Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- Claude Code hit $1 billion fast by transforming real developer workflows.
- Agentic coding built my complex iPhone app in just 11 days.
- Early command-line access gave Claude Code a huge adoption edge.
Anthropic’s Claude Code, a programmer’s tool, just reached $1 billion in revenue a mere six months after its release. I’ve been managing and selling programming tools since I was a mere pup. Take it from me, this has never been a vibrant market. And here’s Anthropic, pulling in a billion since May.
Also: Anthropic just released Claude Opus 4.5 – here’s how it stacks up against other leading models
Let’s put this in perspective. The IT world’s last mega-shift was cloud computing, spearheaded by Amazon’s AWS, which launched in 2006. According to Forbes, AWS reached $500 million by 2010 and was expected to reach $1 billion by the end of 2011. Not to diminish the incredible value AWS has provided for companies worldwide (including my own), but it still took six years for the last major IT earthquake to drive demand into unicorn territory.
Why Claude Code
So what makes Claude Code so special? Speaking personally, Claude Code wrote 19,647 lines of code and 5,139 lines of documentation, created 63 user interface views, and generated 114 source code files for my new iPhone app. Oh, and it did all that in the last two weeks, in and around my primary writing work.
We’ll come back to my app in a minute. First, though, let’s position Claude Code in the market.
Also: The best free AI for coding – only 3 make the cut now
Parallel evolution is a theory that explains why similar traits in very dissimilar species tend to evolve simultaneously over time. It helps to explain how agentic coding capabilities from companies like Anthropic, Microsoft, OpenAI, and Google all seemed to arrive in May of 2025. Prior to May, coding was primarily confined to fairly powerful autocomplete capabilities and chatbot responses to pasted-in code blocks.
But in and around May, each of the companies introduced their agentic coding tools. The difference between the previous generation (where ‘previous’ refers to April, not May) is that the new technology enables the tools to autonomously perform a series of tasks on behalf of the programmer. For example, you could ask the AI to redesign the UI or add a backup and restore system (both of which I did), and the AI will go off, work on the code, and report back.
Also: Microsoft’s new AI agents won’t just help us code, now they’ll decide what to code
Initially, Codex from OpenAI and Jules from Google would perform this work, but only using a web interface and only for coding that resided in a GitHub repository. While many programmers manage their code in GitHub, that approach made testing and tinkering with the code cumbersome.
Microsoft’s agentic offering, GitHub Copilot, took a different turn. It integrated the chat interface into the VS Code programming environment. Meanwhile, Anthropic just made Claude available from the command line, so no matter what terminal program a programmer is using, the coder can simply issue Claude Code prompts just like, and along with, any other command line command.
While all the agents eventually added IDE integration and command-line operations, this early command-line ease of use seems to have driven Claude’s early adoption. I have a group of coder friends with whom I regularly chat over Zoom. While I was making major strides with Codex integrated into VS Code, they were all singing the praises of how easy Claude was to use from the command line and how productive it was making them.
I built an iPhone app
So, in mid-November, a little more than two weeks ago, I decided to see what all the fuss was about. I wanted an iPhone app to help me organize my 3D printer filament workflow, and I decided to write my own (as one does).
Initially, I utilized Claude Code, integrated into the Apple Xcode development environment, leveraging Apple’s Code Intelligence. It was more like Code Meltdown. Xcode kept hanging and crashing with Claude Code embedded. To be fair, when I switched to OpenAI’s Codex within Xcode, I encountered the same problem.
Also: I’ve tested free vs. paid AI coding tools – here’s which one I’d actually use
However, when I used Claude Code in the Mac’s Terminal app, everything took off. The development experience was fluid.
I have programmed iPhone apps before. Back in 2008, right when the App Store launched, I wrote 40 pinpoint iPhone apps. But that was 17 years ago, and it was in Objective C. Most iPhone apps are now written in Apple’s Swift language and SwiftUI framework. I had never coded in that environment.
And, to be honest, I still haven’t. I have a working iPhone app with many complex and advanced features, but I did not write a single line of code. I did spend 11 days, on and off, managing and directing Claude Code, but it wrote the entire thing.
My app is not a simple project. It includes the ability to read and write NFC tags using the iPhone’s hardware. It can take pictures and import them from the Photos app. It can perform super-fast searches and data filtering. It can sync photo and database data with iCloud and other Apple devices on the same iCloud account. It can make and restore backups. It can analyze photos to create color swatches.
Also: So long, SaaS: Why AI spells the end of per-seat software licenses – and what comes next
It also implements a Quick Move workflow, which involves multiple NFC scans that dynamically update the database, saving a ton of time in managing my inventory.
I have seven active 3D printers. Four of them support multi-spool holders for multi-color printing. At any given time, I could have up to 18 different filament spools assigned to the printers. I also have an inventory of 100 filament spools not on the printers. My problem was, as I worked on projects, it was very hard to keep track of which spool was where, how many spools of a given material or color I have, and keep an updated inventory.
Before this tool, I made voice notes into my Apple Watch, and then, when back at my computer, I processed them into a crude Notion database. But the inventory always slipped out of date, because a simple decision to move a spool from one printer to another might never get recorded.
