Agentic AI & GenAIDeep Dive120min
2026: A Spaceship Copilot Odyssey
An AI-assisted engineering escape-room workshop where players solve spaceship puzzles under time pressure using brains, search, or LLMs. It explores how AI tools affect problem-solving, compares model performance, and shares gameplay metrics and strategies to reveal where LLMs help, struggle, and boost productivity.
talk.summaryAiDisclaimer
Lucía Conde MorenoInfo Support B.V.
A failing spaceship. You, the only astronaut awake. 1 hour till the oxygen runs out. Can you save your crew, fix your ship, and figure out what happened?
You are about to play the very first AI-assisted engineering text adventure escape room (what?!), developed by yours truly. Lots of rooms to explore, items to collect and, above all, puzzles to solve: from deciphering logs to cracking passwords. Use just your brains, or a search engine, or any LLM chatbot. But remember: the tools you use might affect your problem-solving speed, and determine the fate of your crew.
Once the hour is up, we will gather everyone's player statistics, and discuss the results and findings. How did the usage of LLM coding assistants influence the solutions? Where did agents succeed, and where did they struggle? What strategies did people use to solve each puzzle? Did the model choice influence the quality of the outputs?
After this workshop, participants will leave with a fresh view on AI-assisted engineering: which kind of tasks LLM-based tools excel at or struggle with (and why), which specific models work best depending on the task, how context affects results, deterministic strategies and prompting workarounds to improve the models' performance and reduce common failures, and when these tools really do increase productivity or not (backed with actual metrics from their -and fellow players’- gameplay).
You are about to play the very first AI-assisted engineering text adventure escape room (what?!), developed by yours truly. Lots of rooms to explore, items to collect and, above all, puzzles to solve: from deciphering logs to cracking passwords. Use just your brains, or a search engine, or any LLM chatbot. But remember: the tools you use might affect your problem-solving speed, and determine the fate of your crew.
Once the hour is up, we will gather everyone's player statistics, and discuss the results and findings. How did the usage of LLM coding assistants influence the solutions? Where did agents succeed, and where did they struggle? What strategies did people use to solve each puzzle? Did the model choice influence the quality of the outputs?
After this workshop, participants will leave with a fresh view on AI-assisted engineering: which kind of tasks LLM-based tools excel at or struggle with (and why), which specific models work best depending on the task, how context affects results, deterministic strategies and prompting workarounds to improve the models' performance and reduce common failures, and when these tools really do increase productivity or not (backed with actual metrics from their -and fellow players’- gameplay).
Lucía Conde Moreno
Lucía Conde-Moreno is Head of the AI Research Center at Info Support, where she also works as a consultant software engineer specializing in data and AI applications. She is known as a Jack Of All Trades by her colleagues, having worked in varied roles ranging from .NET or Java developer to data scientist or machine learning engineer. She has worked for different national and international clients, in diverse fields such as finance, healthcare, energy, or education. She is part of the AI Champions chapter for promoting AI-augmented engineering tools, and she is responsible for supervising internal research in subfields of AI like explainability or computer vision. When she is not working, she is busy switching across random hobbies, from filmmaking to DJing. She holds a MSc in Computer Science, and a BSc in Telecommunications Engineering.