Maeve

Every other chapter of this journal is a record of something I built. This one was written first, before any of it existed, and it sat that way for months. On 16 September 2026 that stopped being true: one commit, one branch, and the first object on this page is real. What follows is still the whole specification, written down in full, for the thing Frozen Dawn is pointed at. It has just stopped being only that.

1commit, and it is not merged
19files on the branch
IN PROGRESSimplementation status
DESIGN DOCstill further ahead than the code

Everything the reader has just spent seven chapters on, the collapse, the heaters, the corporation, the creatures, the hive, the rocket, was built first and explained afterwards. That is what a dev journal is. You do the work, then you write down what the work turned out to be.

This is the opposite. The specification came first and it sat in a document for months while I built everything around it. The mod already contains the world Maeve is supposed to wake up into. As of this week it also contains the first nineteen files of her, on a branch nobody has merged. That is still not the same as containing Maeve.

Chapter VIII

I put off writing this page for a long time, because a design document is not an achievement. It is a promise with a deadline attached to it. But leaving it out would have been dishonest in a way that bothered me more. The whole second half of Frozen Dawn is shaped by a system that has barely started to exist, and a journal that only shows the finished parts is a sales page, not a record.

What is written down here

So: this is the blueprint, and construction started this week in one corner of it. The laws it has to obey, the two brains it runs on, the one door it walks through, and the seven objects it is built from, one of which now exists. Where the document says a value is locked, I have said so. Where I still do not know the answer, I have said that too.

The player spends the apocalypse learning how Frozen Dawn works. Once Maeve wakes, Frozen Dawn begins learning how the player works.

That sentence has been the first line of the specification since the day I started it. Everything below is the machinery for making it true without cheating.

What it is

The one chapter that is not a record. It is the design document for the thing the other seven chapters are pointed at, written in full before any of it existed.

What you do with it

Section 05 has her running. Feed her sightings until she commits to a guess about you, then contradict her while the bet is still open and watch what it costs her.

Why it is odd

She is allowed to be wrong, in public, on a timer. Nearly every rule here exists to stop her being right too often to be readable.

Left is what she is and what she does to you. Right adds the class names, the developer dump and the order it gets written in. Less is hidden here than anywhere else in the journal, because on this page almost nothing has been built yet.

Six rules sit above everything else, including the difficulty setting.

Before there was a single class name in the document there was this list. They are marked LOCKED, which in my own notes means I am not allowed to argue with them later when something is inconvenient to build. Every mechanic further down has to be checked against all six, and a mechanic that fails one is not shipped in a weakened form. It is cut.

LaggedLaw 01Adaptation answers what the player has done historically. It never answers what they are doing this second. A director that reacts to your current input is not intelligent, it is a cheat, and players can feel the difference immediately.
LegibleLaw 02Every counter has to be observable and understandable from inside the game. If the player cannot eventually work out what changed and why, the system has not been adaptive, it has just been unfair.
ExploitableLaw 03Every adaptation creates a counter of its own. If it can be learned it can be baited, including on purpose. Feeding it false evidence is a legitimate strategy, not an exploit to be patched.
FallibleLaw 04It is allowed to hold beliefs that are uncertain, out of date, or flatly wrong, and it is allowed to act on them. Being wrong in a way the player can see is the feature.
BoundedLaw 05Attention, memory, inference depth, reconnaissance and simulation all have budgets. Nothing about it is unlimited, and the limits are fixed numbers I have to defend.
CausalLaw 06Knowledge has to trace back to something that was actually observed in the world. No server omniscience wearing a costume. If it knows where you sleep, something saw you sleep there.

The one I expect to fight with is Legible. It is easy to write an opponent model that produces genuinely clever behaviour and completely impossible to read from the player’s side of the screen. That version feels like the game is cheating even when it is not. A slightly worse decision that the player can reconstruct afterwards is worth more than a better one they cannot.

And the one that most people would call a bug is Fallible. A director that is always right is not a mind, it is a difficulty multiplier. The document is explicit about this: perfect knowledge is a design failure.

The architecture is stolen, deliberately, from Alien: Isolation.

