Yash16

I have tried a lot of the agenting frameworks, especially Ignore, Open AI Agent SDK, and Language Graph, but I am not happy with all of this framework. I am still missing speed, performance, and the capability, like Hermes and Open CLO Agent.

I am thinking too: if I do not have a good, underrated agent framework, please suggest. Otherwise, I am thinking of building my own agent SDK. I know this will be best of the time, but I am thinking of it now, seriously, because I am building a lot of projects. I am thinking that I can create my own personalised agent at the end of the work that I can change according to my use cases. It's

thanks to @kiro

Can I add a deploy option for thisa app on my platform https://clawsifyai.com ?

try google gemini model for ui

Has anyone tried Flux 2 Klein? Lately, I’ve stopped chasing new models. I’m genuinely happy with Nano Banana Pro—the results are solid, so I’m building the entire app around just this one model.

https://picxstudio.com

Are you using a coding package or graph-based approach for this, or is it standard image generation? I’ve seen very similar results on https://picxstudio.com —it’s an AI image generation platform.

I agree with you — the model itself is strong. The real challenge is prompting. Not everyone knows how to write detailed, high-quality prompts, and that’s exactly the gap we’re solving with https://picxstudio.com .

We’re building curated prompt collections and an AI agent that uses prompt templates + RAG-based knowledge to generate better prompts automatically. In some cases, structured (JSON-style) prompts give much better results — but getting those right consistently is hard and honestly very trial-and-error.

Our goal is simple: users shouldn’t have to be prompt experts to get high-quality results.

Bro, you got laid off — I actually quit my job.

They kept pushing extra work with zero upside, so I made a call: If I’m going to grind harder, it might as well be for myself, not someone else.

That’s how my indie hacking journey started — and honestly, you can start too.

Right now is the best time to build your own AI SaaS. If you’re into content creation but feel stuck, don’t overthink it — use AI to move faster. It’s not cheating, it’s leverage.

I’ve just started myself. https://youtu.be/BOqNr6a3N_0?si=ntsT22hJGGsbXIqQ

One real talk thing though: This path isn’t easy. Building is the fun part — but marketing + sales matter more than code. Focus there, or the product won’t matter.

If you’re serious, start now. Timing won’t get better.

By 2026, I plan to stop writing code myself and go all-in on AI coding tools like Claude. My focus will shift entirely to marketing, building products, and scaling my own company. AI is changing the game—soon, one person will be able to build a billion-dollar company with AI doing the heavy lifting.

This works fine for casual use, but for professional, high-quality result especially with Nano Banana Pro —you need proper tools. Platforms like https://picxstudio.com and Higgsfield AI deliver way better output with advanced controls. That’s why I use those—and honestly, image editing directly on X just doesn’t cut it.

Do you support integrations or provide open-source APIs? I want to integrate this with https://picxstudio.com and enable easy, browser-based downloads in multiple image formats.

how ?

Why do you think that?

Great product—love the domain and the Nano Banana model choice.

I’m building picxstudio.com , focused only on image generation (no video for now). Since you’re doing video, maybe we can feature each other—users often want both.

I’ll probably explore video later when VEO-5 drops. Let me know what you think

The red looks great overall. Changing the text to white might improve contrast and readability. I experimented with red in my app, but it didn’t work visually, so I moved to a snowfall theme.

I like this because most of the time I need random stuff—numbers, quick searches, or ideas—and this helps instantly.

What’s the purpose of creating this?

I built an agent that uses RAG to fix the "low-quality" issue common with short AI image prompts.

It retrieves relevant descriptors from a database of 500+ professional prompts to expand your simple input into a high-detail prompt before generation. This provides more stylistic variety than a standard system prompt.

Yes,earlier we were fully focused on B2C for the image-editing app. At the time, AI photo transformation was trending hard, so adoption was fast.

But a big chunk of users started misusing the paid plans—jailbreaking prompts and pushing boundaries. Even though we’re using Google’s Nano Banana model (which doesn’t generate full NSFW), it can still produce partial outputs like lingerie-style images. That still feels wrong, especially when people misuse images that aren’t theirs.

Because of that, shifting to B2B makes more sense. The Nano Banana Pro model really shines there—strong text rendering, true 4K quality, and it solves real business use cases. I’m confident we’ll attract higher-quality users and build something more sustainable in the B2B segment.

I was thinking the same thing. Once we remove free credits and move fully to paid plans, we’ll automatically reduce NSFW usage. We can also increase pricing—maybe even double it.

B2C is tough because pricing becomes a blocker; users just want to casually edit images and won’t pay much. But if we shift focus to B2B, the pricing becomes affordable for them because it’s tied to real business value. Brands don’t see it as editing images—they see it as saving time and money.

I built picxstudio.com, an AI image generator. Got $850 in revenue. Then users started abusing it for NSFW content.

The problem started 3 years ago with my first image gen app. Used Stable Diffusion with ComfyUI on Azure GPU VMs. Users from Asia uploaded photos of women trying to generate nude images. I tried every safeguard. They jailbroke all of them. I shut it down.

This year I tried again. Used Replicate and Fal AI instead of self-hosting. Built proper UI. Used Nano Banana model for headshots and viral images. Small team curating prompts. Started getting traction.

Same problem came back. NSFW abuse.

What I tried:

Content moderation APIs. Users found edge cases. Banned keywords. They used synonyms. Removed free tier. They paid anyway. Manual review. Doesn't scale. The truth is B2C image generation attracts the wrong crowd. The product wasn't the problem. The market was.

Now pivoting to B2B. Businesses want product photos, branded visuals, AI headshots. They don't upload random photos of women.

What we're building different:

Brand Kit: save your brand URL, we extract logo/colors/fonts, every image stays on-brand 4 Variations: generate using GPT, Claude, Gemini, Mistral. Each has unique strengths. Team workspace without per-seat pricing No monthly subscription. Pay for credits, not monthly fees. Questions for HN:

Anyone dealt with NSFW abuse in consumer products? How did you handle it? Anyone did B2C to B2B pivot? What worked? Any content moderation tools I'm missing? If you failed at something I'm about to try, I want to hear it.

I have made custom GPT for Lego avatar!, Brick Avatar, a GPT model that transforms photos into custom Lego-style characters. I focus on creating playful and unique Lego avatars from your images, respecting privacy and personal details. Excited to build some fun Lego transformations with you!