Also: Is that an AI image? 6 telltale signs it’s a fake – and my favorite free detectors
My app uses NFC tags, situated on every spool, on every storage rack shelf, and on every 3D printer. Once a spool’s information is recorded, data such as vendor, color, material, thickness, and spool size remain associated with that spool. When I move a spool, I just tap it with my phone, and it reads the NFC tag. Then I tap the destination NFC tag with my phone. The location is automatically updated. I don’t even have to touch the phone’s screen. It’s fast and seamless.
A similar app could be used to manage all kinds of collections. I built something that tracks my 3D printer filament, but the same kind of project could be used for your wardrobe, book collection, music collection, knitting patterns, sheet music, yarn stash, fabric stash, hobby management, model kits, LEGO sets, photo management, spare parts, and more.
I started with the hard stuff
No matter the coding target or development environment, making lists of items with detail pages isn’t that difficult a project. It can be challenging to populate your lists with all the information you want while also achieving attractive formatting. Overall, though, it’s a fairly straightforward process.
Also: Google Workspace now lets you create AI agents to automate your work – how to get started
However, reading and writing NFC tags using proprietary Apple APIs is significantly more challenging. So is interfacing with the photos subsystem to take and manage photos. And so is syncing all that data to iCloud, especially when syncing photos as well.
So rather than start with the easy stuff, I set out to see if Claude could handle the harder things. If it could read and write NFC tags and take and save photos, it could surely do the rest.
And it did. By the end of my first day’s run (approximately 4 hours), I had a simple test program that could read NFC tags, take photos, and sync to iCloud. I didn’t write any code. Claude Code did it all.
Now, to be clear, I didn’t just tell it to give me an app. I have had to manage it constantly. I’ve had to cajole it into doing what I wanted. I’ve had to send it back to try again (sometimes 20 and 30 times) until it actually got something to work. Just because I didn’t need to write the individual code lines doesn’t mean this wasn’t work.
It has taken all my years of product marketing and product design skills to supervise and manage Claude Code and guide it into building my app.
Also: Amazon wants to demolish your tech debt like it did this AWS server – here’s the plan
That said, across spool management, quick move workflow, colors view, machines and locations, the list system, entity picker sheets, settings, backup and restore, NFC tools, iOS shortcuts, NFC tag system, core data and persistence, and iCloud support, there are about 365 individual features. The spool management system has 72 features alone.
I know my programming speed. I don’t normally get to spend whole days programming. When I do get the time, it’s an hour or two here, an hour or two there. So any given feature usually takes me a weekend to a week to produce, and that’s in a language I’m fully familiar with. I haven’t coded in Swift, so it’s quite reasonable to expect that a given feature could take significantly longer. The NFC prototype would probably have taken me a month on its own.
With 365-ish features, I could never have written this app on my own. The best-case scenario would have been about 365 weekends, or, if I worked on it a bit every day, about two years before I had anything to show for my efforts. That’s the best case. Instead, I’m just 17 calendar days in. With only 11 days of actually working with Claude Code, I have a fully working app.
Also: Not a developer? AI could still take your job, MIT study finds
Sure, there are a few features I still want to add. For example, I don’t have a spool retirement feature in place. I want to be able to tell the app that I’m done with a spool, whether that’s because I’ve used it up, have to discard it (sometimes filament goes bad), or donate it.
I have a few other features I’d like to add. I’m also thinking of moving the app to the Mac and Apple Watch. But in terms of whether the app works well enough to help me manage my spool collection, it’s working now.
$1 billion, $100 at a time
Like most AI tools, Claude Code is sold based on usage tiers. I tried the $15 plan, but used it up after about five hours of work. So, I upgraded my plan to the $100/month Max plan, and so far, I haven’t hit any walls.
Also: Amazon says new DevOps agents need no babysitting – you can try them here
While some coders use higher-level tiers, everyone I’ve talked to is using the $100/month tier. So if Claude Code brought in a billion bucks over six months, and coders are spending $100/month, that’s about 1.6 million coders using Claude Code. That’s one heck of a lot of adoption in a very short time.
But look at what it can do. My case is a good example. But remember, it takes a lot of management. A lot of testing. A lot-lot-lot of patience. And Claude Code makes massive mistakes on a regular basis. So it requires a lot of cajoling, persuading, explaining, and sometimes just starting over. It is an astonishingly powerful productivity tool, but it won’t create an app for you just because you wish hard. You definitely have to work it.
Even so, we can see how Claude Code is getting such astonishing uptake. My app is real. It’s not some template-based prototype. I don’t see complete neophytes being able to do product development or truly succeed while coding with Claude Code. But people with some experience? For them, this is an astounding level of force multiplication.
Also: I did 24 days of coding in 12 hours with a $20 AI tool – but there’s one big pitfall
What about you? Have you tried any of the new agentic coding tools that landed this year, and did they change how you build software? Do you see Claude Code’s rapid adoption as a sign of a major shift, or is this surge mostly about early excitement among developers? How do you feel about Anthropic pairing Claude Code with Bun, and do you think faster JavaScript tooling will meaningfully affect your own projects? If you’ve experimented with coding AIs in the terminal, VS Code, or Xcode, which workflow felt most productive for you? Let us know in the comments below.
You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.