Workflows, comparisons, and engineering notes from building Inboxer.
Outlook users were second-class citizens during the first wave of AI inbox tools. That gap has mostly closed — here are the five options worth knowing about, organised by what they actually do.
Read more →Fully-automated AI email sounds great until the model ships something you wouldn't have. Why every credible product in 2026 keeps a human approval click between the draft and the send.
Read more →Transcription is solved. Extraction — turning the transcript into useful, non-bloated tasks — is where most meeting-AI tools still fall short. Here's what good extraction actually looks like.
Read more →AI drafts have a recognisability problem. Voice matching is the technical answer — fewer 'I hope this finds you well' openers, more drafts indistinguishable from the email you'd have written.
Read more →Every 'AI email' product claims smart triage. Almost none explain the mechanic. Here's the unmagical breakdown: how the classifier decides, what categories matter, and the failure modes nobody talks about.
Read more →Most knowledge workers tolerate inbox dysfunction the way we tolerate slow wifi: by adapting around it. Here are the five most reliable signs you've crossed the threshold where a tool starts paying for itself.
Read more →A practical workflow for keeping any inbox at zero — triage, drafts, and meeting follow-ups, all driven by AI but kept under human control.
Read more →Building Inboxer for both Gmail and Outlook surfaced real differences in how the two providers expose their APIs. Here's what matters for users.
Read more →Inboxer's most common user is a founder drowning in 200+ unread emails. Here's the exact playbook the busiest of them follow to get to zero.
Read more →