Hello, welcome to my thoughts on tariffs and inflation. Here's a bit of background: I sell a digital product online and have a significant number of Canadian and European customers, along with others around the world. My business extends well beyond the United States.

When the idea of tariffs was introduced, I immediately started researching. I realized I needed to read the USMCA and examine Canada's list of retaliatory tariffs on U.S. goods. I had to consider the implications if the U.S. imposed tariffs on Canada, Mexico, or the EU, and if retaliatory tariffs would affect my product.

The good news is that the USMCA prohibits tariffs on digital goods. Digital goods are inherently difficult to tariff because they don't cross borders in the traditional sense or go through customs. It's challenging to imagine a scenario where digital goods are specifically targeted without a prolonged trade war or a significant effort from one side. Currently, neither the U.S. nor Canada has the mechanisms to impose tariffs on digital goods effectively.

The real concern isn't the potential 25% tariff itself—I'd be willing to offer a 25% discount to my customers to offset it. The core issue is the impact on customer confidence in the EU and Canada when considering purchases from U.S. companies. If the U.S. starts imposing tariffs broadly, it could erode trust and confidence in U.S. businesses.

To address this, I need to develop a marketing campaign that communicates our stance against tariffs. We could emphasize that we absorb the tariff costs and offer discounts to our customers who are affected by them.

Even if digital goods aren't directly targeted by tariffs, U.S. suppliers in Canada and the EU might face negative repercussions. Companies may instruct their procurement teams to be cautious about buying from U.S. companies if alternatives exist, due to the uncertainty tariffs introduce.

Many claim that tariffs are inflationary. However, the term "inflation" is often misused. Inflation should refer to what we now call "monetary inflation," a purely monetary construct. It's important to distinguish this from price changes, as inflation is the result of market dynamics.

In conclusion, while digital goods might be safe from direct tariffs, the broader implications on customer confidence and international trade relationships are significant concerns that need addressing.

Hello! Today, I want to discuss something intriguing that caught my attention a couple of months ago and has been on my mind ever since. It appears that Google has found a way to scrape data from an iTunes music library on the same device through one of their apps. However, I'm uncertain which app is responsible, and it doesn't seem to involve any information sharing or consensual data sharing with Google.

I have a specific set of music, primarily tracks I copied into my iTunes music library years ago, which are now part of my Apple Music library. These tracks have never left the app, and I don't search for these artists or songs on YouTube. Interestingly, this niche music from my library began appearing in YouTube notifications and recommendations around November or December, seemingly out of nowhere.

YouTube started suggesting, "Hey, you don't listen to music on YouTube. Why don't you try YouTube Music? We already have all of your music from your library. You could switch right now and start listening." This was surprising because my music was siloed in iTunes and Apple Music. I'm unaware of how Google could access it without some OAuth mechanism.

In iOS app development, there's a mechanism to add a "Sign in with Apple Music" button, allowing third-party apps to access data. However, it seems Google has found a way to scrape song names and artist names from the entire iTunes music library, possibly through one of their installed apps on my device. This data appears to be used to promote YouTube Music, a competitor app.

It's puzzling because one would assume Apple would go to great lengths to prevent a competitor app from scraping this data. Presumably, the only use for this data is to advertise within Google apps, encouraging users to migrate from Apple Music to YouTube Music. It's surprising that Google has managed to achieve this, despite Apple's likely efforts to prevent it.

Before jumping to conclusions, I've thought about this extensively. This isn't like those instances where people Google something and then see related ads elsewhere, not understanding how tracking works. I'm aware that phones aren't spying on us in that manner. While this situation might sound similar, I've considered it for months and can't think of any way Google could access this information except through an app installed on the same device.

Perhaps it's easier on macOS, and maybe Google used the Chrome binary to read my iTunes music library, which would be even more surprising. Whatever method they employed to obtain this data is likely interesting, and I imagine there are some Apple developers sending angry emails to Google about not adhering to developer terms.

