Former Assassin's Creed Hexe Director Shares Frustrating AI Coding Experience

A former director working on Ubisoft's forthcoming Assassin's Creed Hexe has recounted a difficult personal experience using AI tools to learn coding, describing early versions of widely available AI systems as unreliable for producing functional code.

The account adds to a growing body of evidence that AI-assisted development tools, despite rapid commercial uptake, remain uneven in their practical utility across different professional contexts — particularly in specialized fields like game development where precise systems integration matters.

Reports of this experience, published via the Telegram channel pirat_nation on 26 May 2026 at 11:00 UTC, describe how attempts to use AI for learning coding were undermined by outputs that were, in the former director's characterisation, broken or incorrect.

Assassin's Creed Hexe and the AI Development Question

Assassin's Creed Hexe represents Ubisoft's continued investment in its flagship franchise, with the title positioned as a departure from the recent open-world formula that has defined the series for the past several years. Development on Hexe has been underway for some time, with the project reportedly exploring narrative and mechanical directions that differ substantially from the recent Valhalla and Mirage entries.

The emergence of generative AI as a potential development accelerant has presented game studios with a dilemma: the technology offers obvious efficiencies for routine coding tasks, asset generation, and documentation, but its reliability for complex, interconnected systems — the kind that underpin AAA game engines — remains contested among practitioners.

Ubisoft has publicly signalled interest in AI integration across its workflow, though specifics around which tools and how deeply they have been adopted remain closely held. The company is not unique in this ambiguity; across the industry, studios have varied widely in how openly they discuss their AI practices, partly due to ongoing public scrutiny around the displacement of human creative labour.

What the Frustration Reveals About Early AI Limitations

The former director's account centres on a specific failure mode that developers have flagged since the earliest consumer AI coding assistants: the systems can produce code that looks structurally correct but fails at execution, either through logical errors, incorrect API usage, or assumptions that do not hold in real-world environments.

For a developer learning to code or attempting to integrate unfamiliar systems, this presents a particular hazard. An output that appears authoritative but does not actually work can mislead the user into believing they have understood a concept when they have not, or consume significant time debugging code that was never sound to begin with.

AI companies have addressed these early shortcomings in subsequent model iterations, with improvements in reasoning, citation accuracy, and the ability to handle complex context. The experiences of professionals who encountered these tools in their first twelve to eighteen months of broad availability now constitute a distinct cohort: their assessment of AI utility often reflects a period when the technology was genuinely less reliable than it has since become.

This temporal dimension matters for any evaluation of AI in professional settings. The same tool that frustrated a developer in 2023 may offer meaningfully different value today.

Industry Patterns and the AI Learning Curve

The games industry has not been passive about AI adoption. Studios ranging from small indie teams to major publishers have experimented with AI for code review, procedural content generation, localisation, and prototyping. Reports from within major studios suggest a bifurcated experience: some teams have found genuine productivity gains for well-defined, repetitive tasks, while others have encountered the kinds of failures that the Assassin's Creed Hexe director described.

The pattern mirrors what has played out across software development broadly. Surveys of professional developers consistently show enthusiasm tempered by caveats. AI tools are most reliably useful for tasks with clear correct answers or well-documented solutions — generating boilerplate, translating between familiar code formats, explaining unfamiliar APIs. They are least reliable for novel architectures, edge cases, and tasks that require understanding of systems-level interactions.

Game development sits somewhere in the middle: it involves both highly standardised elements (animation pipelines, UI frameworks) and deeply idiosyncratic ones (custom physics systems, proprietary rendering solutions). Whether AI tools help or hinder a particular developer depends substantially on which category their immediate work falls into.

The former Assassin's Creed Hexe director's experience points to a specific scenario — using AI to learn coding rather than to accelerate known tasks — where the limitations may be most pronounced. Learning implies encountering the foundations of a discipline, where conceptual understanding matters more than output generation. AI tools, which excel at producing rather than teaching, may be structurally mismatched to that use case.

What Comes Next

The evidence from practitioners suggests that AI tools in game development are settling into a productive but bounded role: valuable for accelerating specific workflows, but not transformative in the way early hyperbolic coverage implied. Studios that have integrated these tools most successfully appear to have done so through careful task scoping — identifying where AI outputs are verifiable and high-stakes consequences are limited.

Whether Assassin's Creed Hexe makes substantive use of AI in its shipped product remains to be seen. Ubisoft has not announced specifics. The former director's experience reflects a personal learning journey rather than a commentary on the studio's technical choices.

What the account does offer is a reminder that the narrative around AI and work is more granular than either advocates or sceptics typically allow. The technology is genuinely useful in some contexts and genuinely frustrating in others. Knowing which is which remains a skill that experience, not tools, can provide.

Desk note: Monexus treated this as a practitioner-account story rather than a technology-futures polemic. Wire coverage of the former director's experience appeared limited, reflecting a broader pattern where individual developer accounts of AI tools receive less sustained attention than announcements from the companies selling those tools.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://t.me/pirat_nation/5823