# Blog | Flusterduck

Canonical: https://flusterduck.com/blog

Field notes

## Notes on friction, fixes, and conversion

What we learn watching real users get stuck, and what to do about it before it costs you sign-ups.

- [August 20, 2026 The field where signups go to die Form abandonment is almost never about the form. It is about one field, and the form has usually been telling you which one all along.](/blog/the-field-where-signups-go-to-die)
- [August 13, 2026 Building an agent that cannot file a lie Our nightly agent is not allowed to report a bug it cannot prove. On moving hallucination from a model problem to a systems problem, and what that trade cost us.](/blog/an-agent-that-cannot-lie-to-you)
- [August 6, 2026 PostHog's Replay Vision sends your users' screens to Google PostHog's new AI feature watches your users' sessions and ships the video to Google to do it. Here's what it does well, what it doesn't, and what it actually costs.](/blog/posthog-replay-vision-sends-your-screens-to-google)
- [August 6, 2026 We gave our AI a card with a hard limit Before our detection agent ever ran unsupervised, we built it an accounting system. Notes on the least interesting thing we've shipped and why it is load-bearing.](/blog/we-gave-our-ai-a-card-with-a-hard-limit)
- [August 2, 2026 How to tell when a deploy breaks your UX A before-and-after method for catching UX regressions within hours of a release instead of weeks later, and for proving your fixes actually landed.](/blog/how-to-tell-when-a-deploy-breaks-your-ux)
- [July 30, 2026 Your error tracker thinks your site is fine Every bug we caught on our own sites last month shipped without throwing a single exception. Notes on the failures that live outside the console.](/blog/your-error-tracker-thinks-your-site-is-fine)
- [July 26, 2026 How to find UX issues without session replay A signal-based method for discovering what frustrates users, without recording screens, watching videos, or storing anything personal.](/blog/find-ux-issues-without-session-replay)
- [July 23, 2026 Teaching an agent to skim Our nightly agent reads every session from the day before deciding what to investigate. It can only afford that because of the least glamorous part of the system.](/blog/teaching-an-agent-to-skim)
- [July 19, 2026 How to reduce form abandonment A practical method for finding the exact field where your form loses people and fixing it, with the field-level measurements that make each fix verifiable.](/blog/how-to-reduce-form-abandonment)
- [July 16, 2026 Our nightly AI agent kept dying at turn 12 The agent that investigates our sites every night started leaving half-finished runs behind, always at the same turn. What I found when I went digging.](/blog/our-nightly-agent-kept-dying-at-turn-12)
- [July 8, 2026 An AI investigates our website every night We pointed our own detection agent at our own sites, expecting noise. It keeps finding real bugs we shipped with confidence. Two stories and what they taught us.](/blog/ai-investigates-our-site-every-night)
- [June 30, 2026 The confused users you never hear from For every person who emails support about a confusing screen, a crowd hit the same wall and left silently. The math of silent confusion, and how to see it.](/blog/confused-users-you-never-hear-from)
- [June 24, 2026 Your launch traffic didn't convert. Here's where it went. Launch day sends you the most curious, least patient visitors you'll ever get. Most teams learn nothing from them. A post about the leak nobody instruments.](/blog/where-launch-traffic-goes)
- [June 12, 2026 What rage clicks tell you A rage click is a user telling you something is broken, in the only language they have. Here is how to read them and what to do next.](/blog/what-rage-clicks-tell-you)
- [June 9, 2026 Session replay won't tell you what to fix Watching recordings feels like research. It mostly isn't. Here's the difference between seeing what happened and knowing what to change.](/blog/session-replay-wont-tell-you-what-to-fix)
- [June 5, 2026 Did your fix actually work? Most teams ship a UX fix and never check whether it helped. The verification loop is the cheapest reliability win you're not using.](/blog/did-your-fix-actually-work)