The booking page. The invoice tracker. The little CRM. The form tool that dumps into a spreadsheet. Most of them do one simple thing, and most of them charge you $15 to $50 a month to do it, forever.

Here is what changed this month. Building your own version of those tools with AI is not new. You could do it a year ago, but it was a slog: lots of back-and-forth, half the features broken, a fragile result you did not trust. What changed is that it got easy. You describe the tool once and get back something that actually works, in a single sitting, and you own it outright.

The jump in capability is real and recent. Anthropic's newest model came back online on July 1 after a government ban, and it now one-shots apps that took a hundred prompts a year ago. People are building real working software in the plain chat window, no coding tool required.

This week: how to replace one of your paid apps with something you build in an afternoon, and the model jump that made it easy.

The Old Way:

You need a small tool. A way to track invoices, book clients, log leads, collect form responses. You do not want to build software, so you rent it. You find a SaaS product, pay the monthly fee, and accept its limits: the features you do not use, the ones you wish it had, the price that creeps up every year.

Stack a few of these together and you are paying $100 to $200 a month to rent simple tools, none of which quite fit how you actually work, all of which you will pay for again next month and every month after. The math never stops.

Building your own was never realistic, because building even a simple working app meant either learning to code or hiring someone. So you kept renting.

The Replacement:

You describe the tool you need. AI builds a working version. You host it for free and own it outright.

The change that makes this real is capability. Today's top models, Claude Opus 4.8 and the newly returned Fable 5, can take a plain description and return a complete working app rather than a broken shell. Simon Willison, a well-known developer, built a fully working software library in the standard Claude.ai chat window, not a coding tool, in one session. Anthropic's own framing for its newest models: apps that took a hundred prompts a year ago now get one-shot (source: Anthropic).

For a simple business tool, an invoice tracker, a booking page, a client list, a form that saves to a spreadsheet, this is well within reach today, from one prompt, with no coding knowledge.

What this replaces: a $15 to $50 per month subscription, times however many simple tools you rent.

What it costs: a Claude plan you may already have, plus free hosting.

Setup time: an afternoon for your first one. Minutes for each one after, once you know the pattern.

The Pro section below includes:

What you get: the simple tools you rent, rebuilt to fit you and owned outright.

→ List every software subscription you pay for. Circle the ones that do just one simple job. Those are the ones you can replace. Use the Subscription Audit and this weeks workflow.

→ Read the Big Story. The reason all of this is suddenly possible is a single model that came back online last week.

→ View the new Monthly resource: The Claude Code Design Systems Pack: 10 named design systems with exact colors, fonts, and spacing tokens that stop Claude building the same generic purple-gradient website every time.

🔒 Pro members get all 6 build specs, the make-it-work follow-ups, and the free hosting steps. Build your first replacement this week.

"I set up one automation from the toolkit last weekend and saved 6 hours in the first week alone. It feels like having an extra team member running in the background."

Elena M. DTC Founder.

Anthropic quietly shipped Claude Science, a dedicated research workbench, and bundled it into every paid Claude plan at no extra cost (macOS and Linux beta).

It is built for the kind of work that eats a day: pulling together sources, reasoning across dense material, and producing a structured, cited writeup you can actually use. Think of it as a focused workspace for deep research, separate from the regular chat, so a serious research task does not get lost in a conversation thread.

Why this week: it is genuinely new, it is free if you already pay for Claude, and it targets a universal pain. Anyone who does market research, competitor analysis, due diligence, or has to get up to speed on a complex topic fast has a use for it. If you pay for Claude, you already have it. Check out the link to find it.

🔒 Implementation checklist is for pro members. Get a breakdown plan to implement AI into your workflow.

🔓 Claude Fable 5 is back after an 18-day government shutdown. The US pulled Anthropic's most powerful model on June 12 over cybersecurity concerns, then lifted the export controls on July 1, and it is live again on Claude.ai, Claude Code, and Cowork (source: Anthropic). Why this matters for you: it is the engine behind this week's workflow, and the Big Story below.

🧭 Google's Gemini 3.5 Pro is finally rolling out after missing two deadlines. After promising it in May, then June, Google began a gradual release in early July. Its headline feature is a 2-million-token context window, the largest of any frontier model, double what Claude and GPT offer. Why this matters for you: for jobs that involve feeding an AI huge amounts of material at once, an entire codebase, years of documents, long histories, this is now the only frontier option, and it is priced below its rivals.

🌏 A cheap Chinese open model is closing the gap with the top US AI. Z.ai's GLM-5.2 is drawing attention for coding and agent performance approaching Claude Opus 4.8 and GPT-5.5, at a fraction of the cost, and it is open-weight. Why this matters for you: frontier-level AI is getting cheaper fast. More capable models at lower prices means the kind of building in this issue keeps getting more accessible.

📉 Meta cut about 8,000 jobs while reassigning 7,000 to AI teams. Part of a wider wave of over 100,000 tech job cuts in 2026, many tied directly to AI automation (source: Crescendo). Why this matters for you: the work being automated first is repetitive knowledge work. Owning the tools, like building your own instead of renting, is how you stay on the right side of that shift.

On June 12, the US government did something it had never done before: it pulled a working, publicly available AI model off the market. Anthropic's Claude Fable 5, its most capable model, went dark for everyone over cybersecurity concerns. Eighteen days later, on July 1, the export controls were lifted and Fable 5 came back online (source: Anthropic).

The ban grabbed headlines. The capability is the real story.

Fable 5 is built for long, autonomous work. It holds a million tokens of context, an entire mid-sized codebase, in a single session, and it can run for hours on one task, planning, building, and checking its own work as it goes. Stripe gave it a 50-million-line codebase and it did a migration in a day that would have taken a team of engineers over two months by hand (source: Anthropic). A Wharton professor with early access called it a genuine leap over every model he had used, and built a full academic paper from a single prompt plus one piece of feedback.

Here is what that means for someone who is not an engineer. The wall that kept normal people from building their own software was never really about ideas. It was that building anything that actually worked required knowing how to code or paying someone who did. Fable 5 lowers that wall to the ground. You describe what you want, it builds the working thing, and increasingly it gets there in one shot.

An Anthropic engineer put the shift plainly this month: with a model this capable, the limit on what you can build is no longer the model. It is how clearly you can describe what you actually want. Backed by a study of 400,000 real sessions, the finding was that people still make around 70% of the important decisions, and the clearer your thinking going in, the more the model does for you.

🔒 The full breakdown (what one person can now realistically build and own, where the real leverage is for a small business, and how to think clearly enough to get the most out of a model this capable) is for Pro members.

New This Week:

Prompt Library Updates:

NEW Resources:

Each resource lives permanently in your Pro account. Use them whenever you need them.

The simple tools you rent were built for the days when you could not make your own. Those days just ended. You will not replace every app, and you should not try, but the long tail of simple subscriptions that quietly drain a few hundred dollars a month is now yours to build and own.

If you do one thing before next Thursday, pick the simplest app you pay for and rebuild it. One afternoon, and you stop paying for it for good.

👇 What’s behind the paywall:

Pro members deploy AI in their work an average of 5-8x more often than free readers (based on reply data from past issues). The difference is having the exact setup, not the concept.