a16z“How We Can Eliminate Crime” | Ben Horowitz and Garrett Langley
Erik Torenberg and Ben Horowitz on a technology-and-policy blueprint to deter crime without mass incarceration.
In this episode of a16z, featuring Ben Horowitz and Garrett Langley, “How We Can Eliminate Crime” | Ben Horowitz and Garrett Langley explores a technology-and-policy blueprint to deter crime without mass incarceration They argue that the most humane crime strategy is high certainty of getting caught (deterrence) rather than simply increasing incarceration, because prison is costly and often destroys future opportunity.
At a glance
WHAT IT’S REALLY ABOUT
A technology-and-policy blueprint to deter crime without mass incarceration
- They argue that the most humane crime strategy is high certainty of getting caught (deterrence) rather than simply increasing incarceration, because prison is costly and often destroys future opportunity.
- They propose a “Teach for America” model for law enforcement—using student-debt relief and other incentives to address a staffing crisis largely driven by cultural stigma and reputational decline in policing.
- They describe a modern “technology stack” for public safety—cameras, gunshot detection, drones, license plate readers, and an AI orchestration layer—to convert abundant raw data into actionable intelligence and safer, more accountable police responses.
- They explain why clearance rates (even for murder) have fallen nationally: less witness cooperation, more randomized/organized crime, evidence volume outpacing tools and skills, and the loss of experienced detectives due to early retirement and understaffing.
- They emphasize governance and legitimacy: transparency features, configurable data retention/sharing policies, and community policing are positioned as necessary to build trust and avoid the inequities of privatized security.
IDEAS WORTH REMEMBERING
5 ideasCertainty of apprehension is framed as the key lever for reducing crime.
They contend that when people believe “you will get caught,” crime becomes less economically and socially attractive, reducing the need for extreme sentencing or “harsh models” like mass imprisonment or Singapore/El Salvador-style punishments.
Policing’s staffing crisis is described as primarily cultural, not demographic.
Langley argues the supply of potential recruits hasn’t fundamentally changed; stigma and vilification drove early retirements and recruitment collapse, which then forced some departments to lower standards with dangerous consequences.
A debt-relief service pathway could rapidly expand the public-safety workforce.
A proposed program would trade student-debt retirement for 2–4 years of service in patrol or civilian policing roles, easing staffing shortages and raising the skill baseline without requiring military service.
Modern public safety requires an end-to-end sensor stack plus an AI ‘sensemaking’ layer.
They describe combining cameras/LPRs, gunshot detection, and drones with AI-driven orchestration so departments can act on data quickly and consistently rather than leaving footage unwatched and leads unpursued.
Better intelligence can reduce both crime and violent police encounters.
Horowitz claims Vegas saw a large drop in police shootings after deploying cameras/drones because officers approach incidents with clearer identification, backup planning, and less uncertainty—shifting from “subjective” to “objective” policing.
WORDS WORTH SAVING
5 quotesIf you don't enforce crime, what you end up is with lost generations.
— Ben Horowitz
Outside of Vegas, the national average is around 47% clearance rates, so you have a coin flip.
— Garrett Langley
The irony of defund the police is defund the police for poor people. To privatize the police for rich people.
— Ben Horowitz
Do you realize if the federal government wanted to find you, a license plate reader is the dumbest way to do it. I will just get a cell phone dump. And I will know your exact location in real time at all times.
— Garrett Langley
We helped return over 450 missing children this year.
— Garrett Langley
QUESTIONS ANSWERED IN THIS EPISODE
5 questionsWhat would a “Teach for America for law enforcement” actually look like in practice—what roles would be civilian vs. sworn, and how would training be shortened without lowering standards?
They argue that the most humane crime strategy is high certainty of getting caught (deterrence) rather than simply increasing incarceration, because prison is costly and often destroys future opportunity.
Horowitz claims Vegas has a 90%+ murder clearance rate—what specific operational practices (community policing, tech, DA/judge alignment) explain the gap versus the ~47% national average?
They propose a “Teach for America” model for law enforcement—using student-debt relief and other incentives to address a staffing crisis largely driven by cultural stigma and reputational decline in policing.
If the goal is deterrence through certainty of capture, how do departments measure and communicate that certainty without creating a perception of blanket surveillance?
They describe a modern “technology stack” for public safety—cameras, gunshot detection, drones, license plate readers, and an AI orchestration layer—to convert abundant raw data into actionable intelligence and safer, more accountable police responses.
Which part of the ‘tech stack’ delivers the highest ROI first: LPRs, drones, gunshot detection, camera networks, or the AI orchestration layer—and how does that vary by crime type?
They explain why clearance rates (even for murder) have fallen nationally: less witness cooperation, more randomized/organized crime, evidence volume outpacing tools and skills, and the loss of experienced detectives due to early retirement and understaffing.
Langley argues privacy criticism is ‘falsely focused’ and trust is the real issue—what concrete governance/audit controls would persuade skeptical communities (e.g., independent oversight, warrant policies, public reporting)?
