Avsnitt

  • We carry more compute in our hands than sits in the entire cloud, yet almost all the money flows to data centers. For five years we've built on-device first at Detail and Subwave, shipping local rendering, captions, and enhancement while keeping video on the device. Through NP-Hard we backed Mirai, pushing LLM inference to 1,000 tokens per second, and at Detail we use Argmax's on-device speech models. The building blocks exist, but there's still a wide gap between what developers want to build and what's actually available to them.

    Every developer I spoke to at WWDC had the same story: a wish list of AI features they'd build if inference cost weren't a factor, and expensive cloud tokens they'd happily swap for a local model running on the device already in their user's pocket. The demand is real. The SDKs aren't there yet.

    That's why we're launching Desert Ant Labs, a European on-device AI lab focused on shipping dozens of small, opinionated audio and visual models that drop into any product with a few lines of code. No inference cost, nothing leaving the device, running on the 6 billion devices people already own. When cost drops to zero, you stop trading capability against budget on every feature decision. The first models already power Detail and Subwave, and a dozen more will be available to third-party developers before the end of the year.



    Published on Subwave
    https://subwave.app/@paul/post/little-brains-in-every-product

  • Two new on-device AI audio models are shipping across Detail and Subwave this week, and both were built in-house. Until recently, the benchmark for audio enhancement was Dolby: server-side processing, per-minute pricing, and a generic output that sounded identical regardless of the app it came from. Building the alternative became possible because of four things arriving at the same time: open-source base models, years of Detail recordings as training data, affordable hardware fast enough to train overnight, and tools like Claude Code handling the engineering scaffolding around the model itself.

    The Clear model cleans up recordings and the Uhm model detects filler words, both running entirely on-device, processing a 10-minute recording in 10 seconds without the audio ever leaving your phone. The audio enhancement model was shaped around a specific sound target: warm and present, closer to a podcast studio than a phone call. The clean reference audio it learns from is itself lightly enhanced and normalized toward that target, so the model learns the sound of Detail recordings, not clean speech in general. Iteration looked more like product work than research: train overnight, listen back in the morning across tens of real recordings, run blind A/B comparisons against Dolby and against previous versions, pick the winner, repeat.

    Because both models carry zero variable cost, they change what's possible as product defaults. In Detail, Auto Edit runs both models on every recording without being asked. In Subwave, audio enhancement applies to every post by default. The filler word detection model processes a 57-minute interview in under 30 seconds. The result is not a generic enhancement layer dropped into the app; it is a trained opinion about what a Detail recording should sound like.



    Published on Subwave
    https://subwave.app/@paul/post/designing-on-device-ai-models

  • Saknas det avsnitt?

    Klicka här för att uppdatera flödet manuellt.

  • Modern podcasting requires managing content across every platform where audiences actually live. YouTube became the largest podcast platform in 2024, with Spotify hosting half a million video podcasts. Industry research shows 57% of new podcast discoveries now start on social feeds, not the full episode.

    Independent creators face a brutal workflow tax: six different tools costing $200-300 per month, plus 11-19 hours per episode spent on distribution alone. That's half a working week on top of recording and editing. The biggest shows solved this by building teams, but most creators with 500 or 10,000 subscribers end up skipping the very surfaces that would grow their audience.

    Subwave aims to solve this workflow problem by turning one recording into everything: audio episodes, video pages, social clips, article drafts, newsletters, captions, chapters and show notes. Built for independent creators who need the distribution capabilities of a media company without hiring a team.



    Published on Subwave
    https://subwave.app/@paul/post/the-hidden-cost-of-modern-podcasting

  • Building a platform for creators starts with being your own first user. When Detail launched five years ago, the team kept reaching for other tools because the product wasn't ready. The rule became simple: make all marketing content with Detail, or the product isn't ready to ship. Every broken workflow became input for improvement.

    Subwave is being built the same way. The entire team is publishing on the platform while it's still rough, hitting the bugs before anyone else does. Building in public isn't just transparency, it's product strategy. Each post is a test, each new feature we iterate on and gets real use before our customers see it.

    The approach extends to early partners too. Working with a small group of journalists, podcasters, and creators to build their channels and grow audiences together, for free. Not generosity, but the best way to learn how to make people successful on the platform.



    Published on Subwave
    https://subwave.app/@paul/post/build-and-create

  • The geopolitical landscape around digital platforms has shifted dramatically. The Trump administration banned former EU commissioner Thierry Breton from the US for enforcing content moderation rules, while the State Department actively opposes European data sovereignty laws. US cloud providers control 85% of the European market, creating infrastructure dependencies that now carry significant political risk.

    European countries are responding by seeking alternatives. Denmark is piloting open source classroom technology, French public media is evaluating independent hosting, and governments across Europe are actively searching for sovereign digital infrastructure. The discomfort with platform dependency is no longer abstract or fringe.

    Europe largely gave up building consumer social and media platforms after companies like Spotify and SoundCloud emerged over a decade ago. But European platforms built natively within EU digital governance frameworks now have a structural competitive advantage. The opportunity is to build the best independent publishing platform in the world with global ambition but a European foundation, designed for the regulatory environment rather than fighting against it.



