Joshua Sanchez-Lopez has been arrested 35 times.1 His rap sheet reads like a bingo card of antisocial behavior: involuntary manslaughter, drug offenses, car theft, and a charming habit of joking about his ankle monitor on Snapchat while fleeing from police.2 He is, by any reasonable assessment, a man who has demonstrated - repeatedly, flamboyantly, almost recreationally - that he poses a danger to the people around him.
So when Las Vegas Justice of the Peace Eric Goodman ordered Sanchez-Lopez released on bail with electronic monitoring in January, the Las Vegas Metropolitan Police Department did something unusual. They said no.3
Not quietly. Not through back channels. They flat-out refused to place Sanchez-Lopez in the monitoring program, citing his history of ignoring its rules, his bench warrants, his failure to appear in court, and - one suspects - a general institutional exhaustion with processing the same man through the same revolving door for what must feel like the hundredth time. When Judge Goodman threatened contempt proceedings, Metro escalated to the Nevada Supreme Court, asking it to settle the jurisdictional question once and for all.4
The legal merits of this standoff are genuinely interesting. Nevada statute gives the sheriff authority to determine whether electronic supervision poses an unreasonable risk to public safety.5 The judge, meanwhile, has constitutional authority over bail and pretrial release. Both sides are invoking real law. This isn’t cops gone rogue - it’s a structural collision between two branches of government that were never designed to disagree this publicly about something this straightforward.
But here’s the question nobody in the legal debate is asking: How did we get to a place where a judge can make a decision this obviously, measurably terrible, and the only recourse is a dramatic standoff that requires the state supreme court to intervene?
The answer, it turns out, has less to do with judges than with voters. And the fix might be simpler than anyone thinks.
American criminal justice policy is controlled by two tribal coalitions, and neither one is optimized for the thing that actually matters: whether people are being harmed.
The “tough on crime” package bundles genuine public safety concerns with drug war zealotry, mandatory minimums for nonviolent offenses, and a generalized enthusiasm for incarceration as a solution to problems it cannot solve. The “reform” package bundles genuinely necessary corrections to those excesses with a reflexive skepticism toward detention that doesn’t always distinguish between a first-time marijuana possessor and a 35-time arrestee who treats the justice system like a game of tag.
The result is a system where drug offenses remain the single largest driver of federal incarceration at roughly 44–45% of all Bureau of Prisons inmates,6 while state prisons hold hundreds of thousands more on nonviolent charges7 - all at a median cost of roughly $61,000 per prisoner per year.8 Meanwhile, a small population of prolific, demonstrably dangerous offenders cycles through a revolving door that produces a predictable stream of victims.
The criminological research on this is remarkably consistent and not particularly controversial: somewhere around 5 to 8 percent of offenders account for roughly half of all crime.9 A landmark Swedish study tracking 2.5 million people found that just 1 percent of the population was responsible for 63 percent of all violent crime convictions.10 The first Philadelphia Birth Cohort study found that 6 percent of the cohort committed 52 percent of all juvenile criminal activity.11 A system that was calibrated to actual harm production - radically lenient where there’s no victim, genuinely serious where there demonstrably is one - would probably mean dramatically fewer people in prison overall and dramatically fewer victims of violent crime.
Nobody designed this dysfunction. It emerged through path dependency, tribal consensus, and the political incentive to maintain bundled ideological packages rather than adopt a harm-focused framework that would require each side to concede significant ground. It is theology masquerading as policy, and people are dying because of it.
But here’s the thing: the people responsible for this - the judges, the DAs, the legislators - are, in most cases, elected. The democratic feedback mechanism exists. It’s right there. The problem is that it can’t function because voters have almost no idea what any of these people are actually doing.
Quick quiz: Who is your local district attorney? What are their charging policies for violent repeat offenders versus first-time nonviolent drug offenses? How do their prosecution rates compare to their predecessor's? What happened to crime rates after they took office?
If you’re like the vast majority of Americans, you don’t know the answer to any of these questions. Not because you’re stupid or apathetic, but because finding out would require you to attend hundreds of court proceedings, dig through public records databases, cross-reference crime statistics, and perform the kind of longitudinal policy analysis that would take a trained researcher weeks to complete.
The same goes for judges. Voters in jurisdictions where judges are elected - which includes Nevada - typically know nothing about judicial candidates beyond their party affiliation, if that. The idea that voters could evaluate Judge Goodman’s bail decisions against those of his peers, track the recidivism rates of defendants he released, or compare his risk assessment accuracy to the jurisdictional average is laughable. That information exists in principle, scattered across court records and police databases. In practice, it might as well not exist at all.
This is the core democratic failure. By the time a voter is making a decision about a judge or a legislator, they’re operating on such a thin and distorted slice of information that their “democratic input” is essentially noise shaped by tribal affiliation. The feedback loop between policy decisions and electoral consequences is so degraded that officials can pursue ideologically motivated agendas for years with zero accountability - not because voters approve, but because voters literally cannot see what’s happening.
