Vibe coding a mobile app with Claude Opus 4.5
How will no-code building evolve as AI coding keeps improving?
Good Morning,
To understand where AI is heading we have to understand where it’s bringing the most ROI. In terms of return on investment in Silicon Valley, that’s directly being seen in engineering teams with remarkable progress in AI coding over the last 18 months (mid 2024 to 2026). In 2025, this also began a redesign of the role of product manager (PM) itself.
In my opinion Jeff Morhous is one of the top writers and guides on understanding all things software engineering at the intersection of Generative AI, like Claude Code and many aspects of AI coding tools. His Newsletter the AI-Augmented Engineer is required reading.
Claude Opus 4.5 was one of the most anticipated models of 2025. It did not disappoint. While Gemini 3 Pro really dethroned the momentum of OpenAI, where GPT-5 related models disappointed, Anthropic I’d argue was the real winner as it will increase its annual recurring revenue (ARR) in 2026 to a degree that is hard to fathom.
Claude Opus 4.5 is considered state-of-the-art, especially for complex, autonomous tasks like coding (SWE-bench leader) and computer use (OSWorld leader).
Increasingly even as a non-developer, there’s more that you can do. Jeff’s writing style is really accessible:
Building a mobile app with Claude Code
How I vibe coded an iPhone app that I now use every day
Everything new in GPT-5.1-Codex-Max
Using NotebookLM to learn new things
Anthropic says software engineering is done (it’s not)
🚀 What makes Claude Opus 4.5 stand out in late 2025/Early 2026?
Deep repository understanding:
Uses a new hierarchical attention mechanism to track dependencies across 500K+ token contexts.Self-correcting execution loops:
Integrates lightweight sandboxed execution during generation, allowing it to run, test, and fix code iteratively— optimized for software engineering.Strong TypeScript/Python/Go support:
Trained on 4.7x more high-quality, post-2023 open-source code than Claude 3.5. Major advance.Tool-use reliability:
Near-perfect accuracy in using linters, type checkers, and test runners via function calling—critical for real-world agent workflows.
Claude Opus 4.5 is a significant leap 🏆
It can do a lot more! It’s so good it will allow Anthropic to grow faster globally and with ARR in 2026 than most expect. You will see what I mean in the coming months. I’m so bullish on them people ask me if they are a sponsor (they’re not, I will disclose all sponsors 🎉). Still, you got to give them credit:
Enhanced enterprise and everyday productivity: Significantly improved for deep research, working with spreadsheets, slides, documents, and complex workflows like financial analysis or multi-step reasoning.
Advanced vision and multimodal features: Anthropic’s strongest vision model yet, enabling complex visual interpretation, multi-step navigation, and tasks involving images or layouts.
Hybrid reasoning mode: Combines instant responses with optional extended thinking for deeper problem-solving, allowing control over effort, latency, and cost.
Dramatically reduced pricing: Costs $5/$25 per million input/output tokens—one-third the price of predecessor Opus 4.1—making frontier-level intelligence more accessible.
Robust safety and alignment: Most aligned frontier model released by Anthropic, with strong resistance to prompt injection and extensive safety evaluations.
200K context window: Supports large inputs for handling extensive codebases, documents, or prolonged conversations without losing coherence.
Not everything is all rosy though, as usage grows surprise usage limits seem to be appearing. Nor are they without controversy in some of the claims the CEO has made, many of which have not come to pass. His claim that AI would write 90% of code within months; ended up being misleading hype as it didn't align with reality. Amodei suggested in mid-2025 that AI could write nearly all code within months.
Claude 4.5 Opus introduces several structural improvements over previous generations, most notably the ability for users to control the model's depth of "thinking" through new parameters.
If you look at how their flagship model has evolved compared with its previous iterations, that’s where the AI talent of Anthropic really shines. In my opinion, Anthropic hasn’t gotten enough credit for being the frontier leader in AI coding. We took many of those benchmarks for granted but we don’t easily understand how this scales to execution in the 2026-2030 period. What comes next keeps the AI boom alive and well.
Anthropic is likely a mega AI startup to go IPO in late 2026 or early 2027, and then the world will know it better. We are AI enthusiasts, we so already know. It could go public at a valuation close to $500 Billion, almost unthinkable just a few years ago.
The most important AI IPO in History
Anthropic might be the biggest AI IPO of 2026, if I’m right - that they overtake OpenAI as the “generational” AI company of the early Gen AI era. OpenAI will have a bigger valuation and raise more, but is delivering less ROI to customers. This is the beginning of what I call “BigAI”. BigTech has major ownership in OpenAI and Anthropic, and Nvidia is deep in xAI.
Anthropic is owned in a big way by BigTech. Google has an equity stake of 15% and Amazon likely even more between 15-20%. Anthropic is able to leverage these Cloud partners in a better light than Microsoft having a 27% equity stake in OpenAI that’s not as collaborative. Anthropic’s Enterprise AI reach is going to be immense.
It’s my top pick for builders who want to work with a product that takes trust and saftey seriously. Anthropic’s AI talent are the highest tier of world class researchers. Their product teams are clearly very focused.
Is Anthropic worth paying for?
The simple fact that this company has AI coding supremacy means its trajectory is very high.
The catch? To use Claude Opus 4.5, you do need a paid Claude account. See pricing.
Claude Pro: $20/month (or $17/month billed annually). This includes the “Model Selector,” allowing you to choose Opus 4.5 specifically for your chats.
Claude Max: $100–$200/month. This tier is for “power users” and provides much higher usage limits for Opus 4.5 than the Pro plan.
Claude Team: $30/user/month (min. 5 seats). Provides administrative controls and shared workspaces.
🙋♂️ What are others saying? 🗳️
There were some definite winners in 2025 and their momentum is going to change the landscape even more in 2026.
Read our primer on Claude for digital products and Creators:
Let’s now get into the vibe coding of mobile apps with these beast. 🔥
Vibe coding a mobile app with Claude Opus 4.5
By Jeff Morhous.
2025 has been an incredible year for model releases, and no doubt 2026 will be even more. Fierce competition has been great for incentivizing new model development that crushes benchmarks.
Claude Opus 4.5 was one of those models. As of today, it’s Anthropic’s flagship model. It beats every other state of the art model on the SWE-bench benchmark, making it a top choice for code generation.
Claude models can be used to write software in a few popular ways:
Back and forth chats in the Claude web app
As a model choice in Claude Code
As a model choice in Cursor
 | A guest post by
|