FIVE YEARS ago today, we released Heaven's Vault, which was designed to be our follow-up to our previous narrative game, 80 Days.

The game started from a core idea: SPACE ARCHAEOLOGY, that we felt we hadn't seen before; and translation as a mechanic, which we had seen in the 1980s Infocom game Infidel, and in two late 90s indie IF games, The Gostak and The Edifice.

The translation prototype was the first thing we built. We went through about nine iterations which were all dire - too easy, too random, too arbitrary - until we found one with enough internal logic to be compelling. We talked about that process at EGX just before release.

For the story, we wanted an adventure with no fail states, no back-tracking, little inventory management, and characters with memories who cared what you did - and when - and how.

We talked about our dialogue design ethos at Adventure X a year before release, with a talk that went a little viral with game writers.

In terms of the structure, we wanted to improve on the model established in 80 Days. That game was designed around fast play with lots of freedom - and lots of consequence: going this way *always* meant not going that way.

For Heaven's Vault we wanted to fold that design into something more linear. That meant creating fewer "episodes" but allowing the player to reach them in different orders, and changing what each meant based on that - with a robust "knowledge model" to track what the characters knew / believed with high fidelity.

This knowledge model paid huge dividends - it allowed us to produce a vast array of conversational options, all gated by when they were relevant, so there's always a snippet to suit the particular situation you find yourself in. It also allowed us to do the Story So Far feature... when you relaunch the game, it gives a single page, detailed summary of your adventure to date. (We think it inspired the similar feature at the start of Return to Monkey Island, but that's not been confirmed to us.)

It's also the same knowledge model is what makes the characters in Overboard! work - we've used it on every game since. In fact, we designed a whole feature (LISTs) in our scripting language, ink, just to make this work.

This model also meant we could hit a design goal - supporting incorrect theories. Most games won't accept an answer to a puzzle until it's correct, because the consequences of a mistake can be far-reaching, and can mess things up for the player and the story further down the line.

We wanted to embrace mistakes, to capture the reality of archaeology as a "story-telling science". We wanted people to have answers they were sure of, ones they knew were wrong, and some in between, and preferably have them for the whole duration of the game.

Since launch we've had a lot of feedback from practising archaeologists: we were even invited to deliver a seminar about the game at Cambridge University[arch.cam.ac.uk] and have seen several articles and papers about the game's depiction of the science.

As a production, Heaven's Vault started risky and got riskier. We began the game during the indie boom, and the "indiepocalypse" - when indie studios saw sales and press coverage collapse in favour of larger, bigger budget titles - happened mid-development. We entered 2019 with <2000 wishlists but were saved by a "puzzle" post on IMGUR. The cards had been originally produced for handing out at a Rezzed event where they... didn't move the needle at all. We collected them into one post a few weeks before release - and our Discord blew up, with thousands of posts of theories for the "stone tablet challenge".

One fan later got their own back by creating Vaultle[alek2ander.github.io], a World-in-ancient playable online.

For the art, we wanted 2D characters with lots of expression and character. Aliya and Six were designed by Jeremy de la Gaza, and we then hired Anastasia Wyatt to fill out the 1000s of frames - she's drawn all our character art in all our games since. This is Jeremy's concept...

... and Annie's first test sketch that got her the job:

For the environment art we hired a fantastic team led by Laura Dilloway, who also drove the architectural design of the world. She started by building out a "style history" of the Nebula spanning the millennia of game time (which is why you can date sites by their doorways.)

Making and designing the Nebula itself - the rivers in space - was a colossal technical challenge. Streaming assets in Unity was glitchy, so the whole scene is loaded at once. Clouds are rendered once and then projected across the level.

Sailing was and is controversial - after release we added a Fast Travel option that we'd debated a lot internally before launch. Ultimately, we wanted sailing to be like riding your horse in Shadow of the Colossus - easy to navigate, but atmospheric and with a sense of scale.

The "rivers" concept came partly from the Egyptian Book of the Dead - a quote from which opens the game - but also from the Scottish landscape, and rivers winding through valleys. That same visual concept was behind the design of A Highland Song.

Another feature big on scale was the timeline - that could expand to show minute by minute discoveries, and compress to show thousands of years worth of history, and all smoothly animated.

The game's New Game+ mode - where you restart with the language confirmed in your last game, leading to more complex translations each run - was fluke. Translations were picked at runtime based on what you know anyway... so we just commented out the code to reset the dictionary. Then we just wrote lots of alternative lines for every item. The opening "brooch" translation can be, by the eighth or ninth game, a whole screenful of text! But with the extra detail comes extra interpretation, lore, and understanding of the world - more dialogue for the knowledge system to deliver.

The music was a huge part of the feel. Initially, Laurence Chapman tried middle-Eastern instrumentation but the result felt tired (there's similar instrumentation in the new Dune movies, btw). Instead, he used a cello to create the resonate theme.

Throughout our fans have been the best. We've loved every piece of fan art, every wild theory, every argument on Discord over how to write "gecko" in Ancient. Without their encouragement, we'd never have written the novels. We've even received artefacts and books of our own. We were also surprised to be an answer on University Challenge last year, one of the UKs longest running TV quiz shows. (80 Days had previously been an answer on Jeopardy!)

Overall the game took 5 years to make - from initial designs in 2014, to shipping in 2019 - with another year to port to Switch, and another to write the novels. It's our biggest and most detailed game to date. If you've been part of the journey - thank you! And may the waters bring you home.