Does the $100 AI Ultra plan offer good usage of Opus 5.5?

Hi, I’m a software developer, but I like telling stories. I was thinking of making scientific stories for kids. I tried making them as comics using AI, but I’m wondering if someone would accept this quality or if it’s too low.

Could you check it out and give me your feedback?

https://science-stories-episode-1.vercel.app/book/comic/

Hi,
I have been working on a customer support agent, that can control app UI when the app owner embeds it inside his app, I designed it to work primarily on flutter and react native apps. My question: Is there really a market need for this?

The scenario is:

"A customer asks why his latest order charged him $34."

Instead of immediately escalating to a human, the agent investigates inside the app, but first it asks for permission and states its plan. After that It navigates to the billing and charge details screens, reads the actual order data, and instead of returning plain text, it returns a rich UI breakdown: food subtotal, service fee, delivery fee, and tip.

Then the user says the delivery fee changed than the one he saw before placing an order.
The agent does not guess. It checks the charge again, asks the customer what previous value they saw, and then reports the issue with the relevant context.

When the user asks when he will get a reply, the ai honestly replies that it doesn't have specific timeline, but the issue is reported.

This is the behavior I think customer support agents should follow:

1. AI should investigate before escalating.
2. It should use the actual app state, not generic support scripts.
3. It should explain findings clearly, ideally in custom UI, not just text.
4. It should ask follow-up questions when information is missing.
5. It should report actionable issues with context.
6. It should escalate to a human only when it really cannot resolve or investigate further.
7. It should be honest when it doesn't know something, or isn't in its knowledge base.

So this won't be another “chat with bot and maybe it tells you where to tap.”
but “An agent can understand the current app, navigate it, inspect the right data, and help the user complete the task.”

There is also a dashboard where you can see analytics, handle escalations, automation, etc..

So what do you think?

Hi,
I have been working on a customer support agent, that can control app UI when the app owner embeds it inside his app, I designed it to work primarily on flutter and react native apps. My question: Is there really a market need for this?

The scenario is:

"A customer asks why his latest order charged him $34."

Instead of immediately escalating to a human, the agent investigates inside the app, but first it asks for permission and states its plan. After that It navigates to the billing and charge details screens, reads the actual order data, and instead of returning plain text, it returns a rich UI breakdown: food subtotal, service fee, delivery fee, and tip.

Then the user says the delivery fee changed than the one he saw before placing an order.
The agent does not guess. It checks the charge again, asks the customer what previous value they saw, and then reports the issue with the relevant context.

When the user asks when he will get a reply, the ai honestly replies that it doesn't have specific timeline, but the issue is reported.

This is the behavior I think customer support agents should follow:

1. AI should investigate before escalating.
2. It should use the actual app state, not generic support scripts.
3. It should explain findings clearly, ideally in custom UI, not just text.
4. It should ask follow-up questions when information is missing.
5. It should report actionable issues with context.
6. It should escalate to a human only when it really cannot resolve or investigate further.
7. It should be honest when it doesn't know something, or isn't in its knowledge base.

So this won't be another “chat with bot and maybe it tells you where to tap.”
but “An agent can understand the current app, navigate it, inspect the right data, and help the user complete the task.”

There is also a dashboard where you can see analytics, handle escalations, automation, etc..

So what do you think?

Demo for the libary

Every support tool I've used does the same thing, you open a chat bubble and you talk to a dummy bot, and when things get serious it redirects you to a human.

I built an SDK that takes a different approach: the AI reads the React Fiber tree at runtime. Every Pressable, TextInput, Switch, Picker is discovered automatically. When it needs to act, it finds the element's onPress handler from the Fiber node and calls it directly. With no screenshots, OCR, nor selectors.

You can talk to it by text or just by voice.

And the usage is dead simple, just wrap your app with it, and that's all. You can set your backend as the ai source:

<AIAgent
  proxyUrl="https://myapp.vercel.app/api/gemini"
  proxyHeaders={{ Authorization: `Bearer ${userToken}` }}
  provider="gemini"
  voiceProxyUrl="https://voice-server.render.com"  // only if text proxy is serverless
  navRef={navRef}
>

  {/* your app */}
</AIAgent>

or for local development you can use the key directly

<AIAgent
  provider="openai"
  apiKey="YOUR_OPENAI_API_KEY"
  // model="gpt-4.1-mini"  ← default, or use any OpenAI model
  navRef={navRef}
>
  {/* your app */}
</AIAgent>

Default is Copilot mode — AI navigates and fills forms silently, pauses before anything irreversible (place order, delete, payment). You can also add aiConfirm to specific buttons as a code-level safety net.

You wrap your app with one component, and the AI can:

1. Answer questions from a knowledge base.
2. Navigate to the right screen and show the user where things are.
3. Fill forms on behalf of the user (pauses before anything irreversible).
4. Complete full flows end-to-end with permission.
5. Escalate to a live human over WebSocket when it can't help.
6. Do all of this by text or voice in real time.

It also works as an MCP server: for example connect Claude Desktop to your emulator and test your app in plain English. No Detox setup needed.

You can call custom business logic you register with a useAction hook: like checkout, apply promo code, or cancel an order without needing a visible button for it and ai agent will trigger it when needed.

There's an optional dashboard too, which provides escalation inbox with live WebSocket chat, interaction analytics, heatmaps, automations, etc..

Links

• npm: https://www.npmjs.com/package/react-native-agentic-ai
• GitHub: https://github.com/mohamed2m2018/react-native-agentic-ai

Tell me what do you think, Will you use such library in your apps, and how do you think it can be better?