Mom
9:41landed, waiting on the bags. call you in 20
Product preview · Made for iPhone
Chicken Wire is being built to use AI to sort campaign blasts, loan pitches, and obvious scams in Messages.
Try the preview below. The iPhone app is in development.
“Pre-approved: up to $25,000 deposited as soon as tomorrow. No credit check…”
landed, waiting on the bags. call you in 20
sending the lease over tonight — one last read?
Your verification code is 408912.
4 new · campaign fundraising, coverage offers, loan pitches
Filtered2 new · prize fee, courier impersonation
Filtered“You won the $1,500 gift card drawing. Pay the $4.95 processing fee…”
“We declined a $312.40 charge because it did not match your usual activity.”
What gets sorted
Sales pitches can go to Promotions. Clear scams can go to Junk. Messages that could matter to you pass through the filter.
Campaign fundraising and mass canvassing, judged the same way across candidates and parties. Ballot receipts and polling-place notices are not solicitation and pass through.
Sales outreach for new plans and Medicare lead generation. Your claims, renewals, and pharmacy notices are a different thing entirely, and they are left alone.
New-credit solicitation: pre-approvals, cash advances, debt consolidation. Updates on an application you started can pass through.
The sitewide-sale blast, the subscription upsell, the signals group promising to double your money. Clear pitches, no personal context.
A prize behind a processing fee. A delivery text asking for your card details. Only the unambiguous patterns, because Junk silences the alert.
Codes, deliveries, fraud alerts, appointments, school and election notices, replies from people, and anything ambiguous. Uncertainty is not a reason to hide a text.
Illustrative demo
Pick an example text. See where it would land and what clues led there, from a campaign blast to a bank alert worth keeping.
An illustrative demo. Every message here is invented, and each outcome is the one the prototype's own rules produce — not a live model call.
politicalpromotion — Sorted into Promotions. Alerts follow your Messages settings.FINAL NOTICE: Our campaign is 412 donors short of the midnight deadline. Chip in $25 now to triple-match your gift: [link] Reply STOP to opt outHow it works
A loan pitch and a loan payment receipt share a lot of words. Chicken Wire asks AI to tell the difference.
Messages hands off eligible SMS and MMS from unknown senders. Your conversations stay in Messages.
With your opt-in, iOS sends the eligible text to our service. We drop the sender and screen the text for sensitive details before asking OpenAI what it is about.
Only confident results move a text. If the service is offline or the answer is unclear, Chicken Wire leaves the message to iOS.
iPhone setup
Here's how setup works when you have a build on your iPhone.
Open Chicken Wire, read what cloud filtering shares, and opt in when you're ready.
In Messages, open Filters → Manage Filtering → Text Message Filter and choose Chicken Wire.
Look under Filters in Messages. You can turn off Chicken Wire in the app or change your filter in Messages.
Questions
The iPhone app is in development. You can try the demo here today.
Eligible SMS and MMS from unknown senders. iMessage, saved contacts, and senders you have replied to at least three times are outside this filter’s scope.
iOS hands the filter one eligible incoming text at a time. Your history and outgoing texts are outside its reach.
OpenAI receives screened message text and classification questions. Our service discards the sender and masks detected links and phone numbers. Texts flagged by our privacy checks skip the model call.
The filter is designed to sort new eligible texts into Promotions or Junk. You manage deletion and blocked numbers in Messages.
Look under Filters in Messages for Promotions or the spam folder. Promotion alerts follow your Messages settings.
Open Filters in Messages to find it. If iOS offers a recovery action for that folder, use it to move the message back. You can add a sender to Contacts to keep future texts from them outside the filter.
The demo illustrates expected behavior. We are still testing how often the filter catches spam and how often it gets a legitimate text wrong.