A checkout page malfunctions on mobile. An email validation script misfires. A dead link sits three clicks into your product. And the user? They close the tab and move on. Maybe forever.
This is the silent erosion most product teams know but struggle to quantify. According to Lucent, a startup emerging from Y Combinator's current batch, the vast majority of people who encounter a software bug never report it. They just leave—taking their frustration, and often their business, elsewhere.
It's a dynamic that helps explain the surge in session replay tools over the past few years. Platforms like PostHog, FullStory, and Hotjar have become standard equipment for product teams trying to understand what actually happens when users interact with their software. The problem, though? Someone still has to watch those recordings. And when your SaaS product is logging thousands of sessions a week, the math gets untenable fast.
Enter Lucent's pitch: What if an AI watched every single session for you?
The two-person San Francisco team—founded by Alisa Rae, who previously built and sold an edtech startup before becoming a founding engineer at MagicBrief (later acquired by Canva)—is betting that automated bug detection from session replays is a category waiting to happen. Or perhaps more accurately, a category that's already happening, with Lucent jostling for position alongside both venture-backed competitors and AI features now rolling out inside the very platforms it integrates with.
Last week alone, the company says it analyzed north of 45,000 session replays. That volume, in itself, speaks to a certain kind of traction—or at least to how much recorded user behavior is piling up, unwatched, across the software industry.
The Mechanics
Lucent's workflow is purposefully uncomplicated. Connect your PostHog account (the integration they're highlighting for now), and the system begins ingesting sessions automatically. When something breaks—a payment that fails silently, a form that rejects a perfectly valid email, a 404 error three layers into your navigation—Lucent flags it. More importantly, it groups similar issues across users, so twenty people hitting the same broken link don't generate twenty separate alerts. You get one, with reproduction steps and a list of everyone affected.
The alerts route into tools teams already use: Slack for immediate visibility, Linear for ticket creation. No dashboard to check, no separate queue to manage. The aim is near-real-time notification—find out your checkout flow is broken this afternoon, not three weeks from now when someone finally bothers to email support.
Rae demonstrated the product in a LinkedIn post earlier this year, and the examples felt refreshingly mundane. Real bugs. The kind that chip away at conversion rates and customer trust but rarely surface in support queues. A checkout failing for no obvious reason. Dead ends in navigation. Overzealous validation scripts.
One commenter on that post claimed, with the casual hyperbole of social media, "I thought Den (YC F25) had 0 bugs till I used Lucent." Formal case studies haven't been published yet, but the company is offering to analyze the last 300 PostHog sessions for free—a low-commitment way for skeptical teams to see what's been slipping through the cracks.
Crowded Territory
Here's where it gets interesting. Lucent is hardly the only company to spot this opportunity.
Amplitude mentioned a "session replay agent" on its Q3 2025 earnings call—AI that reviews thousands of sessions and surfaces friction points automatically. Mixpanel shipped AI-powered summaries and "Frustration Signals" for its replay tool around the same time. LogRocket has been layering AI into its product under the name "Galileo," targeting rage clicks, dead clicks, and other telltale signs of user struggle.
Even within Y Combinator's own portfolio, Lucent has company. At least two other startups—Decipher (Winter 2024) and Prism (Fall 2025)—are positioning themselves as AI agents that watch session replays and tell developers what to fix. The category, in other words, is forming in real time.
So the central question isn't whether AI-assisted session analysis makes sense. Clearly it does, or this many teams wouldn't be building it. The question is whether a dedicated, standalone tool can win against native features inside platforms teams already pay for—or against fellow startups chasing the same workflow.
Lucent's angle appears to be speed and simplicity. Setup takes minutes, not days. It plugs into existing replay sources rather than trying to replace them. And the core loop—watch, detect, alert, ticket—runs without human intervention. Whether that's differentiated enough remains to be seen. The company closed a pre-seed round last October (sources report either AU$2 million or roughly US$1.3 million, depending on the telling), with backing from Long Journey Ventures, Horizon Ventures, Browder Capital, and Weekend Fund.
What It Means
The technical architecture isn't fully public, but the value proposition is straightforward enough. If you're a product manager staring at a backlog of unreviewed sessions, or an engineering lead who suspects bugs are slipping through but lacks the hours to hunt them down manually, Lucent promises to do the watching for you. And then tell you what broke, who it broke for, and how to reproduce it.
Pricing hasn't been announced—access is currently via demo request and a waitlist at lucenthq.com. The 300-session free analysis suggests a usage-based or volume-tiered model might be in the works.
Whether Lucent carves out a sustainable position will hinge on execution: detection accuracy, integration friction, and whether teams find the alerts useful or just noisy. The competition is real, and it's coming from multiple directions—both well-funded incumbents and YC batchmates targeting the same pain point.
But the underlying problem isn't going anywhere. Bugs that never get reported still cost revenue. They still lose users. They still corrode trust, one silent exit at a time. And if the vast majority of people genuinely won't speak up when something breaks? Well, that's a pretty good reason to stop waiting for them to file a ticket.
