Ivo Gospodinov

the prompt injection and jailbreak problem is the one everyone's talking about, but i keep thinking about the other end - agents also need a verified sense of who they're acting on behalf of. right now there's no standard for that. earlycore secures what agents do, which is critical. who they represent is still an open problem. been thinking about that identity layer with northr identity and...

the "skills and visual identity" feature is the right call - the biggest gap with ai agents isn't capability, it's that every session starts from scratch and the agent has to rediscover who you are. curious how runable handles that cold start for new users. been working on exactly that layer with northr identity - the idea that your context travels with you so the agent knows you from message...

the stale claude.md problem is real and you've nailed the fix for project context. what i keep running into is the other half of the cold start - claude doesn't know who the developer is either. role, preferences, how they like to work. codeyam handles what the codebase needs claude to know. been working on the human side of that same gap with northr identity. feels like these two would stack...

the 45-second context switch problem is real and annoying. the "stuck on a form field" use case is the one i'd lead with in marketing - it's the most universally relatable. curious whether you're seeing people use it more for quick lookups or longer analytical questions like contract review. congrats on the launch krunal.

the "context into claude" framing is exactly right. chordio loads the product context. the gap i keep seeing is the layer above - claude also doesn't know the pm or designer using it. role, preferences, how they think about tradeoffs. been building northr identity for that piece. product context + human context together is a much better starting point. congrats on the launch ehud.

the SQLite memory layer with relevance decay is the right architecture. project context that fades naturally is much better than stale context that poisons the agent. the gap i keep seeing is the layer above - who the developer is doesn't persist anywhere either. working on that with northr identity. same problem, different stack.

the "wrestling between tools" framing is exactly right kane. that context switching cost is real - and most of it is the same problem every time. chatgpt doesn't know what you're building, claude doesn't know who you are. been building northr identity to fix that specific gap. every session starts with full context. congrats on the launch.

the kurzgesagt-style bet is the right call. the uncanny avatar problem has killed more ai video tools than anything else. curious how the pricing evolves - the $150/5min comment is the main objection you'll keep hearing. congrats on the launch ritam.

the voice signature + story bank approach is exactly right. been building the same thing from a different direction - northr extracts identity from AI conversations so it's available the moment you open chatgpt or claude. teract from the social side, northr from the AI side. weirdly complementary. congrats on the launch rade.

the "describe the edit, it executes" framing is sharp. curious how the agent handles first-time users with no editing history - does it lean on the clarifying questions a lot early on? the cold start problem for understanding someone's aesthetic is one of the harder ones. congrats on the launch @saksham_aggarwal7

the semantic layer approach is exactly right. agents shouldn't guess at infrastructure the same way they shouldn't guess at who the user is. been building the identity side of this problem - the agent knows what to build with insforge, but does it know who it's building for? the cold start for user context is the same problem you solved for infra.

the memory and personality per agent is the right call. most team AI tools feel stateless, like starting over every session. curious about the cold start for new team members though. agents build context over time, but what about day one? been using northr identity for that exact gap, portable user identity that loads into any AI session instantly. combined with cochat's memory layer that could...

the chatgpt prompt mockup use case is underrated. that's the hook format killing it on short form right now. nice that you solved your own problem and shipped it. what's the most popular platform people are generating?

the chat-based building interface is the right call - getting to a working app without touching code is where this space is going. curious how greta handles returning builders. every session starting from scratch feels like the main friction point once you're past the first build. noticed you're using claude under the hood. been thinking about this exact context persistence problem with northr...

@benln @topealabi_ solid launch tope. the sql transparency is the right call - trust killer if you can't verify the query. one thing that moved the needle for me: ran it alongside northr identity so dex had context on my role and what i'm working on, not just my data. the recommendations sharpened up noticeably. data without identity is still half the picture.

I’m a bit sceptical this would actually happen, but it’s an interesting thought experiment. A licence gets you data. An acquisition gets you distribution, incentives, and control over how the product evolves - especially the feed and monetisation. That’s probably the real difference. The part that feels risky to me is exactly what you mention: Pinterest works because it feels human and...

Non-technical founder / operator here. For me, the biggest value from “AI + automation” hasn’t been fancy workflows, but tools that actually force execution and quality without engineers involved. UseMotion has been a big one. It’s basically our real project management system now. Tasks aren’t static - time and priority actually matter, and when things slip, it adjusts. That alone improved...

Hey Product Hunt 👋 I built northr.ai after repeatedly watching traditional planning systems fail the moment real life intervenes. Most tools assume goals, priorities, and time stay stable. In reality, they change weekly. When that happens, plans either get ignored or constantly rebuilt. northr.ai is built around adaptive weekly cycles instead of static plans - you set direction once, and the...