In conclusion, I find this situation fascinating and worth pondering. How Google managed to access this data remains a mystery, but it certainly raises questions about data privacy and app permissions.

Hey, so the question on everybody's mind today is, did the AI bubble just pop? I think no. The stock market thinks yes. Let's explore why I believe that perspective is a bit short-sighted and doesn't consider the bigger picture.

Here's how I use AI in 2025 for tasks that weren't possible just 24 months ago:

Two years ago, these applications of AI were not possible because the models simply didn't exist. We're only a couple of years into this transformation, and the potential applications of AI are still largely untapped.

Currently, AI can't handle tasks like doing taxes or providing legal advice. The training data exists—case law is just a collection of books that someone spends years understanding. While AI might not take over the legal industry, it could make knowledge of tax law or other legal fields more accessible through well-trained models, but we're not there yet.

The demand for AI hasn't changed. Despite headlines claiming that a model was trained for $6 million to compete with OpenAI's cutting-edge model, I believe that's mostly propaganda. Significant advancements include activating fewer parameters per token and using a mixture of experts. In the coming days, OpenAI is expected to disclose how GPT-4 Turbo was made cheaper and more resource-efficient than previous versions.

Building a data center to run these models still requires hundreds of millions of dollars. It's not like new models will suddenly run on different chips. They will continue to operate on H100s in data centers that require significant electricity and human resources to build and maintain. We are still not meeting the demand for AI infrastructure, as more people want to use tools like Stable Diffusion or ChatGPT than there is capacity for.

The assertion that AI demand has decreased is misleading. There are still new applications for AI models, and the infrastructure to support them is continually evolving. The claim that a model was trained for $6 million on half a billion dollars worth of GPUs is part of a larger story about the ongoing development and deployment of AI technologies.

Hey there, welcome back. I'm James, and this is my indie maker update for Saturday.

What you're hearing is a new audio file format. Optionally, you can add a logo identity to the beginning of your personal podcast updates. This addition makes it sound a little bit more professional, a little bit more like a real podcast. However, right now, it's limited to one very generic, very inoffensive logo identity.

I handpicked and edited the logo to be short and very easy to add to any spoken audio track without requiring many tweaks to the way you record your audio. It works best if you start your audio in a certain way, like "Hey, welcome back" or "Hey, I'm James," instead of the more typical, boring way I normally start my audio recordings.

Hopefully, if this test goes well, this will be the track that is published and introduces the new feature. Thanks for listening.

I've been working on a new app, a Progressive Web App (PWA) for music streaming. Despite its incredible performance and how much I enjoy using it daily, I hesitate to release it. Why? It's due to a counterintuitive principle regarding how users perceive utility apps.

From my experience at HBO, I've learned that certain types of apps are expected to function flawlessly, yet users don't necessarily care about the app itself. These apps often serve as conduits to other content, causing the app to fade into the background. Unlike video games or entertainment media, users don't love the app; they love what it delivers. Consequently, they rarely leave reviews or provide feedback. In essence, you receive zero points for a perfect experience.

The app review landscape is a negative-sum game. On the App Store, I built a tool to aggregate reviews for our app at HBO and analyzed competitor apps. A common pattern emerged: users don't flood the platform with five-star reviews for a seamless experience. Instead, they leave one-star reviews for trivial negative experiences. These are users who have paid for the app for years, use it frequently, and generally have a great experience. Yet, a minor issue can provoke them to leave scathing reviews.

For instance, a user might complain about a subtitle color being slightly off-white, despite the app's overall success. As a developer, it can be baffling to see such minor issues result in negative reviews, especially considering the challenges overcome to make the app work.

When building a utility app or one that serves as a content delivery conduit, this is the reality you face. Users expect perfection because they've paid for it. You get zero points for excelling 29 days of the month; a single negative experience can render your app worthless in their eyes. The phrase "I'd give this app zero stars if I could" is a common refrain in reviews, often for seemingly insignificant problems.

