I Vibe Coded an App with Opus 5 to Master The Finals

According to software writer Ruben Circelli, building custom AI productivity tools using Anthropic’s Opus 5 model offers significant efficiency gains for competitive gaming preparation, though developers must navigate strict platform token limits and agent limitations compared to competing systems like OpenAI’s GPT-5.6.

Building FISH: The Finals Interactive Study Helper with Opus 5

Competitive first-person shooters like The Finals require fast, dynamic decision-making that traditionally takes hundreds of hours to master, according to Circelli. To speed up this process, Circelli built the Finals Interactive Study Helper (FISH) application using Anthropic’s Opus 5 model on Claude’s Max 5x plan ($100 per month). FISH is not a cheat tool; it operates completely independently of the game files. Instead, it generates random game states—including player states, active gadgets, respawn coins, and team compositions—to test tactical responses and drill users on in-game specializations, maps, and modifiers.

Opus 5 Performance: Efficiency Gains and Limit Drains

While Opus 5 delivers strong reasoning capabilities, its implementation of autonomous agents presents distinct hurdles for developers. Circelli notes that Opus 5 runs significantly faster than previous versions, allowing for more prompts before hitting five-hour usage caps. However, Claude’s default behavior of spinning up 10 to 20 agents simultaneously can exhaust this window in minutes. By comparison, ChatGPT typically defaults to a three-agent maximum and relies on weekly limits.

Furthermore, Opus’s agents struggled with basic web-scraping tasks, failing to extract game mechanics directly from official wikis or web pages without manual link and quote provisioning—an issue absent when using GPT-5.6 agents, according to Circelli. Opus 5 also frequently ignored memory instructions regarding conciseness and context-window tracking.

Feeding YouTube Guides to Create Pro Strategy Drills

To move beyond basic rules, Circelli used Opus 5 to ingest and analyze hundreds of transcripts from experienced YouTube creators. Across dozens of agents and millions of tokens, the model distilled these transcripts into a comprehensive strategy bible. This data enabled FISH to provide nuanced, advanced advice alongside standard rulebook answers. For example, while basic rules dictate pushing a late-round cashout, FISH’s pro strategy advice factoring in teammate loadouts and enemy gateway gadgets generated specific tactical instructions, such as deploying a light class member with glitch grenades.

Comparing Opus 5 and GPT-5.6 for Coding Projects

Previous generations of Anthropic models like Opus 4.6 fell short in direct comparisons against OpenAI’s GPT-5.6. However, Opus 5 successfully closes that intelligence gap while offering superior efficiency for specific workflows. While Circelli ultimately found that GPT-5.6 still produces the best overall results for complex software projects, Opus 5 remains a viable alternative for developers seeking competitive reasoning capabilities.

Frequently Asked Questions

What is FISH (The Finals Interactive Study Helper)?

FISH is a custom study and training application built with Anthropic’s Opus 5 model designed to help players of The Finals memorize game mechanics and practice tactical decision-making through randomized simulations.

Does FISH interact with game files?

No. FISH runs entirely separately from The Finals game files and does not function as a cheat or mod.

How does Opus 5 compare to GPT-5.6 for coding?

According to developer testing by Ruben Circelli, GPT-5.6 generally delivers superior results for complex coding tasks, but Opus 5 matches competitive intelligence levels while offering notable efficiency gains.

What do you think about using AI-driven simulation apps to train for competitive video games? Leave a comment below, explore our latest tech and software reviews, and subscribe to our newsletter for more developer insights.