A weird take. They seem oblivious to the fact that real problem is not distrust of flock, it’s distrust of the police. They are in a situation where the public just barely trusts the local police, and a new surveillance system pushes them over the edge into hostility. They need to be working to gain much more trust.
Police are bounded by evidentiary laws and warrant requirements, all of which legitimately do stand up in court. Flock is not.
So no, actually Flock itself is the problem. If Flock were willing to sign some consent decree that its data can only be accessed by police with a valid judicial warrant, for example, they probably could've headed off this entire mess.
But instead they called people "terroristic" for pushing back. Completely braindead move from the CEO.
I posted this in another thread, and had my comment instantly flagged. Love this site.
Worth looking at how these cameras got approved in the first place. My county's ALPR interlocal passed on the consent agenda in December 2024, one block vote alongside forty other items, no discussion, and went through consent again in April 2025. Nothing was "hidden", it's just that "public" here means a 300 page agenda packet and a four hour YouTube video with no transcript, completely unreadable in practice.
So I spent a week and a half building a pipeline over what the county already publishes. It pulls the CivicPlus portal (agendas, minutes, item metadata, outcomes) and the YouTube channel, does ASR plus diarization, then aligns the transcript against the published agenda so the passage carries the item it happened under and whatever outcome the minutes recorded for it. Search is hybrid lexical and semantic, and the video is synced to the transcript so any claim can be checked against the recording.
The speaker attribution and the alignment was the hard part. Roughly two thirds of passages have no reliable name attached, and diarization gives you clusters rather than identities, so I mark confidence and basis rather than guessing a name. Alignment is very messy because the published agenda order and the order things actually get taken up diverge constantly, and the recorded outcome is often just "Approved the Consent Agenda" for an item that never got discussed. There is a labeling admin area to improve the dataset after initial collection
At my county meeting on Aug 25th (in a few days) I'm bringing up the fiscal side. Ours was approved under an item title that said no funding was required. No accounting has been done of what enforcement costs when someone takes a camera down. Prosecutor and court hours. Jail time, or just holding, if the state charges it. Lost future wages from otherwise upstanding citizens who now have a felony. None of that was in front of the commissioners when they approved it, and none of it is in front of them now.
Most counties run the same portal software and post to YouTube, so this should generalize. Code is at https://github.com/Tostino/civic-watch and the Pasco instance is at https://pasco.watch if anyone wants to point one at their own county. Happy to help if anyone wants.
So no, actually Flock itself is the problem. If Flock were willing to sign some consent decree that its data can only be accessed by police with a valid judicial warrant, for example, they probably could've headed off this entire mess.
But instead they called people "terroristic" for pushing back. Completely braindead move from the CEO.
Worth looking at how these cameras got approved in the first place. My county's ALPR interlocal passed on the consent agenda in December 2024, one block vote alongside forty other items, no discussion, and went through consent again in April 2025. Nothing was "hidden", it's just that "public" here means a 300 page agenda packet and a four hour YouTube video with no transcript, completely unreadable in practice.
So I spent a week and a half building a pipeline over what the county already publishes. It pulls the CivicPlus portal (agendas, minutes, item metadata, outcomes) and the YouTube channel, does ASR plus diarization, then aligns the transcript against the published agenda so the passage carries the item it happened under and whatever outcome the minutes recorded for it. Search is hybrid lexical and semantic, and the video is synced to the transcript so any claim can be checked against the recording.
The speaker attribution and the alignment was the hard part. Roughly two thirds of passages have no reliable name attached, and diarization gives you clusters rather than identities, so I mark confidence and basis rather than guessing a name. Alignment is very messy because the published agenda order and the order things actually get taken up diverge constantly, and the recorded outcome is often just "Approved the Consent Agenda" for an item that never got discussed. There is a labeling admin area to improve the dataset after initial collection
At my county meeting on Aug 25th (in a few days) I'm bringing up the fiscal side. Ours was approved under an item title that said no funding was required. No accounting has been done of what enforcement costs when someone takes a camera down. Prosecutor and court hours. Jail time, or just holding, if the state charges it. Lost future wages from otherwise upstanding citizens who now have a felony. None of that was in front of the commissioners when they approved it, and none of it is in front of them now.
Most counties run the same portal software and post to YouTube, so this should generalize. Code is at https://github.com/Tostino/civic-watch and the Pasco instance is at https://pasco.watch if anyone wants to point one at their own county. Happy to help if anyone wants.
"I built a tool to help you understand how/if your community has agreed to deploy Flock cameras. You can try it out here: [ ].
The hardest part of this was..."