These issues, while minor to most, are crucial to a small subset of users. Even if you assure them that a fix is underway, satisfaction isn't guaranteed. The app review scale feels like it ranges from zero to negative five.

This is why, despite building a music streaming app that I might use for years, I hesitate to release it widely. It may remain a personal project shared with only a few dozen people. I thought I'd share this insight into the challenges of releasing a utility app.

All right, this is probably my last Indie Maker update specifically for Indie.am. This will be my Indie Maker update in general then.

I've basically completed all the features I wanted to complete for Indie.am. Most of this is around GPT and video generation, and those things are working fairly well. I'll continue to use the app and make little tweaks, but I'm pretty happy with where it is. This is more of one of those baseline things that I'm really happy that it exists again. I don't have to go through TestFlight, and I want it just so I can publish my own audio blog and see how it turns out.

Lately, I've been focusing on a different app to stream music and working on some different creative pursuits. I've been writing music, recording, and streaming a couple of games on YouTube, which has been a lot of fun. I'm trying to get in the mindset of creating content in different mediums, playing around with it, and seeing what I can learn from that. I'm having a lot of fun.

I've got like four or five of these Capacitor Framework 7 Vue apps, and each one I build feels a little bit more native and polished. I'm going back and forth between them, trying to re-implement the things I've learned from one app to the other and solving some issues.

For the PWA, I basically have to swap out Framework 7's Vue router for Vue Router to use Vue with Framework 7 in a way that works and feels good. I had to figure out how to get all the animations and swipe back functionality to work. I also usually swap out Framework 7 Store for Pinia and have moved away from any Capacitor APIs unless absolutely necessary. This allows me to build the PWA straight from Vue and only use Capacitor where I need the extra functionality. Building a PWA in Vue with Framework 7 versus a Capacitor build of that PWA results in a completely different set of little works to the app.

I need to go back and update the Mastodon client that I created. I want to get short video working for when TikTok shuts down in a couple of days. I have a TikTok-style video feed inside of that app, but I need to swap out Vue Router for a Framework 7 Router to fix a few bugs and glitches. Once that's done, I can put it back up.

Other than that, I've mostly been focusing on music right now. Maybe I'll share more of that in the next few days, but as far as my Indie Maker update goes, that's pretty much it.

Thanks for listening.

Roughly, here's what I think happened: social media broke society. There's a statistic that's often tracked—it's the number of close friends the average adult has. In 1990, this number was around 13, but it has declined every decade. By 2020, it was down to something like one to three, which is quite alarming.

I believe what happened is that we replaced real-life social interactions with online social media interactions. These parasocial interactions with people online—whom we never actually meet or interact with, like watching YouTube videos—became even more prevalent during and after lockdown.

Elon Musk's influence on social media and the impact of elections further deteriorated the quality of these platforms. Twitter, for instance, used to be tolerable, but now it's filled with bots and garbage. The rise of AI is likely to exacerbate this issue. Not enough people have migrated to other platforms like Blue Sky.

My theory is that AI models like GPT and LLaMA are becoming increasingly affordable. As they get cheaper, any platform based on text content will become completely unreliable. Websites where you scroll through a feed of text posts will turn into a mess, reminiscent of the early 2000s when bots would flood blog comment sections with spam.

I think we're about to experience this bot takeover on any platform that gains traction and relies on text content. It will be more bots than humans. The only way to ensure that a real person is behind a post or piece of content might be to change the format to audio or video. Text and image content, which AI can easily generate, will become so cluttered with garbage that it won't be worth your time.

Platforms that rely on text and image content will suffer. You'll open your feed, and it will be filled with garbage, leading to frustration and eventual abandonment. Platforms like Twitter and Reddit might struggle. Reddit has some human curation through upvotes and downvotes, but if comments become 80% bot-generated, the system will collapse. Text-based platforms like Twitter, Blue Sky, and Mastodon might not survive once AI becomes ubiquitous.

