready for a case it had never seen. That right-order, leap-from-solid-ground feeling, steadier than any lucky guess built on a coincidence, was to Glean and Foretell exactly why they worked the two ends of one bench, and trusted what passed between.
The NeuralQuest ensemble
Glean and Foretell is part of NeuralQuest's distributed-narrative cast. Each character embodies a different curricular primitive; together they teach the full subject.
-
Tag
Labeling — the cheerful labeler who treats every label as a human choice and meaning-making act ('every label is a choice — and you're the one making it')
-
Drill
Training loops — the focused practitioner who treats iteration as rhythm, not race; explicit teacher of when-to-stop ('once, again, again — different this time? Then again')
-
Skew
Bias + data fairness — the bias-vigilance anchor who always asks 'whose data is in here, whose is missing, who decided'; appears in every kit from kit 5 onward
-
Veer
Generalization vs overfit — the wandering scout who treats generalization as travel ('trained here, tested here — now go somewhere new, does it still know the way?')
-
Weigh
Ethics + decisions — the reflective elder who carries the ethics gate at the AI-in-society capstone ('can we build it? Yes. Should we? That's a different question')
-
Foretell
Prediction — the model learned on the past; now it guesses about something it's never seen (the leap is the whole point and the whole risk)
-
Glean
Feature selection — a computer can't look at everything; what you let it look at is what it learns from, so choose the clues on purpose
-
Odds
Confidence — I'm not sure, I'm 80% sure; those are different, and the difference is the whole point
-
Rue
Loss — how far off was that guess? the exact size of the miss tells the model which way to change
-
Verge
Decision threshold — the model gives a number, I draw the line; move the line and you choose what you'd rather be wrong about