Isolation splits its alien into two systems. One brain knows roughly where you are and steers the creature towards you. The other brain is the creature itself, which knows only what it can sense and is genuinely ignorant of your position. The horror comes from the gap between them.

Frozen Dawn does the same split, with one important difference: the second brain is not one creature, it is every Returned in the region, and the Master Architect is the piece that both brains can see at once.

Brain one

The Director

Regional, slow, and never rendered. It holds beliefs about the player, chooses an objective, and issues pressure. It does not have hands.

Runs on a tick budget
Wiped on ERASED
Never an entity

Brain two

The bodies

Individual Returned, Architects, and the Master Architect. They act on what they can actually perceive plus whatever packet the Director has pushed to them.

Local senses only
Can be deceived
Can refuse the packet

The Director never manifests. There is no boss bar, no entity, no health pool, and no model in the files.

This is the single most important canon rule in the whole document, and it is the one I have had to explain the most: Maeve is not a person and she is not a monster. She is the emergent collective will of the hive. The Returned experience Director pressure as something arriving from outside themselves, and the name they give it is Maeve. The myth is the system, perceived from the inside by the things the system is running on.

Difficulty is theologyBrutal does not give the Returned more health. Brutal means Maeve is paying attention to you. The hive treats sustained attention as a religious event, because from where they stand it is one.
She never speaks to youThere is no dialogue, no taunt, no voice line. Everything the player learns about her they learn from Returned behaviour and from Thaeven text written by people who are inside her.
The payoff is in the sourceThe design document says it outright: dataminers finding MaeveDirector is the payoff. She is a class name. That is the whole reveal.

Waking her is a one way door, and the game tells you so in a language you should not be able to understand.

Activation sits at the late end of Phase 6, gated on Major Hearths. It is not a boss trigger and it is not an item pickup. It is a state change on the region that cannot be undone, reloaded around, or walked back.

How you are told

The signal is a Thaeven transmission. Not text on screen, not a cutscene: the player receives the Orsha sensation of a door closing, delivered in their own inner voice. By this point in the game the player has been reading Thaeven for hours in the archive, which means the moment lands as comprehension rather than as a notification.

Design note · first encounter

No counter fires in the first encounter.

The first time the player meets her she does nothing clever at all. She watches. Every counter she ever plays is built out of what happened during the encounters where she appeared to be doing nothing.

That restraint is load bearing. Law 01 says adaptation is lagged, and the cleanest way to prove a system is lagged is to have it visibly do nothing the first time. The player should get through the first encounter thinking it was overhyped, and then spend the second one finding out what she was doing during the first.

The terminal state is called ERASED. When a region reaches it, the tactical belief store is wiped completely: every hypothesis, every confidence value, every piece of evidence. She does not remember how you fight.

She does remember what you did. That is a different ledger, and it survives.

She does not store what you did. She stores what she thinks you did.

This is the difference between an opponent model and a stat tracker, and it is where most adaptive AI in games quietly gives up. It is far easier to read the player’s actual inventory than to maintain a guess about it. The document does not allow the easy version, because Law 06 says knowledge has to trace to observed evidence.

So the unit of memory is a belief record. Every one of them carries its own confidence, its own evidence list, and its own expiry.

A belief record, as the dump prints it Section 9.1
// approach.roof belief the player enters from above confidence 0.68 support 3 observations contradicted 0 times first_seen day 34 last_seen day 41 stale_at day 61 // 20 in-game days after last support evidence ARCHITECT_SIGHT day 34 east ridge HEARTH_WITNESS day 38 roof breach ARCHITECT_SIGHT day 41 rope anchor
One of them, in words. She thinks you come in from above, and she is about two thirds sure of it. Three things put it there: an Architect saw you on the east ridge on day 34, a hearth saw the roof breached on day 38, an Architect found a rope anchor on day 41. Nothing has contradicted it yet. If nothing new supports it by day 61 she starts letting it go.

Staleness onset is a locked value: twenty in-game days after the last supporting observation, a belief begins to decay. It is not deleted, it is demoted. She keeps believing you come in through the roof for a while after you have stopped, which is exactly the behaviour a real opponent has and exactly the behaviour that makes deliberate disinformation worth the player’s time.

