Deep Sentinel.

Somewhere right now, a stranger is being told to leave

It is 2:14 a.m. on a loading dock in a city you have never visited. A motion event fires. Within a second or two, a trained guard a thousand miles away is already looking at the live feed, decides the figure by the roll-up door is not a raccoon, and speaks: calm, specific, by name of the location. A 100-plus decibel siren is one click away. The person leaves. No police report. No insurance claim. No grainy footage to hand a detective the next morning.

That moment is the whole company. Deep Sentinel sells cameras, sure, and an app, and a subscription. But what it actually sells is the few seconds between "something is happening" and "someone is doing something about it." Most of the security industry has spent decades shrinking the price of those cameras. Deep Sentinel spent its years shrinking that gap instead.

An alarm that cries wolf 9 times out of 10 isn't really an alarm

Traditional security has an honesty problem it would rather not discuss at parties. Burglar alarms, by the company's account, throw false alarms more than 90% of the time - so often that police in many cities deprioritize them entirely. And the cameras? Cameras are excellent at one thing: explaining, in high definition, exactly how you were robbed. After the fact. The footage is for the report, the claim, the closure. It is grief, rendered at 30 frames per second.

The founding scene was not a boardroom. In 2016 a home invasion happened in David Selinger's own suburban neighborhood - armed intruders, a family at home. He went looking for a system that could have stopped it, and found that the entire market was built to witness crime rather than interrupt it. The tension that runs through everything since: a camera that records is a historian. A camera that can prevent needs a human attached to it, fast.

Hire the robot to do the boring part, keep the human for the brave part

Selinger was not an obvious home-security guy. He was an early Amazon engineer who reported to Jeff Bezos and ran R&D on the data-mining and personalization team - the machinery that learned to guess what you would buy next. He later co-founded Redfin and founded RichRelevance. His instinct was the one you would expect from that résumé: the boring, repetitive judgment of "is that a threat or a tumbleweed" is a machine learning problem. Co-founder Winston Chen helped turn that instinct into hardware and a 24/7 guard operation.

The bet was counterintuitive in an industry sprinting toward full automation: do not try to remove the human. Use AI to remove the noise - the cars, the joggers, the cats, the wind - so that a real, awake, trained guard only ever looks at things that might actually matter. Computers filter. People decide. It is a division of labor old enough to be obvious and new enough that almost nobody in security had built a business on it.

What you're actually buying is a person who is already paying attention

The stack looks like cameras. It behaves like a guard booth that happens to fit in your pocket. AI-equipped cameras watch continuously; the computer vision filters harmless motion and surfaces genuine threats to remote guards the company calls LiveSentinels. They engage through two-way audio, trip a siren if needed, and escalate to law enforcement - crucially, with verified crime-in-progress information, which is the kind of call police actually move on.

The numbers it likes to be judged on

A skeptic's favorite move is to ask for the receipts. Deep Sentinel's headline claims are about speed and silence: guards engaging within roughly 30 seconds of a verified threat, and a false-alarm rate it puts at zero - against an industry baseline north of 90%. Heading into its Series B, the company said it had doubled sales bookings year over year, with its Bring Your Own Camera program climbing to nearly half of all sales.

The funding chart tells the other half. Egis Capital Partners - an investor with prior stakes in Alarm.com, Brivo and Stealth Monitoring - led the round, calling it their fifth remote-video-monitoring bet since 2008. When the people who funded your competitors fund you, that is either flattery or foreshadowing.

"The best protection money can buy" - aimed at everyone, not just the gated few

The company states its mission plainly: every family deserves to feel safe, and to have the best protection money can buy. Private guards have existed forever - for people who could afford a person to stand at a gate. Deep Sentinel's argument is that AI is what makes a guard's attention divisible, and therefore affordable, at flat per-camera pricing rather than per-bodyguard pricing. One trained person, augmented by software that never blinks, can watch over far more than one driveway.

Whether that fully democratizes the watchman or simply makes a premium product slightly less premium is a fair thing for a skeptic to keep asking. The pricing sits above most camera subscriptions because there is a payroll behind it. The pitch is that you are not paying for storage - you are paying for someone to care, on the clock, at 2 a.m.

If it works, "the camera caught it" stops being a consolation prize

The interesting bet is not on better cameras. Resolution has been good enough for years. The bet is on response - on collapsing the distance between detection and intervention until the crime simply does not finish. As the AI gets sharper at telling threat from tumbleweed, the human guards get to spend their attention only on the moments that are genuinely up for grabs. That ratio - machines filtering, humans deciding - is the thing that scales, and it is the thing every competitor will eventually have to answer.

Back to the loading dock at 2:14 a.m. The old version of that scene ends with a manager watching footage the next morning, the way you watch weather you can't change. The Deep Sentinel version ends with nothing happening at all - a non-event, a silence, a person who turned around and left because a voice made it clear someone was already watching. The company is betting the future of security is boring on purpose. The most valuable footage, it turns out, is the footage of a crime that never happened.

Things worth knowing at a dinner party

• Bezos-adjacent origins. Selinger was an early Amazon engineer who reported directly to Jeff Bezos before co-founding Redfin and founding RichRelevance.
• The guards have a name. They are called LiveSentinels, and they are trained to talk to intruders directly through the camera - often ending the incident before police are even called.
• A real fright, not a focus group. The company traces back to an actual armed home invasion in the founder's own neighborhood in 2016.
• The industry, inverted. Most systems are designed to document crime. Deep Sentinel is designed to interrupt it - which is a surprisingly rare thing to build a company around.