One Small Step
An AI task assistant for people who have difficulty starting tasks. It uses an agent architecture and minimal-step suggestions to help users begin tasks, can step back proactively when the user is absent, and supports a Telegram interface and an MCP server.
Timeline (last 90 days)
One Small Step is an AI task assistant app whose core design is to push reminders when the user needs them and to proactively back off when the user does not respond (asking whether to keep reminding, stay quiet, or drop the task).
One Small Step was initially a general-purpose chatbot and was later refactored into an agent architecture based on 10 tools, making a state assessment every turn of conversation and persisting memory.
One Small Step is actually deployed on GCP Cloud Run and provides its service through the Telegram interface.
Triana also built an MCP (Model Context Protocol) server so that the same set of tasks can be accessed from the cloud, the desktop, or code; this server has not yet been officially deployed, but it is already runnable.
Triana set a quality threshold of 75 points and tested One Small Step on 13 real tasks; only 9 passed, for a score of 69 points, below the threshold.
Testing revealed a production risk of conflicts between SQLite writes and cloud storage; the patch was not sufficient to solve it, so the entire persistence layer was migrated to Firestore, which is now live.
Triana's mentor advised her that a tool built for everyone will ultimately likely be used by no one, including the builder herself; so she narrowed the scope to focus on the specific pain point of task paralysis.
One Small Step provides a dashboard that centrally displays incomplete tasks, streak records, and upcoming deadlines; users can add goals and tasks, or remove completed items and ones they do not want to keep.
One Small Step currently has only one real user (who is also the tester); this user likes the dashboard but reported that the app does not know the user's name.
All Facts (9)
One Small Step currently has only one real user (who is also the tester); this user likes the dashboard but reported that the app does not know the user's name.
50%UnverifiedOne Small Step was initially a general-purpose chatbot and was later refactored into an agent architecture based on 10 tools, making a state assessment every turn of conversation and persisting memory.
50%UnverifiedOne Small Step is actually deployed on GCP Cloud Run and provides its service through the Telegram interface.
50%UnverifiedTriana also built an MCP (Model Context Protocol) server so that the same set of tasks can be accessed from the cloud, the desktop, or code; this server has not yet been officially deployed, but it is already runnable.
50%UnverifiedTriana set a quality threshold of 75 points and tested One Small Step on 13 real tasks; only 9 passed, for a score of 69 points, below the threshold.
50%UnverifiedOne Small Step is an AI task assistant app whose core design is to push reminders when the user needs them and to proactively back off when the user does not respond (asking whether to keep reminding, stay quiet, or drop the task).
50%UnverifiedTriana's mentor advised her that a tool built for everyone will ultimately likely be used by no one, including the builder herself; so she narrowed the scope to focus on the specific pain point of task paralysis.
50%UnverifiedOne Small Step provides a dashboard that centrally displays incomplete tasks, streak records, and upcoming deadlines; users can add goals and tasks, or remove completed items and ones they do not want to keep.
50%UnverifiedTesting revealed a production risk of conflicts between SQLite writes and cloud storage; the patch was not sufficient to solve it, so the entire persistence layer was migrated to Firestore, which is now live.
50%