Tools like Slack and Teams were built for people, while tools like Cowork and OpenClaw were built for working with Al.
Lightbulb brings those two worlds together in one shared workspace. Talk in channels, bring agents into the conversation, delegate work, and review important actions before they happen.
An AI agent needs two different kinds of context:
A skill says how to do the work.
An automation says when to start it.
Keeping them separate lets you improve the workflow without rebuilding the trigger. Setup:
https://t.co/59LGhWnf04
@YubaX01 Hey Yassine. We’re Lightbulb. We bridge the gap between AI work (Cowork and OpenClaw) and communication (Teams and Slack). Would love to connect.
@iAnujVarshney Hey Anuj. Full stack engineers currently use one platform for communication and another for their AI work. Lightbulb brings both into one shared channel. Would love to work together. https://t.co/SGxapsMBUc
Exactly. The useful promise isn’t “replace the founder.” It’s “give the founder fewer repetitive things to hold in their head.” I’d keep customer calls, product judgment, and hiring human, then use agents for prep, follow-up, and keeping the work visible. What would you delegate first?
@39oorr This tracks. The sneaky transition seems to be $25k–$50k, when the founder stops being the fastest person to do every task but the team still depends on them for every decision. What handoff do your members struggle with first: support, growth, or product?
Having the same agent span onboarding, campaign work, and billing is interesting. The hard part is the handoff when it stops being confident. Are you keeping the conversation and account context intact when a person takes over? We’re working on that boundary too and would be happy to compare notes.
Debug production, yes, but read-only first. I’d let an agent gather logs, reproduce the issue, and propose a fix. Deploys, DB writes, scaling changes, and anything that can delete data should need approval with a clear diff and rollback plan. Curious where you’re drawing the line in AtherNull.
@euboid@ferndeskhq The interesting part is Fern checking each claim against the product, not just rewriting docs. How are you handling changes that are technically true but still need a support or product owner to approve the wording before they go live?