Claude Sonnet 4 is here, and Create just leveled up
We added Claude Sonnet 4 to Create, and it’s already making a big difference. This model is fast, accurate, and seriously powerful. We wanted to share what’s changed, what we’re noticing already, and how it’s quietly pushing us closer to something wild: full-stack app builds with near-zero handholding.
Here’s what we’re seeing already:
1. Claude can now work for hours. That means:
One word: flow.
Claude 4 can now stay active for long stretches, like a real developer working at their desk for hours. It remembers what it’s doing. It doesn’t lose track. And it doesn’t need you to constantly nudge it back on course.
That means it can now do real work over time. Multi-step refactors. Complex revisions across many files. Debugging chains of logic that used to trip up older models.
This is one of the biggest leaps in “agentic” behavior we’ve seen to date.
2. Claude scored a 72.5 on SWE-Bench.
The model handles complex logic; the stuff that usually trips up other AI dev tools, with way more clarity and accuracy. Think: complex marketplaces, dashboards with permissioned user flows, logic-heavy state machines, dense integrations (looking at you, Salesforce). These used to be stretch cases. Now, they’re reliably buildable.
We’re already seeing fewer logic errors, more accurate flows, and smoother results without having to decompose every detail.
3. We’re working toward letting Create build for 10+ hours straight without intervention.
So much of our backend infra has been built to compensate for LLM errors. But Claude 4 just lowered the error rate by a lot. Like, fundamentally.
One of our goals with Create is to support full multi-hour autonomous builds: write the prompt, hit go, and come back to a working app.
Until now, that’s been hard to trust. Most models would fumble after 20–30 tests due to too many small errors, broken flows, or context drift.
Claude 4 shifts that baseline. It’s less error-prone. It keeps more in memory. It doesn’t need to stop and ask as often. That means Create can now reliably stay in build mode longer and actually finish things with fewer handoffs.
It’s the closest we’ve been to fully unsupervised builds that actually ship.
4. Context Retention = No More Repeating Yourself
One underrated shift: Claude’s new memory capabilities. It can now hold onto long-term context and even build “tacit knowledge” about what you’re trying to make.
For non-technical founders, this means no more repeating your app vision every 20 minutes. It remembers what you said, why you said it, and what it means in code.
We’re still figuring out how we want to integrate memory into Create (we're cautious about vendor lock-in). But the possibilities are real. Persistent vision, fewer restarts.
5. Fewer Shortcuts, More Robust Code
One issue with older models: they often took shortcuts. Not maliciously, but because they were trained to output what looked right to humans. Turns out, humans like shortcuts. Computers don’t.
Claude 4 drastically cuts these down. Less brittle code. More real engineering. In Create, this means:
- Fewer runtime errors.
- More reliable production apps.
- Better app quality, even for solo founders.
6. Stripe is now allowed in App Store apps. Combined with Claude 4, that changes the economics for devs.
Apple dropping its 30% tax means one thing: more viable business models.
If you're building a content app, an AI-powered image generator, or anything else with high infra costs you used to have to tack Apple’s 30% cut on top of your margins. It crushed lots of ideas before they could start.
Now you can plug into Stripe directly. And with Claude 4 powering Create to ship faster and more reliably, indie devs and solopreneurs have a much shorter path from idea → app → revenue.
7. Parallel tool execution. Quietly, one of the biggest wins.
Create uses tool calls to edit files, scaffold routes, generate tests, etc. Each of those actions used to happen sequentially. That meant slower builds and higher costs.
Claude 4 supports parallel tool calling. That’s… a massive deal.
It makes the Create engine faster and cheaper overnight. This unlocks a totally different development experience. The same builds now run smoother, quicker, and at a lower cost, without us changing a single line.
Claude 4 Sonnet makes Create better across the board.
Faster builds. Fewer bugs. Smarter results. Lower costs.
The future’s getting closer. And if you’ve got an idea for a product, you’ve never had more leverage to go build it.
We’re excited. And we think you’ll feel it too.
Try it now → inside Create.
