Halo is the new Doom and most software will soon be effectively open source (by force or by choice)

Seven months ago I published “Source code is now a common good, and SaaS is mostly dead.” I argued that AI had collapsed the economic cost of clean-room reimplementation, that pure software expression was losing its scarcity, and that the SaaS model built on high-margin code would face an existential squeeze—except where real data moats or network effects remained.

I was wrong about the timeline.

It didn’t take five to ten years.

This month the premonition became visible reality.

Halo everywhere — including 100+ player co-op on PC

In the space of a few days we watched Halo: Combat Evolved appear on platforms it was never meant for, and get radically expanded where it already existed:

• Halo 64 (Jack Tulli) — a playable multiplayer (and emerging campaign) port running on real Nintendo 64 hardware. Built on the halo-ce-universal decompilation, with a custom renderer, four-player split-screen, the original 13 maps, bots, classic modes, and even a Firefight mode the original Xbox version never had. Claude assisted heavily in the coding.
• Halo: Dreamcast Evolved — extended gameplay of the Pillar of Autumn and other levels running on actual Dreamcast hardware. The CPU is more than three times slower and the system has a quarter of the RAM of the original Xbox, so performance is still rough, but the core systems are already ported and being optimized in public.
• Native PC (and beyond) with massively expanded player counts — the same decompilation work that enabled the console ports also produced native Windows, Linux, and Android builds (OpenCE / halo-ce-universal and related projects). These support system-link / online multiplayer for up to 128 players. Campaign co-op limits have been raised as well; 32-player campaign co-op is already being demonstrated, with work continuing toward the same high player counts. Browser versions of the original Xbox game have also appeared with 128-player rooms and split-screen co-op.

These are not asset flips or simple ports. They rest on prior decompilation work that reached near byte-matching fidelity. What changed is the speed of the remaining work. A single developer (or small team) plus modern agents can now finish the last mile—new renderer, platform integration, expanded netcode, extra modes—in days or weeks instead of years. The result is a franchise that once defined the Xbox suddenly running on N64, Dreamcast, modern PCs, phones, and browsers, with player counts the original hardware could never support.

Halo is becoming the new Doom: the game that runs on everything, because the barriers that used to stop it have collapsed.

Demon's Souls Remake: from ~1 FPS to 60+ (and nearly 120) in roughly a month

PS5 emulation provided the other high-visibility stress test.

In early September, Demon's Souls Remake on KyTyPS5 and its forks was crawling at 1–2 FPS even on top-tier hardware after finally reaching in-game states. Within days it jumped into the 10–18 FPS range. By mid-September temporary 30 FPS markers appeared. By mid-October, custom builds were delivering over 60 FPS at 1440p with original PS5 graphics settings on an RTX 5070 + Ryzen 7 9800X3D, and approaching 120 FPS with DLSS Frame Generation. The full game had already become completable earlier in the month.

The historical pace of console emulation does not look like this.

REA — Reverse Engineer Anything

While the ports and the emulator were making noise, a different tool exploded: REA (morluto/rea).

An MCP server and CLI that lets coding agents (Claude Code, Codex, Cursor, and others) reverse-engineer applications, native binaries, websites, and runtime behavior without source code. It orchestrates existing tools (Ghidra, Hopper, IDA, browser instrumentation) and returns evidence-backed explanations. In roughly two weeks it accumulated hundreds of thousands of stars.

The practical effect is that an agent can now be pointed at a feature and produce a usable understanding plus a path to independent reimplementation. Combined with the clean-room pattern I described in March, the barrier between “I can observe the behavior” and “I can recreate the behavior under a different license or on different hardware” continues to fall.

The Adobe and Office “Craft” ports — imperfect, but the signal is loud

Early October also brought the ArtCraft projects (led by Brandon Thomas): open-source, Rust-based applications described as clean-room reimplementations of major Adobe tools — PhotoCraft (Photoshop), FilmCraft (Premiere), LightCraft (Lightroom), and equivalents for Illustrator, After Effects, InDesign, and Acrobat. Built with heavy assistance from Claude.

A few days later the same effort expanded to Microsoft Office: WordCraft, GridCraft (Excel), and DeckCraft (PowerPoint).

None of these are finished products. The teams themselves rate them as early alpha—feature parity and “ready for real work” numbers typically sit in the 45–60% range depending on the metric and app. File compatibility, edge cases, and polish remain incomplete.

They are imperfect. That is exactly why they matter as signals.

These are public demonstrations that a small team + frontier models can now produce recognizable, runnable, file-compatible approximations of two of the most heavily protected, highest-margin software suites in existence in a matter of weeks. The UI patterns, core workflows, and file formats that previously required years of specialized effort are being reconstructed rapidly enough that the projects are already usable for lighter tasks and improving in public view.

What these stories share

They are not isolated stunts. They are symptoms of the same underlying shift I described in March:

• Decompilation and reverse-engineering tooling has matured.
• Frontier models are competent enough at low-level and systems code to accelerate the remaining human work dramatically.
• Agentic workflows (and tools like REA) make the process more systematic and less dependent on rare specialist knowledge.
• The economic cost of “make a working version of X on platform Y” or “understand how Z is implemented so I can rebuild it” has collapsed.

Where is this going?

In the original article I concluded that most software would become effectively open source—by force or by choice—and that the only sustainable thing left to charge for would be convenience. I was thinking primarily about SaaS. I had not fully considered how the same forces would hit other markets.

Now that I am watching it play out in games, I think the game industry is even more exposed.

AAA studios were already struggling with rising costs, risk-averse sequels, live-service fatigue, and audience fragmentation. The ability to rapidly reimplement, port, expand, and redistribute games—especially older or mid-tier titles—removes one of the remaining economic moats: the difficulty of making a working version exist somewhere else. When a decompiled classic can run on N64, Dreamcast, phones, browsers, and PCs with 128-player lobbies within weeks of the tooling maturing, the value of exclusive platform deals, long development cycles, and even some forms of DRM starts to look fragile.

Even games that embrace community content and modding are not automatically safe. Modding communities have historically extended the life of games; the new tools accelerate both the creation of those mods and the independent recreation of the underlying experience itself. The distinction between “official” and “community” versions becomes harder to maintain when the technical barrier to a full alternative keeps falling.

The original prediction still holds, only broader:

• Most software will become effectively open source—by force (because protection becomes impractical) or by choice (because fighting it is more expensive than adapting).
• The residual value will concentrate in convenience, operations, trust, data network effects, real-time services, and experiences that cannot be trivially recreated offline or offline-first.
• Pure expression of code, whether in a SaaS dashboard or a single-player campaign, loses scarcity.

For games this may prove especially brutal. The industry already operates on high fixed costs and hit-driven economics. If the cost of independent recreation and redistribution keeps collapsing, the pressure on traditional publishing and exclusivity models will intensify. Some studios will adapt by doubling down on live services, social layers, competitive integrity, and continuous content that is hard to mirror. Others will not.

Seven months ago I wrote that source code was becoming a common good and that SaaS was mostly dead.

This month Halo became the new Doom, emulators jumped from 1 FPS to 100, reverse-engineering agents went mainstream, and clean-room approximations of Adobe and Office appeared in public.

The direction is clearer than it was in March.

Most software will soon be effectively open source—by force or by choice.

The only remaining question is how many markets realize it in time to adapt.