Clawdbot (Moltbot), Agents and the Age of Personal Software

Clawdbot (Moltbot), Agents and the Age of Personal Software 🔗 **Source:** [Clawdbot (Moltbot), Agents and the Age of Personal Software](https://www.youtube.com/watch?v=KShA22Crumc) > [!note]+ Document Notes > tags: [personal-software, ai-agents, local-development] > author: Syntax > source: YouTube > reference: https://www.youtube.com/watch?v=KShA22Crumc > date created: 2024-12-19 14:30 > > # 🏷️ Title: Clawdbot (Moltbot), Agents and the Age of Personal Software >  ## 🧰 Action Inbox - [H1] Set up Tailscale network for secure local development — enables device-to-device communication #networking - [H2] Install Claudbot for personal assistant automation — connects to Home Assistant, calendar, email #ai-agents - [H3] Experiment with local LLMs using Ollama — reduce dependency on third-party APIs #privacy ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - [Claudbot GitHub](https://github.com/anthropics/claude-3-5-haiku-20241022) - [Tailscale](https://tailscale.com/) ## 📝 Summary - Scott and Wes discuss building hyper-specific personal software using AI agents like Claudbot - They cover local development setups with Tailscale for secure networking between devices - Key focus on privacy-first approaches using local LLMs and keeping data on personal networks - Examples include meal tracking, fitness insights, home automation, and custom productivity tools - Bottom line: AI enables rapid development of niche software that would previously require months of work ## 📖 Main Content Personal software represents a shift toward building hyper-specific applications tuned to individual needs rather than general-purpose solutions. The hosts demonstrate how AI agents can rapidly prototype and deploy these tools while maintaining privacy through local networks. The technical foundation involves Tailscale for secure device communication, enabling developers to run services locally while accessing them from any connected device. This eliminates the need to expose services to the public internet while maintaining convenience. Claudbot serves as a personal assistant that connects to various services (Home Assistant, calendars, GitHub) and can spawn coding agents for specific tasks. The approach prioritizes function over code quality since these applications aren't public-facing. ## 🗒️ Notes #### ⏱️ [00:00](https://youtu.be/KShA22Crumc?t=0) — [H1] Introduction to Personal Software Era - Key idea: AI enables building extremely specific apps in minutes instead of months - Evidence: "The amount of extremely specific apps I have built in the last 6 months, a year is unbelievable" #### ⏱️ [03:45](https://youtu.be/KShA22Crumc?t=225) — [H2] What Makes Personal Software Different - Key idea: Hyperspecific software tuned to individual needs with relaxed code quality standards - Evidence: "I don't give a shit about the code quality if nobody else is using it" #### ⏱️ [05:20](https://youtu.be/KShA22Crumc?t=320) — [H2] Security Considerations for Personal Apps - Key idea: Local-only deployment reduces security concerns about API key exposure - Evidence: "If it's not getting off my network, what the hell do I care if it's exposing API keys" #### ⏱️ [07:15](https://youtu.be/KShA22Crumc?t=435) — [H1] Claudbot Personal Assistant Setup - Key idea: AI assistant that connects to multiple services and spawns coding agents - Evidence: "Connected mine to Home Assistant, calendar, WhatsApp, Telegram, Discord" #### ⏱️ [10:30](https://youtu.be/KShA22Crumc?t=630) — [H2] Home Automation Intelligence - Key idea: AI can suggest automations based on existing device capabilities - Evidence: "Given what you know about all devices on my home assistant, what automations am I missing?" #### ⏱️ [12:45](https://youtu.be/KShA22Crumc?t=765) — [H3] Air Quality Monitoring Discovery - Key idea: AI-suggested automation revealed poor office ventilation - Evidence: "CO2 in your office is four times higher than it should be. You should open a window" #### ⏱️ [15:20](https://youtu.be/KShA22Crumc?t=920) — [H1] Tailscale Network Architecture - Key idea: VPN solution enables secure access to local services from any device - Evidence: "Anything running on my Mac Mini via its ports can be accessible via that IP" #### ⏱️ [18:10](https://youtu.be/KShA22Crumc?t=1090) — [H2] HTTPS Development with Tailscale - Key idea: Automatic SSL certificates for local development servers - Evidence: "Tailscale makes an actual domain with machinename.tailnet.ts and SSL certificate" #### ⏱️ [22:30](https://youtu.be/KShA22Crumc?t=1350) — [H2] Local AI Processing Stack - Key idea: Running text-to-speech and LLMs locally for privacy - Evidence: "Coqui voice quality is super good, could generate an audiobook honestly" #### ⏱️ [25:45](https://youtu.be/KShA22Crumc?t=1545) — [H1] JSON-Based Data Storage - Key idea: Simple file-based storage works well for single-user applications - Evidence: "Number one database for this stuff is just a JSON file - instant saves, no user accounts" #### ⏱️ [28:20](https://youtu.be/KShA22Crumc?t=1700) — [H3] Meal Tracking App Example - Key idea: Photo-based meal cataloging with AI categorization - Evidence: "Taking photos of meals, autocategorizes them, puts them into a bank for food ideas" #### ⏱️ [30:15](https://youtu.be/KShA22Crumc?t=1815) — [H3] Fitness Tracking with OCR - Key idea: Screenshot-based workout analysis with personalized coaching - Evidence: "OCR on workout screenshots, cataloging workouts, giving insights based on goals" #### ⏱️ [35:40](https://youtu.be/KShA22Crumc?t=2140) — [H2] Personalized Meditation Generation - Key idea: AI-generated guided meditations tailored to specific health conditions - Evidence: "Guided meditation talking about my concussion recovery, as good as Headspace but personalized" #### ⏱️ [42:30](https://youtu.be/KShA22Crumc?t=2550) — [H3] Christmas Light Planning Tool - Key idea: Mathematical optimization for LED spacing and material cutting - Evidence: "Exact spacing calculations, cut sheet optimization, 3D printable jigs for drilling" ## 📚 References - [Claudbot](https://github.com/anthropics/claude-3-5-haiku-20241022) - [Tailscale VPN](https://tailscale.com/) - [Ollama Local LLMs](https://ollama.ai/) - [Coqui TTS](https://coqui.ai/) ## 🗣️ Paraphrase - Personal software means building hyper-specific applications for individual use cases using AI agents, prioritizing function over code quality since they're not public-facing - Tailscale creates secure networks between personal devices, enabling local development servers to be accessed from phones, tablets, and computers without exposing services to the internet - Claudbot acts as a personal assistant that connects to various services and can spawn AI agents to build specific applications in minutes rather than months ## 📰 Newsletter **The Rise of Personal AI-Powered Development** Developers are entering a new era where building custom software for personal use becomes as simple as describing what you want. Scott demonstrates how AI agents can create everything from meal tracking to fitness analysis in minutes. The key insight: when building for yourself, you can relax code quality standards and focus on functionality. Combined with tools like Tailscale for secure local networking, developers can rapidly prototype solutions that would previously require months of development. **Key Takeaways:** - Use JSON files for simple data storage in personal apps - Tailscale enables secure access to local development servers - AI agents excel at connecting disparate services and APIs **Try This Week:** Set up a simple Tailscale network and experiment with local AI tools like Ollama for privacy-first development. Resources: [Tailscale](https://tailscale.com/), [Ollama](https://ollama.ai/), [Claudbot](https://github.com/anthropics/claude-3-5-haiku-20241022) ## ✅ Action Items #### DevRel - Developer advocacy: Create tutorials on personal software development patterns, demonstrate Tailscale + AI workflows at conferences - Education: Workshop curriculum on privacy-first AI development, local LLM setup guides - Developer experience: Documentation for rapid prototyping with AI agents, API integration patterns - Developer marketing: Case studies of personal software success stories, developer testimonials - Community management: Foster discussions around personal software ethics and best practices #### OSPO - Policy & Governance: Guidelines for personal software vs enterprise development standards - License Compliance: Review implications of AI-generated code in personal projects - Security Management: Best practices for local AI deployment, network security assessments - Software Inventory: Track local AI tools and dependencies in development environments - Developer Support: Training on secure local development setups, privacy-first tooling - Community Engagement: Support open source personal AI tools and local LLM projects - Strategic Alignment: Measure developer productivity gains from personal software approaches #### Community - Maintainers: Share patterns for JSON-based storage systems, local