There are two memory stores and they are strictly separate. Mixing them was the single worst version of this design I wrote, and untangling them took most of a weekend of notes.

Tactical beliefsWiped on ERASEDHow you fight. Where you come in. What you carry. What you are afraid of. All of it is guesswork with confidence attached, and all of it is destroyed when the region is erased. A new region starts genuinely ignorant of you.
Violation memoryDeath proof, permanentWhat you did. Hearths you burned. Returned you killed. Whether you honoured a truce. This is not tactical information and it does not decay, reset, or forgive. Dying does not clear it.

The split is what makes the ending of a run mean something. Wiping her tactical model is a reward, and it is available: erase the region and she forgets how to fight you. The other ledger is not negotiable. She forgets your tactics and remembers your conduct, which is, I think, the more frightening way around.

Not machine learning. Nothing here trains, and nothing here has weights. It is opponent modelling with counter strategy selection, which is a forty year old idea and a completely honest description of what the code does. I am also not going to market it as Nemesis like: that system is patented (US 10,926,179) and the comparison is both legally careless and technically wrong.

The mechanic that makes fallibility legible is called commitment, and it is the load bearing piece of the whole design.

A director that only ever acts once it is certain is indistinguishable from a director that can read the save file. The player never sees it think, they only see it be right. So the rule is the opposite: once confidence in a belief crosses 0.75, she has to act on it early, visibly, and before it is confirmed, and then she has to hold that commitment for about twenty seconds whatever happens next.

Why she has to hold it

Commitment is sticky on purpose. If she could drop a bet the instant it looked wrong, the player would never get to watch her be wrong, and being watched being wrong is the entire point.

Below is the developer dump from section 9.15, wired up.Below she is actually running. Feed her sightings and watch the confidence climb. Cross the threshold and she commits out loud. Then contradict her while the commitment is still live.

Director diagnostics In‑game day 30
Beliefapproach.roof · the player enters from above
Confidence0.00
Evidenceno observations
Contradictions0
Uncertaintynothing has been seen yet
Objectivehold the hearth
Strategynone selected
Architect packetnone

The evidence list is empty. She has no opinion about you.

The twenty second hold runs at roughly three times speed so it fits in a browser. Every other number here is the one in the document.

Three things in that panel are not tunable, and I have written them down as not tunable so that a future version of me does not quietly soften them when playtesting gets uncomfortable:

One per encounterShe gets a single active commitment at a time. She cannot hedge by committing to two reads of the player at once and taking credit for whichever lands.
Precedence by costWhen two beliefs both qualify, she does not take the higher confidence one. She takes the cheaper bet to abandon. Confidence decides what she believes, not what she is willing to stake.
Contradiction cooldownAfter a commitment is contradicted, she cannot open a new one for a fixed window. Getting caught costs her tempo, and the player can spend that window doing the thing she was about to stop.

Law 03 falls straight out of this. If she commits on evidence, and the evidence is things you were seen doing, then letting yourself be seen doing the wrong thing is a weapon. Walking a decoy route past an Architect three times and then taking the other one is not an exploit to be patched. It is the intended skill ceiling.

Law 05 says she is bounded. In practice that means she can only pay attention to a few things at once, and that number is the difficulty setting.

2focus slots on Cinematic
3focus slots on Normal
5focus slots on Brutal

Attention is the only thing the difficulty slider moves. Nothing gets more health.

I went back and forth on this one for a long time, because the obvious version is a pool. Give her a meter, make her spend it, let it refill. It reads well on paper and it is wrong for two reasons.

Why not a meter

The first is that a pool is farmable. Anything with a refill rate turns into a timing puzzle, and within a week somebody has worked out how to drain her and walk past. The second is that it is off fiction. A hive does not run out of attention. It runs out of places to put it. So attention is a concurrency cap, not a currency: a fixed number of slots, always full, always doing something, and taking an interest in you means dropping an interest in something else.

What she stops watching

Which raises the only question that matters. When every slot is busy and something new happens, what does she stop watching? The answer is ranked by legibility, not by value. She drops the things whose loss the player can most easily read.

