This episode covers a wide range of real-world tech experiments, AI tools, and the line between helpful automation and “maybe don’t give that full access yet.”

We start with money talk (not advice). Damashe shares that he finally opened a Fidelity account and bought his first stock, while Michael talks about using watch lists and trade notifications. They also explore how accessible investing apps are getting, including audio charts and VoiceOver support, plus where accessibility still falls short.

Claude Bot and AI with real power A big chunk of the episode focuses on Claude Bot, an open-source tool that lets you interact with an AI through messaging apps like Signal, WhatsApp, and more. The idea of giving an AI access to your computer is exciting… and a little terrifying. This leads to the show’s unofficial safety rule:

> If an AI tool can take actions on your system, run it on a Raspberry Pi or other isolated setup first.

They break down risks like prompt injection, why connecting AI to email and calendars can be dangerous, and why curiosity should always be paired with caution.

AI for everyday life: meet Cookie On the more practical side, Michael shares a cooking app called Cookie. It reads recipes out loud, lets you ask questions like “What’s the next step?” and even suggests ingredient substitutions. It was not originally built for accessibility, but turned out to be incredibly useful for blind cooks. A great example of AI being used in a focused, practical way.

Smarter notes and personal workflows Damashe talks about using AI with DevonThink to automatically organize documents, and why he’s eyeing Drafts with new automation features. The goal: speak a quick note and have it turn into structured data, lists, or tasks without manual sorting.

Social apps, open source, and platform politics There’s also discussion about:

• A new accessible Mastodon and Bluesky client
• Mastodon instances blocking apps built with AI assistance
• The tradeoffs of open platforms where each server sets its own rules

Linux curiosity returns More blind tech users are experimenting with Linux on the desktop again. The hosts are curious what’s improved, especially with screen readers, and ask listeners to share their experiences.

And yes… Todoist check-ins They wrap with progress (and setbacks) on staying consistent with task tracking.

Support Technically Working by contributing to their tip jar: https://tips.pinecast.com/jar/technically-working

Find out more at https://technically-working.pinecast.co

Send us your feedback online: https://pinecast.com/feedback/technically-working/887fda25-0695-49dd-9853-1586cd71fe1c

Check out our podcast host, Pinecast. Start your own podcast for free with no credit card required. If you decide to upgrade, use coupon code r-431b7d for 40% off for 4 months, and support Technically Working.

This week starts nerdy and only gets nerdier in the best way. Michael and Damashe bounce from Apple rumors to AI customer service fails, with plenty of practical tech talk in between.

We dig into the rumors around Apple’s AI direction and what it might really mean for Siri, a possible Home hub device, and Apple’s partnership with Google’s AI models. Are we finally close to a version of Siri that feels truly useful? We share what we’re hopeful about and what still feels like vaporware.

From Perplexity to Amazon Q, we talk about the shift from “AI that tries to know everything” to smaller models trained for specific tasks. Why focused AI might actually be more helpful and less likely to make things up.

Michael has been setting up Amazon SES and got a firsthand look at Amazon’s built-in AI assistant, Q. We talk about how tools like this can make complex platforms like AWS more approachable, especially when you can ask follow-up questions in plain language instead of digging through documentation alone.

Michael walks through why he’s moving WordPress email over to Amazon SES. The big takeaway: sending email at scale can be shockingly inexpensive if you’re willing to do a little setup yourself. We also cover SPF records, sending domains, and a few beginner tips to avoid common mistakes.

From smart locks that won’t unlock to garage lights that randomly stop responding, we share real-world smart home frustrations. We also talk about Matter, Thread, hubs, and why the future of smart homes should mean fewer extra boxes and more reliable automations.

Damashe shares a frustrating experience with an AI phone system that slowed everything down instead of helping. We talk about what good AI customer service should look like and how companies are missing the point when bots just add extra steps.

Apple now has a Creator bundle subscription that includes Final Cut, Logic, and more across Mac and iPad. We break down who it might make sense for and when it’s probably cheaper to just buy what you need.

Support Technically Working by contributing to their tip jar: https://tips.pinecast.com/jar/technically-working

Find out more at https://technically-working.pinecast.co

Send us your feedback online: https://pinecast.com/feedback/technically-working/e1bbbb85-c1d5-48ad-bd90-4aa7d9cf011e

Check out our podcast host, Pinecast. Start your own podcast for free with no credit card required. If you decide to upgrade, use coupon code r-431b7d for 40% off for 4 months, and support Technically Working.

Show notes (Technically Working, Episode 146) This week starts with a dramatic voice demo and turns into a practical conversation about TTS quality, accessibility, and the friction that slows down real work.

In this episode, we talk about:

• More expressive on-device voices (and why “emotion” in TTS can be impressive but unpredictable)
• Why some AI voices drift over long reads (like losing low end after a few thousand characters)
• The practical fix: chunking text around 3,000 characters at sentence or paragraph boundaries
• The jarring side of expressive TTS: when the tone suddenly shifts mid-training

• Mac code editor accessibility and workflow:

• VS Code feeling clunky with VoiceOver navigation

• Nova being close, but still having VoiceOver quirks (like wrapped-line re-reading)
• Missing the flexibility and simplicity of TextMate
• A quick audio reality check: room reverb, mic position, and loud breathing in the mic
• Why it’s worth listening back sometimes, even if you usually don’t
• “Personal intelligence” assistants: Gemini connecting deeper with Gmail, Calendar, Photos, and Drive, and what that could enable
• Stream Deck Plus on sale (knobs!) and the bigger question: is the software accessible enough?

• Capture friction and follow-up problems:

• Getting ideas out of your head fast

• Using automation to sort notes into reminders, drafts, and follow-ups
• Why the Apple Watch action button might help reduce steps
• PLAUD recording devices: improved hardware button design, but app accessibility still matters
• Local processing ideas: Raspberry Pi options for local transcription and LLM workflows
• Listener feedback: Squarespace questions and a quick look at support options (tip jar vs Buy Me a Coffee)

Feedback and contact: feedback@technicallyworking.show

Support the show: Visit technicallyworking.show and click “Support Us” to leave a one-time tip or set up a recurring amount.

Mastodon: @payown@dragonscave.space @damashe@technically.social @tw@technically.social

Support Technically Working by contributing to their tip jar: https://tips.pinecast.com/jar/technically-working

Find out more at https://technically-working.pinecast.co

Send us your feedback online: https://pinecast.com/feedback/technically-working/1396d590-a7dc-4a88-8455-1b3da1991eb2

Check out our podcast host, Pinecast. Start your own podcast for free with no credit card required. If you decide to upgrade, use coupon code r-431b7d for 40% off for 4 months, and support Technically Working.