Two solutions tackling the hidden costs of AI-assisted development in completely different ways.
I've been watching developers complain about AI-generated code debt for months now. The pattern is always the same: team adopts Cursor or Copilot, productivity shoots up, then six months later someone notices the codebase has turned into a maintenance nightmare.
Two ideas caught my attention recently. Both try to solve problems created by AI tools, but they couldn't be more different in their approach. One targets doctors who can't figure out drug costs. The other monitors your repos for AI-induced entropy. Neither sounds sexy, but both hit real pain points that existing tools ignore.
The Prescriber Affordability Checker (EHR Plug-in) tackles something most people don't think about: doctors have no clue what medications cost. They prescribe a drug, patient goes to pharmacy, finds out it's $400, never fills it. Everyone loses.
The solution is a Chrome extension that shows real-time drug costs while doctors are prescribing. One click to switch to generics. Auto-populated prior authorization forms. Alerts when something will bankrupt an uninsured patient.
This isn't a tech problem, it's a workflow problem. The technology already exists - companies like Surescripts handle prescription data, GoodRx has pricing APIs. But small medical practices can't access this stuff easily. They're stuck with clunky EHR systems that weren't built for price transparency.
The browser extension approach is clever. No IT department, no vendor negotiations, just install and go. But it's also risky. Medical practices are paranoid about security, and "random Chrome extension running on PHI screens" sounds like a HIPAA nightmare waiting to happen.
The market timing feels right though. CMS is pushing price transparency rules, EHRs are opening up APIs, and there's genuine demand from physicians who are tired of playing prescription roulette with their patients' wallets.
The AI-Generated Code Health Dashboard goes after a different kind of invisible problem. AI tools help you write code faster, but they also create subtle technical debt that compounds over time.
File sprawl. Duplicate logic with slight variations. Orphaned types. Empty catch blocks everywhere. You don't notice it at first because the AI-generated code works. But six months later, your repo is a mess and no one can figure out why maintenance is taking forever.
This dashboard connects to your GitHub repos and tracks entropy metrics over time. It can flag likely AI-generated code, detect duplicate patterns, and automatically create cleanup tickets in Linear or Jira when things get out of hand.
The AI attribution angle is genuinely novel. No existing tool tracks which code came from agents versus humans, or correlates that data with technical debt metrics. CodeClimate and SonarQube catch code smells, but they have no concept of AI provenance.
But here's what worries me: GitHub and GitLab are already building AI analytics into their platforms. GitHub's Copilot Metrics API launched this year. How long before they add an "entropy score" tab to GitHub Insights and kill this entire market?
These ideas represent two different philosophies about where to intercept problems.
The prescription checker inserts itself into an existing workflow. Doctors are already prescribing drugs, the tool just adds cost data at decision time. It's workflow augmentation, not workflow replacement.
The code health dashboard creates a new workflow. It assumes engineering managers want to actively monitor and remediate AI-generated technical debt. But do they? Most teams I know are drowning in dashboards already.
The prescription tool has a clearer path to immediate value. Doctor sees $400 brand drug has $12 generic, switches prescription, patient saves money. That's tangible ROI within the first use.
The code dashboard requires more faith. You need to believe that proactive entropy monitoring prevents future maintenance pain. That's a harder sell, especially to engineering managers who measure everything in quarters.
Both markets are underserved today, but for different reasons.
Healthcare moves slowly and enterprise sales cycles are brutal. But once you're integrated into a medical practice's daily workflow, switching costs are high. The prescription checker could build a genuine moat through habit formation.
Developer tools move fast and have shorter sales cycles. But the big platforms (GitHub, GitLab, linear tools companies) can add features quickly. The code dashboard needs to move fast and build network effects before someone else claims the space.
The prescription checker faces regulatory complexity but has clearer monetization. Doctors will pay for tools that help their patients, especially if it reduces administrative headaches.
The code dashboard faces platform risk but has better virality potential. Engineering teams share tools, and a good entropy score could become a status symbol in developer communities.
Both ideas exist because AI tools create second-order effects that no one anticipated. Copilot helps you write code faster, but it also creates maintenance debt. EHRs digitized prescriptions, but they didn't solve the cost visibility problem.
The winners in AI tooling won't just be the companies building AI. They'll be the ones solving the problems that AI creates.
I'm more bullish on the prescription checker in the short term. Healthcare incumbents are slow to innovate, the regulatory moat is real, and the value proposition is immediately obvious.
The code dashboard has higher upside but faces more execution risk. If they can build the benchmark data moat before GitHub ships native AI analytics, they win big. If not, they become a feature request in someone else's product roadmap.
Both teams need to move fast. In AI tooling, timing isn't everything, but it's close.