> ## Documentation Index > Fetch the complete documentation index at: https://docs.useartifact.app/llms.txt > Use this file to discover all available pages before exploring further. # Learning from users > notes from discovery calls and conversations, as i go. **5 september 2026** had a call today while working on the prototypes. one thing that stayed with me: sometimes you don't know what you're building until you've explored for a bit. he described that as a possible way the work unfolds. useful challenge for Artifact: if we're trying to help someone keep their direction, what do we show before that direction is clear? i don't want us to force clarity that isn't there. he also brought up logs, instructions and agent recaps. i still need to understand which of those he actually uses, but it makes the comparison more concrete. what does Artifact help with after someone already has a decent setup? i don't see this call as validation yet. we mostly discussed the idea. next time, i want to start from the last time someone returned to their project and ask what they did first, before explaining Artifact. *these are my takeaways, not direct quotes.* *** **29 august 2026** one part of this call is worth sitting with. after seeing the screens, he imagined Artifact helping him put a blurry vision into words or drawings. not just keeping a direction clear once he already knows what it is. that gives me something specific to explore: can we help someone work out what they mean, without deciding it for them? it doesn't mean we should add a canvas just because one came up in the conversation. he was encouraging about the direction. but the recording only covers the last six minutes, after the demo, and i was asking him to imagine a tool. useful input for an exploration; i still need to see what it helps him do in real work. *** **28 august 2026** went back through a conversation about working with multiple agents. he described keeping a short plan, asking agents to update it, and merging tested work back into main. the sessions themselves can disappear. the product exploration is messier: designs, screenshots, different tries, sometimes a revert. but he also said he relies on his own mental model and doesn't need to keep everything that came before. he'd offered to try a link and give feedback. i didn't have one to send yet. so there was interest in taking a look, but that isn't the same as needing the thing. that's what i don't want to gloss over here. i was trying to understand where Artifact could fit, but messy doesn't automatically mean something he wants fixed. what, if anything, does he find difficult in that process? *** **25 august 2026** this one is a reminder not to assume everyone builds like me. i did that in the call, then had to step back and ask how he actually brainstorms. one specific thing: he gives Claude the options he's already considered and their pros and cons, so it doesn't keep suggesting ideas he's already thought through. that's something concrete to investigate, beyond asking whether context is a problem. he could see value in Artifact, but said his current to-do list works for the amount of feedback he's getting. paying for something like this made more sense to him on a bigger project. i need to keep that distinction. he also brought up a real customer request. i went into explaining how Artifact would handle it instead of finding out what he actually did with it. looking back, that's where i want to spend more time in the next conversation. *these older notes are takeaways too, not direct quotes.*