development best practices - Contributors: Onboard developers interested in AI-assisted development workflows - Sponsors: Fund development of privacy-first AI tools and local LLM infrastructure - Ecosystem: Coordinate between personal software tools and enterprise development platforms ## 🧪 Hypotheses - Personal software development with AI agents reduces time-to-functionality by 90% compared to traditional development - Local LLM deployment will become standard for privacy-conscious developers within 2 years - JSON-file based storage is sufficient for 80% of single-user application data needs - Tailscale-style networking will replace traditional localhost development for multi-device workflows ### 📊 Methodology - Time studies comparing AI-assisted vs traditional development for personal projects - Privacy analysis of local vs cloud AI processing for sensitive personal data - Performance benchmarks of JSON file storage vs traditional databases for personal apps - Developer surveys on networking preferences for local development ### ❓ Questions - How do we balance rapid AI development with maintainable code for personal projects? - What privacy frameworks should govern personal AI assistant access to sensitive data? - How will personal software patterns influence enterprise development practices? ### 🔮 Future Research - Integration patterns between personal AI assistants and enterprise systems - Standardization of local AI deployment for development workflows - Privacy-preserving techniques for multi-user personal software applications ### 💡 Implications - Practice: Shift toward throwaway prototypes and rapid iteration in personal development - Policy: Need for personal data governance frameworks for AI assistants - Security: Local-first development reduces attack surface but requires new security models - Education: Curriculum updates needed for AI-assisted development methodologies - Funding: Investment opportunities in privacy-first AI development tools ### 🧠 Learning Styles Assessment Reading level estimate: 8th grade #### 👀 Visual learners - Present: Architecture diagrams of Tailscale networking, screenshots of applications - Improve: Flow charts of AI agent interactions, visual examples of JSON data structures #### 🔊 Aural learners - Present: Full podcast discussion available in audio format - Improve: Guided audio tutorials for setting up development environments #### 🗣️ Verbal learners - Present: Clear explanations of technical concepts, analogies for networking - Improve: Glossary of AI development terms, expanded examples #### 🧩 Kinesthetic learners - Present: Step-by-step setup instructions, hands-on examples - Improve: Interactive tutorials, sandbox environments for experimentation #### 🧠 Logical learners - Present: Technical architecture explanations, cause-and-effect relationships - Improve: Decision trees for tool selection, performance comparison tables #### 🧑‍🤝‍🧑 Social learners - Present: Community discussions mentioned throughout episode - Improve: Developer forums for sharing personal software projects #### 🧘‍♀️ Solitary learners - Present: Self-contained examples and independent project ideas - Improve: Personal reflection prompts on development practices #### 🌿 Naturalistic learners - Present: Connections to reducing digital waste through targeted solutions - Improve: Environmental impact analysis of local vs cloud AI processing ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Clear heading structure, descriptive timestamps - Improve: Alt text for technical diagrams, ARIA labels for interactive elements #### Keyboard navigation - Present: Standard web navigation available - Improve: Skip links for timestamp sections, focus indicators #### Contrast and readability - Present: Standard text meets contrast requirements - Improve: High contrast mode for code examples #### Text resizing and reflow - Present: Responsive design supports text scaling - Improve: Better line length management at high zoom levels #### Media and captions - Present: Full transcript available - Improve: Synchronized captions with technical term highlighting #### Forms and interactions - Present: Simple navigation elements - Improve: Better error handling for video playback issues ### 🌍 DEIB Assessment #### Cultural breadth - Present: Universal development concepts applicable globally - Improve: Examples from different geographic regions, international privacy laws #### Inclusive language - Present: Technical discussion avoids exclusionary terms - Improve: Define jargon for newcomers to AI