ReconnaissanceSpeculative scouting goes first. You will notice the pressure ease before you understand why, which is exactly the tell I want.
Passive trackingLow confidence beliefs stop being fed. They go stale rather than wrong, and she quietly forgets a habit of yours.
Active commitmentAn open bet can be abandoned under load, but the twenty second hold still runs out first. She never escapes a commitment by getting busy.
Siege orchestrationCoordinated pressure on a shelter collapses back into individual behaviour. The bodies keep coming, they just stop arriving together.
The Master ArchitectNever evictable. Whatever else is happening on the map, one slot is always on the guardian, because the moment it is not, the thing it guards stops mattering.

That last line is a design lock as much as a technical one. The Master Architect is the only entity in Frozen Dawn that Maeve is never allowed to lose track of, and the reason is in Chapter VI: it is not guarding itself.

Design note

Brutal is not a damage multiplier. It is five slots instead of three. Everything that makes the hard difficulty hard is her noticing more of what you are already doing.

There is a strict limit on how much of this is allowed to live in one class, and it is about three hundred lines.

Everything above is a lot of machinery, and there is a very ordinary failure mode waiting at the end of it: one enormous god object called MaeveDirector that does belief storage, inference, attention, commitment, planning, strategy selection and logging, and that nobody, me included, can safely change a year from now.

So the name is a facade and nothing else. It owns no logic. It holds seven collaborators, delegates to them in a fixed order once per director tick, and stays small enough to read in one sitting.

MaeveDirector~300 lines, hard cap
  • BeliefStoreHolds the beliefs, their support, their contradictions and their decay. Knows nothing about what any of it means.
  • WorldModelHearth states, routes, thermal geography, where the bodies are. The facts, as distinct from the opinions.
  • AttentionManagerOwns the slots and the eviction order. The only thing allowed to decide what she stops watching.
  • CommitmentPolicyThreshold, hold duration, precedence, cooldown. The three non tunables live here so they are hard to quietly change.
  • MissionPlannerTurns a selected strategy into orders that bodies can actually carry out, given where they are.
  • StrategySelectorPicks the counter from the unlocked tree. Bounded by the counter cap, which sits between five and eight.
  • DirectorDiagnosticsPrints the developer dump. It exists because a system this opaque is untestable without a window into it, and because I would rather read her reasoning than guess at it. The panel two tabs back is this class, rendered in a browser.
The dump, in full Section 9.15
// DirectorDiagnostics, one tick, one focused target
BELIEF               approach.roof            0.81
EVIDENCE             4 observations, newest day 47
CONTRADICTIONS       1                        // day 44, ground entry
UNCERTAINTY          moderate, one recent contradiction
DIRECTOR OBJECTIVE   deny the northern hearth
SELECTED STRATEGY    SEAL_HIGH_APPROACH       // counter 3 of 6 unlocked
ARCHITECT PACKET     reposition east ridge, 2 slots, hold 20s

That dump is the honesty check for the whole chapter. If I can read those seven lines and understand why she did what she just did, the system is legible. If I cannot, then neither can the player, and Law 02 has been broken somewhere upstream.

And this is the one place in the journal where the name appears on purpose. Dataminers finding MaeveDirector is the payoff. There is no entity called Maeve in the game files, no dialogue, no boss bar, no health value. There is a class that coordinates the hive, and a people who feel that coordination from the inside and gave it a name.

There is one slice of code, out of eight. What there mostly is instead is an order to write them in, and a set of conditions that say when each piece is allowed to be called finished.

Eight slices. Each one has a testable done condition, and no slice starts until the one before it has met its bar. That constraint exists because I know exactly how I would otherwise build this: all of it at once, in a single branch, for two months, and then discover none of it is debuggable.

