I Turned Coding Agents Into a Strategy Game — Ido Salomon, AgentCraft
Source
youtube.com
Author
AI Engineer
Date
Context
AgentCraft applies a strategy-game interface to the problem of supervising multiple coding agents. Its creator, Ido Salomon, describes agents as units on a map, with the file system providing the terrain and heat maps showing where work is happening. The space bar jumps to work needing attentionThe mechanism that lets a model weigh which earlier words matter for the word it's currently processing — the core operation of a transformer.Full definition →, according to the presentation description.
The intended benefit is easier supervision: the interface makes activity and requests for attention visible in one spatial view. This addresses the human effort of directing and reviewing agents. The presentation supports that interface concept, without establishing a measured improvement in how many agents someone can manage.
Key takeaways · AI-distilled
Salomon argues the limit on running many agents is human capacity to steer, direct and review them, not model capability, and that strategy games like StarCraft already taught those skills.
AgentCraft draws agents as units on a map where the file system is the terrain, uses heat maps to show where work is happening, and binds the space bar to jump to whatever needs attention.
For autonomy it adds orchestrators, loops and a review kit that presents visual evidence of what agents did; for collaboration, shared rooms put people and agents in the same workspace.
An early, simpler mobile-game-style mode targets the roughly 90% of people Salomon says are not yet AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → power users.
Terms in this piece · Glossary
attention — The mechanism that lets a model weigh which earlier words matter for the word it's currently processing — the core operation of a transformer.
multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
videoRethinking Claude Code Best Practices — Lydia Hallie, Anthropic
videoStop Shipping LLM Settings That Only Win on Paper - Rohan Pandey & Balaji Varadarajan, DigitalOcean
videoYour SKILL.md Is 70% Stuff the Model Already Knows - Nikolay Petrov, Intel
videoYour Spec Has Bugs. Catch Them Before Your Agent Codes. - Rajath Mahendrappa Kotyal, AWS
