Chalked prepares the reply you would actually send from the thread on screen, your live calendar and sourced facts you have accepted before. Press Tab to insert it, hold fn to revise it, then review and send yourself.
I built Chalked because the hard part of replying usually isn’t writing a sentence. It’s reconstructing what you agreed, whether you’re free, and what this person actually needs.
Most tools either improve wording after you’ve made the decision, or pull you into another inbox. Chalked stays in the conversation you already opened: it reads a bounded slice, prepares a reply in the notch, Tab inserts it, and holding fn lets you change the intended outcome by voice. You still review and send it yourself.
The longer-term bet is that communication should become resolved work. Sourced commitments and decisions can improve the next reply—and give the AI tools you already use cleaner working context.
I’d especially value blunt feedback from agency owners, founders, consultants, recruiters and anyone whose important conversations span several apps: where does Chalked save real reconstruction, and where does it still get in the way?
Report
@atticusjxn I’ve tried similar tools before, but treating communication as resolved work is a really interesting angle. soon i'll check
Report
abt the sourced facts u accepted before, if one goes stale do i hunt it down and delete it myself or does it age out on its own
Report
Tab is the risky part. Typing a reply is slow enough that you re-check what you actually agreed to, and one keystroke removes exactly that pause. The failure mode I'd worry about isn't a bad reply, it's a good one with the wrong date in it, and that goes out faster than anything you wrote by hand. Anything it pulled from the calendar or an old thread should have to survive a glance before Tab does anything.
@asadmalik901 Hey, thanks for your feedback! Our goal is to ensure that Chalked will only give a suggested reply when it confidently knows the context behind it. That's why we're working on making the model deterministically choose which replies have enough context to come down with a reply for. It is also precisely why the fn key can be used to update the response or change a statistic around. Often times people need to correct structure and context pulled for a reply but want to make minor tweaks to the details which is where this comes in to ensure every reply is correct.
Chalked
@atticusjxn I’ve tried similar tools before, but treating communication as resolved work is a really interesting angle. soon i'll check
abt the sourced facts u accepted before, if one goes stale do i hunt it down and delete it myself or does it age out on its own
Tab is the risky part. Typing a reply is slow enough that you re-check what you actually agreed to, and one keystroke removes exactly that pause. The failure mode I'd worry about isn't a bad reply, it's a good one with the wrong date in it, and that goes out faster than anything you wrote by hand. Anything it pulled from the calendar or an old thread should have to survive a glance before Tab does anything.
Chalked
@asadmalik901 Hey, thanks for your feedback! Our goal is to ensure that Chalked will only give a suggested reply when it confidently knows the context behind it. That's why we're working on making the model deterministically choose which replies have enough context to come down with a reply for. It is also precisely why the fn key can be used to update the response or change a statistic around. Often times people need to correct structure and context pulled for a reply but want to make minor tweaks to the details which is where this comes in to ensure every reply is correct.