And the civic foundation isn’t exactly robust to begin with. The Annenberg Public Policy Center’s annual civics survey found that 35 percent of Americans cannot name all three branches of government, and only 7 percent can name all five rights protected by the First Amendment.12 As APPC director Kathleen Hall Jamieson put it, people are unlikely to protect freedoms they don’t know they have, and will struggle to hold leaders accountable if they don’t understand how the branches of government work.13
James Madison designed a system that assumed an informed electorate evaluating representatives based on performance. What he got - what was probably always inevitable at continental scale - is an electorate making decisions based on vibes, yard signs, and which team their preferred cable news network told them to root for.
Now imagine a website. Call it something boring and trustworthy - GovLens, CivicPulse, whatever focus groups like. You visit it. You enter your address. And it shows you, in plain language, what your elected officials have actually been doing.
Your local judge? Here’s a dashboard. Judge Goodman released 47 defendants classified as high-risk in 2025. Of those, 31 were rearrested within 90 days. The jurisdiction average for comparable judges is 22 out of 47. Here are the three most serious outcomes, including a defendant arrested for his 36th offense while on electronic monitoring he’d previously violated.
Your district attorney? Here’s their charging breakdown by offense category, compared to the state average. Here’s the conviction rate. Here’s what happened to violent crime rates in the two years since they took office. Here’s how they allocate prosecutorial resources between violent offenses and drug possession.
Your state legislator? Here’s every vote they cast this session, linked to measurable outcomes where data is available. They co-sponsored a bail reform bill - here’s the before-and-after on violent crime in affected jurisdictions. They voted for a spending package - here’s where the money went.
Now here’s the key: the system asks you what you care about. Property taxes? School quality? Public safety? Small business regulation? It surfaces the information most relevant to your priorities, not a news editor’s judgment about what’s nationally important. A parent sees school board outcomes. A small business owner sees regulatory impact. A crime victim sees judicial sentencing patterns. Everyone sees verifiable data tied to the specific officials on their specific ballot.
This isn’t a filter bubble. It’s not telling you what to think. It’s not framing information to favor one side. It’s personalized relevance with objective data - showing every voter the slice of governance that matters most to them, grounded in outcomes rather than rhetoric.
The reason this didn’t exist ten years ago isn’t that nobody wanted it. It’s that the human labor costs were prohibitive. Organizations like AllSides and Ground News do genuinely valuable work ranking media bias and presenting multiple perspectives on major stories. But they’re constrained by editorial bandwidth. They can cover the presidential race, a handful of Supreme Court decisions, maybe a few hot-button national issues. They cannot possibly cover your county sheriff’s race, your local DA’s plea bargaining patterns, or your city council’s vote on that rezoning proposal three blocks from your house.
AI eliminates the bandwidth constraint. Natural language processing can ingest court records, legislative text, budget documents, and regulatory filings at a scale no newsroom could match. The same analytical framework that evaluates Judge Goodman’s bail outcomes can simultaneously evaluate every judge, every DA, every city council member, every school board, in every jurisdiction in the country. Not because an editor decided it was newsworthy, but because the computational cost of doing it is trivially small once the system exists.
The technical components are largely available. Public records are increasingly digitized. NLP can extract structured data from legal and legislative documents. Statistical methods can link policy decisions to outcomes with appropriate uncertainty. The missing piece isn’t technology - it’s assembly, institutional design, and the commitment to genuine neutrality.
That last part is the hardest, and it’s worth being honest about it. The moment such a system is perceived as having a thumb on the scale - weighting certain outcomes, framing data to favor one tribe - it gets sorted into the partisan landscape like every other information source and becomes useless. The methodology has to be transparent. The outputs have to be auditable. The governance structure has to be credibly independent. This is a serious design challenge. But it’s an engineering challenge, not a human-nature challenge, and those tend to be the kind we can actually solve.
Here’s the thought experiment that matters: If this website existed - if you could, in five minutes, see a clear, verified, nonpartisan summary of what your elected officials had actually done and what outcomes their decisions produced - would you use it before voting?
Most people, when asked sincerely, say yes. Not because they’re policy wonks, but because they’re frustrated. They know they’re voting in the dark. They know the yard signs and attack ads aren’t giving them real information. They’d welcome something better if it existed and they trusted it.
And here’s what makes this approach different from every other reform proposal in the “fix democracy” space: it doesn’t require anyone to give up power. It doesn’t require a constitutional amendment. It doesn’t require restructuring federalism or passing controversial voting reforms. It doesn’t even require changing human nature. It just requires building a tool that makes existing public data legible - and then letting the democratic mechanisms that already exist do what they were designed to do.
Judges like Goodman currently operate in an informational vacuum where terrible decisions produce no electoral consequences because voters can’t see them. Fill that vacuum with transparent, verifiable outcome data, and you don’t need to redesign judicial accountability from scratch. The accountability mechanism is already there - it’s called an election. It just needs inputs that aren’t garbage.
The Founders designed a system that assumed citizens could evaluate their representatives. For 250 years, that assumption has been mostly fiction, because the information costs were too high and the intermediaries were too compromised. For the first time in history, technology exists that could make it true.
A 35-time arrestee is sitting in a Las Vegas jail cell right now because the cops decided a judge’s order was too dangerous to follow.14 That’s a crisis. It’s also a symptom. The disease is that voters in Clark County had no way to know what Judge Goodman was doing until a police department decided to make a public stand. That’s not democracy working. That’s democracy failing, with a dramatic workaround.
The question isn’t whether we need better judges. It’s whether we’re willing to build the tools that would let voters figure out which ones they already have.