Belief storeBeliefs, confidence, evidence, contradictions, staleness, provenance. Observations write to it. Nothing reads from it yet. Done when a developer command dumps a populated store after a normal session and every entry traces to a real observed event. Started 16 September 2026. The developer command arrived with it, so slice 02 has already leaked into slice 01.
ObservabilityThe developer dump, built second rather than last, because nothing after this slice is debuggable without it. Done when the tool can answer, for any belief, what she believes, why, and what contradicts it.
CommitmentThresholds, precedence, the one per encounter cap, the cooldown. Done when a blind tester, unprompted, describes an Architect waiting somewhere that turned out to be wrong.
World MVPAccess points, danger zones, heat sources, derived unknowns, and spatial beliefs keyed to bearings rather than named exits. Done when sealing one side of your base makes her walk into the discovery.
AttentionFocus slots, eviction priority, visible eviction, minimum dwell. Done when a tester can reliably shed a stalker by making noise elsewhere and reports it as a tactic rather than a bug.
ReconnaissanceMission packets and information gain as utility. Done when a tester says some version of that one wasn’t attacking me.
DepthConditional player models and strategy performance memory. Depth on a proven core, never as a substitute for one.
OptionalPersistent individual Architects, tactical traits, bounded forward simulation. Only if everything above is stable, and honestly only if I still want to.

Slice 3 is the one I care about. Its exit criterion is not a unit test, it is a sentence out of a stranger’s mouth, and the document says what happens if that sentence never comes: stop and fix it before continuing, because if nobody notices her being wrong, nothing else in this system will land either.

Then the tuning, which is the part that actually scares me. The window between reads as stupid and reads as cheating is narrow and only findable in play. Thresholds first, then how fast confidence climbs, then decay, then slot count last because it interacts with everything above it. Targets: roughly two thirds of her commitments correct, and a tactic that works three to five times before the answer arrives.

And the rule I wrote down mostly as a warning to myself: never tune two variables in the same session. There are six interacting numbers here, and adjusting all of them by feel is how a system like this becomes unfinishable.

One capability was designed in full and then deliberately cut. Maeve could track which counters you have already seen, and then pick a familiar one on purpose to bait a prediction before switching to something else. A feint.

Why deception is not allowed

It is cut because a feint is definitionally illegible. Its whole function is to be misread. Every other counter in this system telegraphs: a changed search pattern, preemptive cover, a held position. Deception cannot, and when it works the player has no way to tell the difference between she deceived me and the game did something random and I lost. Small fraction of the emotional payoff, large share of the quality assurance surface, highest risk of the whole thing reading as unfair. Law 02 wins.

Four things are still genuinely unresolved, and I would rather leave them written down as questions than pretend the document is finished:

Veteran legibilityIf an individual Architect remembering you is never perceivable, that layer is engineering nobody experiences. Either veterans get a legible tell, or persistent individual history gets cut entirely.
The desperation phaseAfter an interrupted heal, retreat utility is zeroed for a minute, and what happens in that minute is currently emergent. It is also the emotional peak of the fight. Last stand or last wall, and I have to pick one on purpose.
The shield counterIf blocked hits void the freeze riders, a shielded player invalidates the entire melee loop. It needs an in character answer, not a weapon swap.
Twenty secondsThe commitment hold is the number this whole chapter rests on and it is currently a guess. It has never been tested against real encounter pacing.
The branch contract

There is a branch contract, and it was written long before there was any work for it to govern. Maeve gets a protected integration branch cut from a clean main, one short lived branch per slice, pull requests that target the integration branch and never the mainline, and a merge bar that includes the exit criterion from the list above. Notion records the contract. GitHub enforces it. The first branch cut under it is feat/maeve-belief-store.

Why it was written first

I wrote all of that before a single line of it existed, which was either good discipline or a very elaborate way of not starting. Probably both, and it is no longer hypothetical either way. I have built enough of this mod now to know which parts of it hurt, and every rule on this page is a scar from something earlier in this journal: the Architect that knew too much, the systems that were clever in ways nobody could see, the week I spent tuning four numbers at once and could not tell you afterwards which one had helped.

Where that leaves it

So this chapter is still mostly a promise rather than a record. The rest of the journal is what I built. This is what I am going to build, written down carefully enough that a year from now I will not be able to pretend I meant something else. What changed this week is that the document finally has something to be checked against.

The player spends the apocalypse learning how Frozen Dawn works. Once Maeve wakes, Frozen Dawn begins learning how the player works.

She is not awake yet.

1 commit · 19 files · feat/maeve-belief-store, unmerged
This chapter is not listed anywhere