This story is being published in the Product Hunt Weekly Newsletter. If you'd like to read more stories like this, subscribe here.

Hey Product Hunt!

I'm Aloke, Engineer #1 at Warp and lead eng on Warp's new coding features.

We're all in on agentic coding at Warp, but we also recognize that even the best agents need some human guidance. We just launched a suite of new features to help you closely iterate with agents code review panel, file editor, file tree, slash commands, WARP.md (or use your existing agent.md file).

I noticed that Product Hunt discontinued the teaser / coming soon section this week, and I m curious if it did help you build genuine traction? 

My take (from hunting PH launches):

After my AI startup (trieve) got acquired, I returned to an old idea: building a better Patreon for content creators making serial content like long running comics, shows, or books. The domain patron.com was perfect, but it was owned by World Media as an investment. Instead of assuming it was impossible, I cold emailed the owner Gary Millin through LinkedIn and regular outreach.

The standard advice you get from AIs on this topic is to double down on "strong Instructional Defense" essentially, just write more forceful prohibitions in the system prompt.
This is the approach many teams are taking.
If you ask for a better method, the AI's answer is simply to make the prohibitions even stricter.

The problem is, this doesn't actually seem to work in the real world.

My hypothesis is this:
telling an AI "Do not do X" paradoxically makes the model fixate on X. It forces the model to linger on the forbidden subject, which ironically makes it more susceptible to being 'gaslit' and manipulated into violating its core instructions.

So, I believe that instead of negative reinforcement ("don't do this"), we need to focus on positive reinforcement ("do this instead").