Welcome to 328 members who joined this week 👋

Hey there, I’m Joshua, and welcome to the NCF Weekly newsletter where we share actionable guidance on building a business as a non-technical founder -- using no-code tools and AI.

Check out our other newsletters: Tiny Empires and Seedstrapped

May has been a busy month. In the last few weeks we've had a new Claude model, a new ChatGPT model, a major Cursor release, and Google Labs quietly launching a no-code builder that nobody is talking about yet.

Most of the coverage has been aimed at developers. This edition is aimed at you.

Here's what actually dropped, what it does, and whether you need to care about it.

Sponsored by

Most App Store developers don't realize their earned revenue sits locked for 30-45 days before Apple pays out. That's cash you can't use for ads, features, or tools right when momentum matters most.

AppCap advances up to 85% of your earned App Store revenue within 48 hours after approval. No equity given up, no long-term debt, and no commitment.

Rates start from 0.34% for a 5 day advance. 

Anthropic released Opus 4.7 in mid-April. If you're using Claude to write, think through problems, or run longer research tasks, this is a meaningful upgrade over 4.6.

The headline improvement is coding -- Opus 4.7 cleared 70% on CursorBench versus 58% for 4.6. But the changes that matter more for non-technical founders are subtler. It follows complex instructions more accurately over long sessions, produces better-quality docs and slides when you ask it to, and handles ambiguous requests without going off-script in the same way 4.6 sometimes did.

Vision also improved. If you're feeding Claude screenshots, mockups, or design references, the resolution it processes is now higher, which matters when you're asking it to replicate or analyse a UI.

You don't need to do anything to get this. If you're on Claude Pro, you're already using it.

OpenAI dropped GPT-5.5 on April 23rd. Unlike most point releases, this one was a ground-up rebuild rather than an update to the same architecture.

The short version: it's faster than GPT-5.4 despite being more capable, which shouldn't be possible but apparently is, because it was co-designed with NVIDIA's latest hardware. It's also natively multimodal, meaning it handles text, images, audio, and video in a single system rather than separate models stitched together.

For founders using ChatGPT to build and automate things: GPT-5.5 is noticeably better at agentic tasks -- long, multi-step workflows where you hand it a messy brief and expect it to figure out the sequencing. One engineer described it as "the first model I've used that understands the shape of a system."

Whether you use Claude or ChatGPT day-to-day is honestly a preference thing at this point. Both are excellent. The gap between them has never been smaller. If you're already a ChatGPT user, it's worth revisiting what you've been using it for -- this version handles longer, messier tasks than most people are currently giving it. Try handing it something you'd normally break into three separate prompts and see what happens.

Cursor shipped version 3 on April 2nd. This is the biggest release since the product launched.

The main change is the Agents Window -- you can now run multiple AI agents working in parallel across different parts of your codebase. They work in isolated environments, open pull requests when they're done, and can be triggered from Slack or GitHub while your laptop is closed.

If you're not a developer, this might sound like it has nothing to do with you. But if you're using Cursor to vibe-code a product -- which a growing number of NCF members are -- version 3 changes the speed at which you can iterate. Instead of one conversation building one feature at a time, you can run several builds simultaneously and see which approach works. If you've never used Cursor before, think of it as a chat interface where you describe what you want to build and AI writes the code for you -- similar to Lovable or Base44, but with more control over the output.

It integrated Claude Opus 4.7 on day one.

On April 29th, Google Labs launched Opal in public beta. It's currently US-only, but worth understanding now.

Opal lets you build AI-powered mini-apps by chaining together prompts, models, and tools using natural language and a visual editor. No code. You describe what you want the app to do, and Opal builds it.

The difference between Google Opal and something like Lovable or Base44 is the use case. Opal isn't trying to build you a full SaaS product. It's for building small, focused tools -- a custom research assistant, an AI content workflow, an automated report generator. The kind of thing you'd currently pay $20/month for a SaaS tool to do, but built exactly the way you need it in an afternoon.

It's early and experimental. But Google Labs has a habit of quietly releasing things that end up reshaping how people work. Keep an eye on this one.

A few practical things to take away from all of the above.

If you're building with Claude, Opus 4.7 is your default now. The improvement in instruction-following over long sessions is real and worth experiencing.

If you're vibe-coding anything with Cursor, version 3's parallel agents mean you can test multiple approaches without waiting for one to finish before starting the next.

If you've been relying on a SaaS tool for a specific workflow -- a report you generate, a brief you write, a research task you run -- it's worth opening Google Opal and spending 20 minutes seeing whether you can build a version that fits your needs better.

The bigger pattern here isn't any individual release. It's the pace. In the last four weeks, four major tools updated or launched. Each one is meaningfully better than what it replaced. If you haven't revisited the tools in your stack since the start of this year, some of what you're paying for monthly is already outdated.

GPT-5.5 and Claude Opus 4.7 are now going head to head in Cursor, which supports both. Most developers using Cursor say Opus 4.7 handles complex multi-file tasks more accurately. GPT-5.5 has an edge on pure terminal/command-line work.

Claude Cowork -- the desktop agent that controls your Mac or PC on your behalf -- is now generally available on both macOS and Windows on the $20/month Pro plan.

Gemini CLI launched a free tier with 1,000 AI requests per day. For founders who want to experiment with AI in their terminal without paying, it's a zero-cost starting point.

Hit reply and tell me which of these you're actually using. Always curious what the NCF community is building.

Get full access to all our resources with Pro Membership for a one-time payment of $149.

Sneak peek at what’s included 👇

💰$50k in Perks - Access Exclusive offers from the top no-code tools who have partnered with us. Including $15k waived Stripe fees, $500 Bubble credit + $1000 Coda credit View all

💻 No-Code Operating System - Advanced Notion template that replaces all your productivity tools. Everything you need to launch and manage your business from one page. Learn more

🎓 Course: Tiny Empires Method - Learn how to build a 6-figure business that works around your life and not the other way around. Learn proven frameworks to stop wasting time and start making money. Learn more

🎓 Course: Sales for Introverts (and people who don’t like selling) - Learn how to sell in a way that fits your personality and delivers consistent, reliable revenue for your busines. Learn more

🛠️ 75+ Curated No-Code Courses and Resources - We’ve curated 2000 no-code videos into 75+ easy to navigate courses. Save yourself hours of watching Youtube videos that don’t move your knowledge forward. View all