    Published on Subwave
    https://subwave.app/@paul/post/time-to-build

  • AI is collapsing the cost of building interfaces to near zero, which means the traditional moat of controlling the user experience is disappearing. When personal agents can curate custom feeds and apps can be assembled on demand, what matters shifts to what's built beneath the surface: content networks, identity layers, moderation standards, and trust relationships.

    Bluesky demonstrates this shift with their AT protocol powering over a thousand apps and 20 billion public records, all drawing from the same content and identity layer. Subwave is applying this same architecture to media publishing and subscriptions, supporting RSS, AT protocol, and open APIs while keeping creator control at the center.

    The key insight: nobody builds a beloved platform by leading with protocols. The product experience must come first, always. But underneath that great experience, open standards make platforms more durable and impossible to replace. As AI agents begin consuming and redistributing media, platforms built on open foundations will be positioned to thrive while walled gardens fight a losing battle.



    Published on Subwave
    https://subwave.app/@paul/post/the-best-product-beneath-the-interface

  • This episode introduces Detail and Subwave, two products built to solve the friction in video creation and distribution. Detail is a video creation app for iPhone, iPad, and Mac that includes features like an adaptive teleprompter, multi-camera recording, and AI auto-editing. Recently selected by Apple for the Best of App Store awards, Detail helps creators go from idea to polished video in minutes.

    Subwave extends that vision by eliminating the chaotic post-production workflow. Instead of managing multiple platforms, formats, and distribution channels, creators simply record and publish. Subwave automatically generates full episodes, vertical clips, written articles, audio versions, and newsletters from a single recording.

    The platform is built on principles of human-centric discovery over algorithmic manipulation, with a transparent business model where creators keep 90-95% of subscription revenue. Additionally, 30% of Subwave ownership is reserved for creators through the Subwave Collective, allowing contributors to earn equity based on their audience and consistency.



    Published on Subwave
    https://subwave.app/@paul/post/from-idea-to-published-video-in-minutes

  • Over the past twenty years, media has evolved from user-generated content in the social graph to algorithmic feeds, and now to long-form subscription models. Each wave has produced new winners, from Facebook and Instagram to TikTok and YouTube. But the platforms that drove growth are now failing creators: Instagram shows posts to only 4% of followers, while TikTok delivers views without building real audiences.

    The pattern is clear across two decades. Formats keep getting richer, from text to photos to short video to long-form content. Distribution becomes more direct, moving from social graphs to algorithms to subscriptions. The relationship between creators and audiences deepens, from anonymous viewers to paying subscribers. This subscription pivot changes everything because it flips the incentives from optimizing for clicks to building trust.

    Substack proved the model works with over 50 writers earning more than a million dollars annually and 5 million paid subscriptions. But no existing platform can fully capitalize on this shift. YouTube can't abandon its $50 billion ad business, Substack started text-first and is catching up to video, and others are constrained by their existing models. The opportunity exists for a platform that combines better creative tools with a viewing experience that works across video, audio, and text.



    Published on Subwave
    https://subwave.app/@paul/post/the-next-wave

  • Video is the defining medium of the decade, but the platforms are built for addiction and ad revenue, not storytelling. TikTok, YouTube, and Instagram dictate format, pace, and who sees your work while keeping your audience for themselves. Meanwhile, trust is eroding. Whistleblowers revealed how financial incentives drive algorithmic decisions at Meta and TikTok, with safety subordinated to growth.

    But people are moving. Blue Sky gained 43 million users in a year. Substack crossed 5 million paid subscriptions. Video podcasts are exploding, with YouTube now the biggest podcast platform on Earth. The pattern is clear: people want direct, intentional relationships with creators.

    Every platform reflects the culture of its builders. Musk turned X into a political megaphone with zero AI guardrails. Zuckerberg chose to flood Instagram with AI content competing against real creators. What happens if you build with different intentions and perspective.



    Published on Subwave
    https://subwave.app/@paul/post/the-broken-promise-of-video

  • This piece explores how AI agents are fundamentally changing the structure of software by interacting directly with components rather than finished products. Instead of adding another layer of abstraction like previous technology waves, agents pull the stack apart and reassemble APIs, databases, and protocols into personalized configurations that serve one person in one context.

    The shift makes well-documented APIs and open protocols the new product surface, while closed systems get routed around. Infrastructure platforms like Vercel and Cloudflare are winning not by building everything, but by making specific surfaces simple enough for agents to use. Standards and infrastructure become the governance layer, defining what's possible and what's allowed.

    The author shares practical examples of running an agent that manages calendar, email, home automation, and more using whatever interface works: CalDAV, IMAP, MQTT, SQLite, web scraping, and CLIs. None of this infrastructure was built for agents, but all of it works. The products that survive will be the ones that are reachable, open, and composable, because openness is how you get assembled into someone's personal software stack.



    Published on Subwave
    https://subwave.app/@paul/post/the-explosion-of-software

  • After five years building video tools, we couldn't shake a growing discomfort: the platforms we create for have become engineered to keep you pulling the lever.

    Then a 1950 lecture by a Canadian economist made everything click. His theory about civilizations and media explains exactly what's broken today, and why we're now living through the most extreme imbalance in human history.

    We've watched creators using Detail craft amazing stories, only to see platforms demand they cut it to 15 seconds or bury it in the algorithm. Time to build.



    Published on Subwave
    https://subwave.app/@paul/post/why-were-building-subwave