Esta es la versión en español de la guía de Bahkruums. Conserva los hechos, identificadores, enlaces y límites de evidencia del corpus fuente en inglés; los nombres oficiales, teclas y valores entre código permanecen sin traducir. Usa los enlaces internos de esta versión para seguir páginas del mismo idioma.
Bahkruums does not publish an AI specification. Nextbot inspiration is official, but exact targeting, pathfinding, speed, damage, sound response, spawn, and despawn rules are not. A behavior guide should help players separate repeated current observations from one chaotic chase.
Build a controlled observation
Choose a route with two visible landmarks and a simple junction. Record the date, current server, player count nearby, entity cue, starting positions, and the action you want to observe. Change one variable per attempt.
Do not attempt precise tests in a crowded intersection. Other players can cross paths, use communication actions, or become targets. If the environment cannot be controlled, label the result ordinary gameplay observation.
Pathing through turns
Observe whether the threat follows the same branch, changes at the junction, or loses contact. Repeat the route from the same direction. A single failed pursuit can reflect target selection, distance, another player, or geometry; it does not prove the corner defeats that nextbot.
Record wide and tight turns separately. Do not generalize a result from one shape to every map. If the geometry changes after an update, reset the behavior status.
Target-selection observations
Note every visible player and when each crosses the threat’s path. Record whether line of sight changed and whether anyone used Q Scream or T Speak. Do not infer “nearest player,” “loudest player,” or random targeting without repeat comparisons.
A target switch is the beginning of a question, not the answer. Compare attempts where the same positions are used and one action changes. In public play, strong conclusions may remain impossible; Community reported is an acceptable label.
Sound-response test
The game officially provides Scream and Speak controls, but no AI response is stated. To test safely, use comparable route conditions and have one player perform one action while another observes. Repeat without the action. Record what changed and what did not.
Do not spam sound or disrupt other players. If results conflict, retain the unverified status. Audio volume and recognition are separate from AI behavior.
Speed and distance
Avoid unsupported numeric speed. Compare travel between the same landmarks under similar conditions. Note whether the distance closes, stays stable, or expands while using controlled sprinting. Latency, camera angle, route length, and player movement can distort estimates.
Do not produce a fastest-nextbot ranking until identities, routes, and measurement methods are stable. A short edited video is unsuitable for precise comparison.
Doors, openings, and geometry
Record the exact opening shape and whether it was already open, interactive, or blocked. A threat failing to enter once may have changed target. Repeat current observations before calling the geometry inaccessible.
Never call clipping, trapping, or unreachable geometry a permanent safe spot from one attempt. Beta fixes can change pathing, collision, or access. Send uncertain locations to the Safe Spots guide as candidates, not guarantees.
Spawn and despawn
No spawn table or schedule is verified. Record the first visible location, prior event, server age if known, and whether the same cue appears again. Do not conclude rarity from a small personal sample.
For disappearance, note line of sight, distance, target changes, round state, and server events. “I stopped seeing it” is not a confirmed despawn.
Evidence record
Use: date | cue label | route | players | action | observed result | alternative explanations | confidence. Keep original observations separate from later interpretation. Two independent current sources that match can raise a claim to Community corroborated.
When sources disagree, use the newest complete directly relevant record as the working report and show the conflict. Do not choose the largest speed or strongest effect.
Behavior to survival decisions
Convert only repeatable behavior into route choices. If a branch consistently preserves visibility and has a known continuation, it is a useful route even without an AI claim. If a turn result is inconsistent, plan an escape continuation instead of relying on it.
Use Identification for cue stability and Nextbot Survival for the immediate chase loop.
Behavior FAQ
Do nextbots hear Q Scream?
No official source says so. It needs controlled, repeated current observations.
Who do nextbots target?
No targeting rule is verified. Multiplayer makes one-off observations especially ambiguous.
How fast are they?
No reliable numeric dataset exists. Compare current travel along fixed routes before making qualitative claims.
Can a corner stop them?
One corner result is not universal. Record geometry and repeat before using it as a route rule.
Sources checked 2026-08-07 UTC: official nextbot inspiration and controls from the Roblox description. Behavior fields remain explicitly current in-game test items.