I've just published some fixes to the RSS feed. Years ago, I hardcoded 2021 as the year, which caused issues with all subsequent feeds. I've now corrected this mistake. Additionally, I'm using ChatGPT to format transcripts into blog posts, converting Markdown to HTML. This ensures that the content is nicely formatted when opened in podcast apps.

I've also worked on a custom component to integrate the old Podlog. This feature allows users to click an RSS icon in the app, redirecting them to their chosen podcast app based on their platform. It's a small but significant fix.

Transcriptions are now handled by Whisper, which provides high-quality results. However, it's resource-intensive on the server. I've implemented a system that queues jobs as users browse older Indie.am podcasts, re-transcribing them with Whisper for more accurate timestamps, especially for subtitles.

I've reworked video generation but haven't integrated it into the apps yet. This feature will be complex, involving 15 video templates. Users will be able to:

The video generation process will take about 10 to 20 seconds. If you exit, it will save back to the item and notify you once complete. This complex UI feature is still in development.

I've made several improvements around social media sharing, including meta tags and adding a share button. This is essential for sharing content to other social media feeds. You can embed it in Twitter (formerly known as X), but Blue Sky doesn't yet support audio posts. You can link to them, but playback issues occur due to their in-app browser.

There's a peculiar issue with Blue Sky's in-app browser muting audio by default. If I share an audio post to Blue Sky and launch the podcast app via their browser, I can then return and play the audio file. It's a strange behavior, and I'm investigating why this happens.

I've accomplished a lot and thought it would be fun to share these updates. I'm also testing out the long recording feature. I'll stop here and see how this turns out.

Hey, this is a test of posts that I'm creating using the Android version of the NDAM Progressive Web App. Let's see if it works.

Adobe Firefly is a generative AI model that Adobe has been developing for years. In collaboration with Nvidia, Adobe has been at the forefront of this technology, especially since the decline of crypto mining. Nvidia anticipated a shift in their product usage, predicting that as people moved away from buying GPUs for crypto mining, they would start purchasing them to train AI models. This prediction has proven accurate.

Most large language models and generative AI systems don't have a unique "secret sauce." They rely on vast training datasets of images paired with text or text with classification, supplemented by human feedback. The key components are substantial financial investment, a large team, and extensive training data. This is why, shortly after the release of models like DALL-E, others like Stable Diffusion emerged. These models aren't created in isolation; they are developed by those who can interpret open AI research papers and gather the necessary resources.

Adobe and Nvidia, along with other companies, have been collaborating for years to generate AI services revenue. Nvidia is transitioning to become a services company alongside its hardware business. This shift isn't happening in secret; information is available through press releases and on their websites.

Adobe Firefly isn't designed to compete directly with models like DALL-E or Stable Diffusion by selling API access. Instead, it's part of a broader transition that every tech company will undergo. New spaces and markets are emerging, and companies must decide whether to build or integrate AI technology. This shift is akin to the rise of mobile and cloud technologies, offering a competitive advantage.

Adobe's product managers are keenly aware of the competitive landscape. Most Adobe products cater to creative professionals, and the first applications of their Sensei AI will empower creatives to achieve previously impossible tasks. This technology isn't about replacing creativity but augmenting it. For instance, Adobe Stock's vast image library could be complemented by AI-generated images, offering a new business line.

Consider a future version of Photoshop paired with a large language model, an advanced iteration of something like Stable Diffusion. This could enable in-painting and photo editing on the fly, without requiring users to master complex commands or tutorials. The ability to interact with such a tool through speech or simple typing could revolutionize creative workflows.

Adobe recognizes the importance of being at the forefront of this technological shift. By developing elements of these advanced tools, Adobe aims to preemptively compete against potential rivals. This strategic foresight ensures Adobe remains a leader in the creative software market, ready to meet the demands of the future.