Agent-hours outnumber human hours at Anthropic, and entry-level delivery labour is cut first

The protection idea is already contested. Thales, Digital.ai, Guardsquare and Jscrambler all ran AI-resistant protection campaigns in 2026, and Thales already cites tests against AI agents. What none of them offers is an independent, repeatable score of how long a frontier-model agent takes to rebuild working source. That would mean a standing red-team harness for native desktop and game binaries, rerun as models improve, with results published or certified. A new entrant could win as a neutral benchmark and certification layer, the way AV-TEST works for antivirus, rather than as yet another obfuscator. That gives it a chance to partner with or rate the incumbents instead of competing with them on core obfuscation, where they have years of tooling lead. The risk is that the research consensus treats obfuscation only as a cost multiplier: if the scores keep falling with each model release, the result undermines the product being sold. The authorized port studio looks earlier and more open. I found no funded company offering agent-driven ports of publisher catalogs. The demand signal is real: in 2026, unofficial AI-generated ports of Smash Bros, Banjo, GoldenEye and Halo showed one developer can ship a playable port in days. Incumbents like Nightdive and traditional porting houses still work by hand and charge accordingly. The opening is fast, cheap, rights-cleared ports of mid-tier catalog titles to web and handheld platforms. The constraints are publisher wariness about AI after fan backlash against "AI slop" ports, the need for original source or legal clean-room processes, and the possibility that publishers or Nightdive-style studios adopt the same agents themselves once it is proven.

• Thales Sentinel Envelope Plus

  A binary protection add-on for desktop and on-prem ISV software, marketed as defending against AI-assisted reverse engineering. Thales's own marketing says an AI agent found 8 of 10 vulnerabilities in an unprotected app and none in a protected one, and its product page describes a test using Claude. This is the closest existing match to model-tested hardening for ISVs.

• Digital.ai Quick Protect Agent v2

  LLM-enhanced post-build app hardening that runs in CI/CD, launched March 2026. It is aimed at mobile apps (Android and iOS) and is framed around AI speeding up reverse engineering and app cloning.

• Guardsquare polymorphic protection

  Changes the obfuscation and encryption on every build to make AI-scale decompilation more expensive. Mobile-focused.

• Jscrambler LLM-resilient code protection

  Polymorphic JavaScript obfuscation sold specifically as resisting analysis by LLMs. Covers client-side web code only, not native binaries.

• morgana.labs

  A small startup claiming AI can tear down protected apps in under an hour. It proposes binding keys to app integrity. Its claims are unverified.

• pyobfus LLM-Deobfuscation-Resistance Benchmark

  An open-source Python obfuscator that publishes a benchmark of how well its obfuscation resists LLM deobfuscation. This is an early example of the benchmarking idea, but it covers Python, not native binaries.

• Quarkslab (research)

  An established obfuscation vendor that has published work on defeating AI-assisted reverse engineering. It argues obfuscation is a cost multiplier rather than real security.

• Unofficial AI-generated N64 PC ports (BattleShip Smash Bros etc.)

  Unofficial ports of classic console games built with AI by hobbyists, such as a Super Smash Bros port its author says is 100% AI-generated. These are the unofficial ports the port studio idea would be racing to beat. Human-led decomp communities are pushing back against them.

• Nightdive Studios

  An incumbent licensed remaster studio. It reverse-engineers titles whose source code is lost (System Shock 2, Blade Runner) and runs them on its own KEX engine. It works mostly by hand rather than with agents.

• Traditional porting houses (Sickhead Games, Juego Studio, Viridian)

  Studios that port legacy codebases and custom engines to modern platforms for publishers. None that I found markets an agent-driven workflow.

• Pixel Research

  An AI remastering startup that worked with Revolution Software. Its work focuses on upscaling art assets, not code ports.

Checked 2026-10-11