Charmizard's comments

On: If AI Is Doing the Investigation, Version the Investigation

[Charmizard]

When AI get's deep in investigation, it's memories fill up with absolute garbage.

[markfrwc → Charmizard]

True. I spin up a new Claude session for each bug. Keeps the context focused and the transcript clean. Trellis makes this easy — each session is scoped to the worktree you're working in.

On: How do you approach outbound discovery without feeling intrusive?

[Charmizard]

I'm not a founder, but I am a builder and in the growth domain. So I can provide a different view for this.

1. Outbound in general is a hard space in 2026. Folks generally have a predisposition to auto-delete or be sensitive to 'wasting time' with cold outreach with the idea that they will most likely be sold to at some point.

2. Your best case scenario here is to offer free access to beta and then talk to every user you possibly can after that. Because at this point you have 1) ensured the person you are going to talk to has a use for your product and 2) offered them value before asking for something.

TL;DR

Offering value before asking for something is the key. This is true in almost every area.

[D___R___ → Charmizard]

I agree, offering real value upfront feels like the only sustainable way to do this now.

The free beta + conversation after they’ve actually used the product makes a lot of sense. It shifts the dynamic from pitch to feedback loop.

What are you building, and in which space? Would be interesting to hear how you’re applying this in practice.

On: It's insulting to read AI-generated blog posts

[Charmizard]

Idk how I feel about this take, tbh. Do things the old way because I like them that way seems like poor reasoning.

If folks figure out a way to produce content that is human, contextual and useful... by all means.