Adding A Feature: Then And Now

tim-cain game-development production pipelines development-costs fallout the-outer-worlds

Abstract

Problem: Why does adding a single feature to a modern game cost so much more time and money than it did in the 1990s?

Approach: Tim Cain walks through a concrete example — adding a simple "go find an NPC with a cool gun" quest — comparing how it was done in Fallout (mid-90s) versus The Outer Worlds (late 2010s).

Findings: The core reason is the shift from one-person implementation to multi-person pipelines. This introduces four compounding cost multipliers: communication overhead, scheduling offsets, longer individual stages, and more complex bug fixing.

Key insight: The explosion in development costs isn't about laziness or waste — it's a structural consequence of pipelines replacing single-person workflows, where every stage now requires more people, more coordination, and more time.

Source: https://www.youtube.com/watch?v=0lIU5ulehbM

The One-Sentence Explanation

Back in the day, one person did everything. Now it's a pipeline. That's the entire explanation at a high level — pipelines make things more expensive and longer.

The Fallout Example: One Person, A Few Hours

Tim describes watching someone add a quest to Fallout in the mid-90s. The quest: an NPC on one map tells you about another NPC on a different map who has a cool gun for sale.

What it took

  • First NPC: Pick an appearance from existing sprites, plop them on a street. Write dialogue in a preset format — maybe an hour or two total for the character and conversation.
  • Second NPC: Load the other map, place them in an alley. Write dialogue with barter checks (does he like you? how much money are you offering?).
  • The gun: Grab an existing weapon model, rename it, tweak the stats in an editor ("fires faster" or "holds 20% more bullets"). Put it in a locked chest, create a key, throw the key in the NPC's inventory.
  • Done. A few hours of work by one person.

The systemic beauty

Without anyone writing additional code, players could: hear about the gun from NPC #1 or stumble onto NPC #2 organically, buy the gun through the existing barter system, kill the NPC and loot the key, or pickpocket the key and steal the gun. All systemic — no special scripting needed.

The Modern Game Version: A Pipeline of People

The same quest in a modern first-person game like The Outer Worlds explodes in complexity:

Face and body

Every NPC needs a unique face (first-person means the player sees it up close). Body, armor pieces, and visual variety all matter because players notice reuse at high resolution and close camera distance.

Lighting and camera

Where the NPC stands now requires lighting setup. Day/night cycles mean checking multiple conditions. The conversation camera needs to be positioned correctly — no clipping through counters, bars, or walls.

Voice over

Modern games voice all dialogue. Writing for voice is a completely different skill than writing dialogue to be read — Tim notes that Fallout had to rewrite characters when they discovered those NPCs would get clay heads and spoken lines.

Localization

Dialogue gets sent for translation into many languages. Once it's in the localization pipeline, changes become expensive. In Fallout, changing "alley" to "back room" was a trivial text edit. In a modern game, that change means re-recording VO and re-translating across all languages.

Memory and performance

Additional NPCs, face parts, models, and textures all consume memory and CPU. Optimization teams may come back and say "too many NPCs in this area" or "too many unique face parts loaded."

The Four Cost Multipliers of Pipelines

Tim breaks down why pipelines specifically make everything more expensive:

Communication overhead

When one person does everything, there's zero communication cost. With a pipeline, every person needs to understand the goals, intent, and constraints: Can this NPC be killed? Are they part of a faction? Should prices change based on reputation? This requires documentation, meetings, or conversations — all of which take time.

Scheduling offsets

When person A finishes their pipeline stage and hands off to person B, person B may not be available. They have other work. The half-finished task sits in a queue. This requires producers to track — and where the 90s needed one producer, modern teams need four, five, or six just to manage the handoff queue. You may also need a specific person with a specialization, not just anyone.

Longer individual stages

Every stage in the pipeline takes longer than it used to. A weapon in Fallout was a ground sprite, an inventory icon, and a wielded sprite — often reusable across items. In a modern game, you need a new 3D ground model, a held model, an inventory image, all with correct lighting. Low resolution allowed massive reuse; high resolution and close cameras make reuse obvious and unacceptable.

Bug fixing

More complexity means more bugs. More bugs means more time and money. Worse, in a pipeline, finding who caused the bug is itself a task — you may need a programmer (who had nothing to do with the quest) just to diagnose which pipeline stage broke. The person who needs to fix it might have moved to another project, be on vacation, or have left the company entirely. Longer development timelines make all of these more likely.

Tim's Conclusion

He's not defending the cost increases or justifying them — just explaining the structural reality. You could make games like they did in the 90s, but the retro market isn't big enough to sustain most studios. The pipeline is a consequence of modern player expectations for visual fidelity, voice acting, and localization — and pipelines are inherently more expensive than single-person workflows.

References