In-Game Notebook Design — Where Should a Deduction Puzzle Keep Its Memory? (from Obra Dinn to Chants of Sennaar)

Introduction

A grid-ruled notebook sleeps in my desk drawer. It dates from the week I played Return of the Obra Dinn: sixty names, faces, and shoe details scrawled across its pages under arrows and crossings-out. Yet when I played The Case of the Golden Idol—another deduction game—my notebook stayed almost blank. The game had built the writing space into itself. The location of my memory had moved from paper into the game.

I have come to think the experience of a deduction or observation puzzle depends on the design of this "location of memory" as much as on the design of the board itself. Does knowledge live in the player's head, on paper beside the keyboard, or in an in-game notebook? And if in a notebook, how much fills itself automatically, and where must the player's own hand take over? This essay rereads the games from Observation as Play through that single lens, as part of the observation puzzle series.

Return of the Obra Dinn (from the official Steam screenshots)

Designing for Paper — External Memory in La-Mulana and Her Story

La-Mulana, released as freeware by GR3 PROJECT in 2005, is notorious for carving clues into stone tablets that you will need dozens of hours later, somewhere else entirely. The game offers no adequate recording tool; real-world notes were effectively mandatory. I read this less as unkindness than as a deliberate reenactment of MSX-era play—the desk where a magazine's tips column and a grid notebook sat side by side. Placing memory outside the game meant play seeped out into the physical space of the desk.

Her Story (2015) refined this forced externalization into something modern. The game is nothing but a search box over a police-interview video database: no journal, no flags, no checklist. Whom you suspect and which words you have tried accumulate only on your own paper. By refusing to record any progress of reasoning, Sam Barlow handed the player even the authority to decide that the case is solved. The ending itself is externalized.

The strength of paper is the freedom of its format. Some players build tables, some draw relationship charts, some keep timelines; the shape of the notes becomes the signature of the reasoning, and a culture of sharing finished notebooks grew around these games. The cost is just as clear: put the game down for a few days and your own handwriting turns illegible, and the play breaks there. A paper notebook is terribly fragile against interruption.

The Self-Filling Logbook — Obra Dinn's Rule of Three and Outer Wilds' Rumor Mode

Return of the Obra Dinn (2018) advanced the in-game notebook by a full generation. Lucas Pope's logbook appends a sketch and a cast list automatically each time you witness a death. The enormous clerical work of matching sixty faces to scenes is absorbed by the game, leaving the player only the core of the deduction: who died, and how. What filled my own paper notebook, I realize now, were the hypothesis fragments this logbook could not yet hold.

The masterstroke is the granularity of verification. Only when three fates are correctly entered does the handwriting turn to print and lock in. Grade each fate instantly and brute force becomes optimal; withhold all judgment until the end and certainty never arrives. The unit of three is a delicate arbitration—punishing guesswork while periodically returning the satisfaction of a completed read. It remains one of the most elegant answers to the question I raised in hint system design: how to show, and how to conceal.

The ship log of Outer Wilds (2019) pushes automation further and draws the relationships between clues themselves. In Rumor Mode, discoveries become cards, lines connect them to related places, and unexplored gaps glow with a marker that says more remains here. Since the world rewinds every twenty-two minutes and progress lives only in knowledge—as I argued in the time-loop essay—this log is, in effect, the save file itself made visible.

Slots for Hypotheses — Golden Idol's Fill-ins and Sennaar's Verification Pages

If automatic recording takes charge of facts, the next question is where hypotheses should be written. The Case of the Golden Idol (Color Gray Games, 2022) answered head-on with its thinking screen, where words scavenged from scenes are slotted into fill-in-the-blank sentences. The blanks in "X killed Y with Z" work as parking spaces for reasoning in progress: place an uncertain word provisionally, swap it out when a contradiction appears. My notebook stayed blank because the game had already built a home for hypotheses.

The fill-in format carries its own difficulty: the moment a sentence is filled, it looks plausible. What makes Golden Idol's grammar excellent is the breadth of its slots and vocabulary, tuned so that mere completion never guarantees correctness. Reconnecting characters and motives across twelve deaths spanning forty years, the player tears down filled sentences again and again. The slots are designed not as an answer sheet but as scratch paper with an eraser attached.

Chants of Sennaar (2023) transplanted the hypothesis slot into a decipherment puzzle. Against each unknown glyph the player writes a guess as a free annotation; when a journal page assembles, the game asks you to match glyphs to illustrations, and meaning locks in only when the matching is right. Guesses flow freely, confirmation arrives in batches—Obra Dinn's arbitration repeated in the register of language learning. The distance from Fez, which I traced in the decipherment essay, lies precisely in this journal: Fez's decoders needed paper, while Sennaar moved paper's entire role into the game.

When the Notebook Solves It for You — The Line Between Recording and Reasoning

Notebook design always carries the risk of overreach. When recording assistance crosses a line, the notebook begins to perform the reasoning itself. Auto-connect the clues, auto-flag the contradictions, highlight the unfilled slots—each is a kindness, but summed together they turn the player's job into digesting the notebook's instructions. Since the core of an observation puzzle is the sensation of having noticed something yourself, the moment the notebook notices first, the spine of the experience gives way.

Lorelei and the Laser Eyes (2024) is a fine example of where to draw the line. It archives every document and photograph automatically; recording is delegated completely. But between documents the game draws not a single line. Seeing which memo answers which lock is left wholly on the player's side. Storage of facts to the machine, connection of meanings to the human—that division of labor turns a mountain of paperwork into pleasure rather than pressure.

There is also the opposite road: making the notebook deliberately incomplete. The instruction manual that TUNIC (2022) has you collect is written mostly in an unreadable script, with only handwritten marginalia legible. It converts the strategy guide—external memory—into an in-game collectible, then makes that memory itself an object of decipherment. The notebook stops being where answers are stored and becomes a board in its own right: a case where the design of the recording device folds back into the design of the puzzle.

Closing

Let me sort in-game notebook design along three axes. First, how far to automate the recording of facts: La-Mulana and Her Story pushed it onto paper, while Obra Dinn and Outer Wilds let the game absorb it. Second, whether to build a home for hypotheses: Golden Idol's fill-ins and Sennaar's verification pages showed that a slot for uncertain thoughts is what decides how a deduction puzzle feels in the hand. Third, the granularity of confirmation: neither instant grading nor a single final judgment, but intermediate rhythms—three fates, one journal page—that block brute force while still returning the pleasure of a completed read.

If I were to build a deduction puzzle next, I would write the notebook's specification before the board's. What fills itself, and what stays in the player's hands—that line effectively determines the game's difficulty and its texture. And I would leave one field deliberately featureless: free text. A single line written in the player's own words supports their reasoning more firmly than any automatic record. I will leave the question with you as well: how much should your game's notebook remember on the player's behalf, and where should it politely decline to remember any further?