For MobileViews 621 I’m joined by longtime friend Jason Dunn filling in as guest co-host. We kicked off by discussing the “moatless” nature of modern AI tools, where hopping between Gemini, Claude, and ChatGPT is easier than ever, though our deep ecosystem preferences often dictate our ultimate choices. This led into a deep dive on note-taking app migrations. Jason shared his recent departure from Evernote after a shocking 171% price hike under its new holding company, Bending Spoons. Using an AI-assisted plugin, he successfully migrated 16 years of data to a local Obsidian vault. I shared my own frustrations with Microsoft OneNote, where basic tasks like syncing handwriting on my Boox e-ink tablet are quietly draining my limited monthly Copilot AI credits, pushing me to explore Obsidian as well.
I noted my initial experience using Google Spark, an autonomous personal AI agent that recently rolled out to Google AI Pro subscribers. This capability was once restricted to the much more expensive Google AI Ultra tier. In my test, I instructed Spark to review the extensive archive of past MobileViews Podcast show notes alongside current technology news platforms. Utilizing the recurring themes from our previous episodes, the agent generates a compiled tech news digest daily at 6:00 PM US Eastern Time, aligning with the conclusion of the standard workday.
A major topic was Jason’s journey into “vibe coding”. While initially hesitant, he used AI to build a custom Tampermonkey script that saved his team over 150,000 manual clicks by streamlining a web form. He later used AI to develop a Mac OS tool called “File Time Travel” to fix frustrating video timestamp metadata issues in Google Photos—a highly specific problem that software developers have not addressed. We discussed how AI flips the traditional learning curve on its head, empowering non-developers to create deeply personal, single-purpose software solutions.
We wrapped up by exploring the power of voice-as-input for AI. Jason has been using Whisper Flow, an LLM-powered speech-to-text tool that finally fulfills the broken promises of old software like Dragon Dictate by actually understanding conversational context and complex names. This allows for “ramble coding”—speaking massive, context-rich prompts directly into the AI to generate much better code and outputs. Of course, the AI revolution isn’t without its hilariously frustrating bumps; we swapped stories about AI hallucinations, from Microsoft Copilot completely fabricating Yellowstone episode descriptions.
Available via Apple iTunes.
MobileViews YouTube Podcasts channel
MobileViews Podcast on Audible.com