Oliver Nathan

Balancing editability with free writing feels tricky. How did you keep edits from breaking that raw flow?

What kind of tasks are users actually trusting it with first in real projects? More like small fixes or something slightly bigger?

How well does the intent detection hold up with messy, slang heavy Instagram DMs?

How often does the system break when recording flows with Playwright, especially on dynamic pages?

A dedicated leaderboard removes a lot of noise from the usual feed. You can actually compare like for like launches.

Do we become operators of systems instead of building things?

Feels like the real answer isn't brans vs product, it's timing, too much brand early can distort feedback, too little means no one notices you.

This feels like distribution turning into product usage. Do you think behavior sticks once the incentive is removed?

Launch day really does compress months of assumptions into a few hours. What surprised you most about how people actually used Velo?

This feels way more complete than subtitle only localization. How big a difference has on screen text translation made in output quality?

That 20 seconds "aha" feels huge. What was the hardest part of getting users to feel value that quickly?

The SSO versus API rate limits comparison says a lot. How often do you think founders confuse loud requests with real demand?

I think best depends way more on task type than people want to admit. Coding, research, and agent loops usually need different strengths.

This is probably the strongest kind of validation you can get, people were interested even when half the launch was broken.

The interesting part is not even the recording, it's how someone reacts when you say no. Do you think that response tells you more than the tool itself?

I'd honestly pick anonymity. Reddit without that feels like a totally different place. Would people still post honestly if they had to verify with their face?