imo @catchback_cards is one of the coolest startups in our yc batch
mystery packs for Pokémon and sports card collectors. cryptographically verifiable odds (you can actually prove the pull was fair). instant Venmo buybacks + $1 marketplace
they're also one of Lucent's earliest users, and here's a bug we caught for them recently: their pack-opening modal was showing a "Error Loading Pack Opening" on pulls that actually succeeded. card was safely in the user's vault. 198 sessions before the team knew with 0 alerts from Sentry.
@sohan_zhang (co-founder & CEO): "Got fed up with UI bugs across different devices. Used Lucent to figure out what was breaking. Saved us a lot of time."
exactly what we built @lucent_ai for 😊
to celebrate 3 months since lauching @lucent_ai, we're giving away 5 Codex Pro / Claude Max plans 🎁
to enter, like this post + comment which one you'd pick (codex vs claude)
winners will be selected from comments in 5 days 🫶
people often think that the hardest problem to solve for any product intelligence tool is data pipeline or infra but instead the hardest thing to crack is how (rightfully) opinionated can you make your tool.
you can show all the statistics on a dashboard and nobody would care but put a single valuable opinion and look at the product go to the moon.
now i used to think that the opinions that our tool would have are my opinions(my team's) for a specific problem but lately i've realised that it's the wrong path.
the right path is to actually make a product that gets opinions from the data it stores from real actual users.
if onboarding A has more steps, is more annoying to complete, has a hard paywall, asks for a review, bla bla bla but it has a higher conversion rate(as per the sessions) than onboarding B which is much simpler and smaller, @lucent_ai must point that the better onboarding was A no matter what any human thinks.
the next obvious question: how does the product really know when to form an opinion vs when to ignore when it sees a session? there is no right or wrong answer to this but we think of this problem a bit differently.
its not about when to form an opinion vs when to ignore but rather its about in what medium is that data useful, to form an opinion. there must be one.
the medium could be anything:
aggregages: daily vs weekly
features: bugs vs insights vs signalling
end user: developers vs GTM
every error, every successful session, every failed workflow has some magical insight hidden in it and we at @lucent_ai want to uncover it for you.
come try us out(linked in the first comment)
its free for the first 400 sessions btw.
career update:
i have taken upon myself to end the misery of watching session replays by joining @lucent_ai.
we are living in incredible times where we are automating(sorry ik this is worse than he who must not be named in 2026) everything that we are doing but people still have to look through session replays to uncover bugs and insights.
at the same time, vision and text models have gotten smart enough to really understand user intent.
put the 2 & 2 together and you'll realise lucent is solving the most annoying problem with a super obvious solution:
ai watches your user sessions. slacks you the bugs and insights. everything you need to know to grow your product.
super excited to get to solve this at day 0.
LFG
NEW: Amplitude + Datadog integrations
Collecting session replays in Amplitude or Datadog - but no one’s actually watching them?
Lucent does it for you.
Automatically surface bugs and user insights, and send them back to the tools your team already lives in.
Introducing @lucent_ai MCP !! 🥰
Connect it to @claudeai, @cursor_ai, or any MCP-compatible client and your coding agent can pull bug reports, session context, and repro steps directly from Lucent.
Demo:
Lucent now works anywhere.
Drop in our SDK and start recording sessions from any framework. Then AI watches them and finds bugs + UX insights automatically 🐛
we built an AI product analyst that watches every session replay and delivers a weekly report on:
- what’s broken
- what’s frustrating users
- what’s working well
let me know if you'd like to try it :) @lucent_ai
I’m hiring a founding engineer @lucent_ai !!
- unlimited claude code credits
- you know how to build and manage teams of coding agents
- remote / sf (will sponsor O1/E3 visas)
- $120-160k, 1-3% equity
- bonus if you like watching session replays (or hate watching them and want to automate it with AI)
comment/DM me your github ☺️
🚀 @lucent_ai launched! AI That Automatically Watches Your Session Replays
"See every bug your users experience"
🌐 https://t.co/GDdMmhs2kN
Congrats @RaeAlisa_!!
she raised $1.3m just a month after launching her product
she likely had minimum traction at that point so I did some research to understand what pushed investors to make this desicion
found a few things:
> she has very good resume
> sold her previous startup in just 9 months
> was 2nd employee in a startup that was sold to Canva
> the product is cool
> good connections
even one of the investors said that they invested mainly because "they believe in her"
this is a good example when "the team" matters more than the product
congrats to her
AI removed the old bottleneck. Writing code is fast. The real challenge now is seeing what users actually hit in production and spotting the weird breakages that slip through.
Lucent @lucent_ai nails this. It watches real sessions and surfaces the bugs and UX paper cuts that normally hide in hours of replay footage.
Feels like running a 24/7 QA teammate on production.
This will become standard in modern stacks. The demo says it all. This is 2026.