They emphasize governance and legitimacy: transparency features, configurable data retention/sharing policies, and community policing are positioned as necessary to build trust and avoid the inequities of privatized security.
Chapter Breakdown
Deterrence as the humane alternative to “lost generations”
Ben Horowitz and Garrett Langley frame crime enforcement as a societal incentives problem: when the odds of getting caught are low, crime becomes a rational career path and corrodes communities. They argue the goal isn’t mass incarceration, but credible certainty of capture that prevents people from entering the system in the first place.
A “Teach for America” pipeline to solve policing’s staffing crisis
Langley proposes a national service-style recruiting program for law enforcement: trade student debt relief for 2–4 years of service in a police department, including non-sworn roles. The idea targets both staffing shortages and capability gaps while boosting the status of public safety work.
The real “people problem”: culture, stigma, and lowered standards
They argue the hiring crisis is primarily cultural—driven by stigma, early retirements after social unrest/COVID, and a narrative shift from “hero” to “villain.” Horowitz warns shortages can force departments to lower standards, which can compound misconduct and further damage legitimacy.
Recruiting through modernization: tech, visibility, and the “Cybertruck effect”
Horowitz describes using high-visibility modernization (vehicles, tech deployments) to make policing appealing and community-facing. They claim flashy but practical investments can improve recruitment, increase community engagement, and shift perceptions faster than abstract messaging.
Building a city’s crime-prevention technology stack (sensors → AI orchestration)
The conversation shifts to “products”: address crime type-by-type with sensors like gunshot detection, cameras, and drones. The missing piece, they argue, is an AI/orchestration layer that turns abundant data into actionable intelligence while supporting transparency and accountability.
Deterrence vs. incarceration: why certainty of punishment matters more than severity
They criticize systems that combine low capture rates with long prison sentences, calling it both economically costly and socially destructive. Their preferred model is “you will get caught,” which changes behavior upstream and reduces reliance on prison as the primary tool.
Intelligence-based policing + community cooperation (Vegas as proof point)
Horowitz argues intelligence improves safety for suspects, officers, and bystanders by reducing uncertainty and mistaken stops. They connect high clearance rates in Las Vegas to community policing—people share information when trust and professionalism are strong.
Why clearance rates are collapsing nationwide
Langley offers multiple drivers of declining clearance rates: higher evidence standards, reduced witness cooperation, a shift toward more random/organized violence, evidence overload outpacing tools, and the loss of experienced detectives due to retirement and staffing gaps.
Vegas case study: visible community support and practical ROI
Horowitz says the biggest surprise in Vegas is how strongly residents—especially working-class locals—support the technology approach despite national press criticism. He emphasizes that small, targeted investments can create outsized operational improvements and public confidence quickly.
Private funding as a catalyst for police innovation (and why cities can’t start alone)
They describe public-private partnerships as a pragmatic bridge: private dollars fund initial deployment, then cities decide whether to absorb costs later. Examples include police foundations and corporate support (e.g., major employers funding local safety improvements).
Privacy vs. trust: what surveillance debates are really about
Langley argues most criticism labeled “privacy” is actually distrust in local police institutions. They claim license-plate readers operate in public spaces and are less invasive than other modern tracking, while emphasizing transparency tools as a mechanism to build public confidence.
Prison reform and rehabilitation: separating enforcement from corrections failures
Horowitz and Langley support reforms that reduce recidivism and avoid turning non-violent offenders into hardened criminals. They highlight programs that redirect first-time, non-violent cases into structured pathways (work/school/home) and point to successful reentry models.
Crime stats, underreporting, and “gaslighting” about public safety
They argue official crime numbers can be misleading due to underreporting, changed recording practices post-2020, and non-prosecution discouraging victims from calling police. They recommend using victimization surveys and lived experience as critical reality checks.
Policy knobs in tech: data retention, data sharing, and cross-border crime
Langley describes configurable governance levers: how long data is retained and who it can be shared with. They argue local values and state laws should determine settings, but warn that strict borders for data sharing can hinder solving crimes that cross jurisdictions.
Organized and sophisticated crime: from retail theft to logistics fraud
They contend modern crime includes highly organized, business-like operations that exploit weak enforcement and policy loopholes. Examples include retail theft rings and sophisticated cargo/logistics schemes that appear legitimate on paper while moving stolen goods at scale.
The future of policing: intelligent, precise, and drone-enabled real-time response
Langley envisions a system where integrated sensors, shared hotlists, and real-time crime centers enable precision interventions. Drones and analytics reduce the need for risky, force-heavy responses and can shift officers’ time from paperwork toward community presence.
Success stories: recovering missing children and high-stakes real-world impact
They close with outcomes they view as unambiguously positive, especially rapid recovery of missing children and incidents where fast identification prevents escalation. These stories are used to argue that the benefits of real-time intelligence become obvious when stakes are personal.
EVERY SPOKEN WORD
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome