Hey HN, I built Tickr because I was tired of being the human reminder system on my team. Every project I've been on has the same failure mode: someone creates tasks in JIRA, nobody updates them, the PM spends half their day chasing people for status, and standups become 15 minutes of "I'll update the ticket after this meeting."

Tickr is a Slack bot that does the project management work that humans hate doing:

- Nudge engine: Automatically follows up with assignees when tasks go stale. It factors in priority, time estimates, grace periods, blockers, and snooze state, so it's not just a dumb cron job pinging people. - AI standup generation: Synthesizes recent task updates into a daily standup summary. No more meetings where people read tickets aloud. - Update quality evaluation: When someone posts a vague update like "working on it," Tickr pushes back and asks for specifics. Uses Claude to judge whether an update actually communicates progress. - Slip detection: Scores tasks on 6 signals (staleness, estimate overrun, update quality trend, blocker chains, priority, velocity) to predict which tasks are about to slip before they do. - Thread-to-task extraction: Mention @Tickr in any Slack thread and it parses the conversation into a structured task with assignee, priority, and estimate.

The whole thing runs on Slack's native UI, no context switching to a separate app. Tasks live in the channel where the work is discussed.

Tech stack: Python, Slack Bolt, Claude via AWS Bedrock (tool use for intent parsing), DynamoDB, ECS Fargate. The AI layer uses Bedrock's Converse API with tool definitions, the model returns structured tool calls like create_task or update_task, which get dispatched to handlers. No prompt-and-pray; it's constrained tool use.

What's different from other Slack task bots: Most Slack task tools (Asana integration, Trello power-ups, etc.) are just notification bridges, they tell you something happened in another app. Tickr is the app. The task lifecycle happens entirely in Slack. And the AI isn't a chatbot you talk to; it's an autonomous agent that nudges, evaluates, and detects problems without being asked.

Honest limitations: - No Gantt charts, no sprint boards, no 47-field ticket templates. If you need that complexity, JIRA is probably right for you. - AI quality evaluation occasionally misjudges technical updates as vague (calibrating)

Free 30-day trial, no credit card: https://heytickr.com

I'd love feedback from anyone who's dealt with the "nobody updates their tasks" problem. What did you try? What worked?