development #### Neurodiversity support - Present: Well-structured content with clear sections - Improve: Visual breaks between technical concepts, summary boxes #### Representation - Present: Focus on technical content over personalities - Improve: Diverse examples of personal software use cases #### Psychological safety - Present: Encouraging tone toward experimentation - Improve: Explicit encouragement for developers at all skill levels ### ✅ Logical Validity #### Strong reasoning - Deductive strength: Clear progression from personal needs to technical solutions - Inductive strength: Multiple specific examples support broader personal software trend - Causal reasoning: Direct cause-effect relationships between tools and outcomes - Analogical strength: Effective comparisons between traditional and AI-assisted development #### Fallacies - Appeal to novelty: Some enthusiasm for new tech without sufficient critical analysis - Hasty generalization: Limited examples may not represent all use cases #### Tone of voice - Overall tone: Enthusiastic and exploratory, appropriate for cutting-edge technology discussion ### 🛡️ DARVO Analysis - Deny: No defensive patterns observed - Attack: No attacks on critics or alternative approaches - Reverse victim/offender: Not applicable to technical content - Emotional cues and rhetoric: Positive enthusiasm for technology with appropriate caveats about privacy and security ## 🎯 Conclusion Personal software development with AI agents represents a significant shift in how developers approach problem-solving. By relaxing code quality standards for single-user applications and leveraging local networks for privacy, developers can rapidly build solutions tailored to their specific needs. The key insight is treating these applications as disposable utilities rather than long-term software projects, enabling faster iteration and experimentation. Combined with tools like Tailscale for secure networking and local AI processing, this approach offers a compelling alternative to generic SaaS solutions. For practitioners, start small with simple JSON-based storage and gradually incorporate AI agents as comfort with the tools grows. The focus should be on solving real personal problems rather than building impressive technical demos. ## 🧱 Refactor Appendix ### 📚 Literature Split Pad Purpose: Stage CEE units that become Literature notes. Split level: H3 --- ## Full Transcript [00:01] Welcome to Syntax. Today we're talking [00:03] about building personal software. We're [00:06] going to be talking about how you can [00:07] build software that is just for you, [00:09] runs on your network of things, and [00:12] possibly anyone else you want to give [00:14] this to, and stuff that, well, is kind [00:17] of disposable, but at the same time, [00:19] things you might not want to spend all [00:21] day every day building. Because now you [00:24] can just throw Opus 4.5 on it, knock it [00:27] out in a couple of minutes. Next thing [00:29] you know, you got a feature. My name is [00:31] Scott Tolinsky. I'm a developer from [00:33] Denver. With me as always is Wes Boss. [00:35] What's up, Wes? [00:36] >> Yeah, man. Like, the amount of extremely [00:40] specific apps I have built in the last I [00:42] don't know, 6 months, a year is is [00:45] unbelievable. [00:46] >> It's outrageous. Yeah. [00:47] >> And I I know people like laugh at this [00:48] whole like personal software thing, but [00:50] like it really is true. Like normally [00:53] when I would have have reached for some [00:56] like I don't know crappy SAS that covers [01:00] everybody's use case or I would have [01:02] reached for a spreadsheet uh and and had [01:06] to like do something weird in a [01:07] spreadsheet. [01:08] >> I don't have to do either. [01:09] >> Spending months building stuff. Yeah. [01:11] Just months. I I worked on my habit app [01:13] for a long time and even though there's [01:15] other people using that thing, I built [01:16] it for myself and I'm still maintaining [01:19] it and it's still a big old project [01:21] where most of the stuff I've built, [01:23] let's face it, in the last 4 days, which [01:25] has been a number of things which we'll [01:28] talk about. When I say I built it, I [01:30] mean I project managed the building of [01:33] it. The number of things that have [01:35] gotten done there has been like [01:36] outrageously fast. And there's been a [01:39] lot of like little even like technical [01:40] things that I've learned about what [01:42] makes some of this stuff easy and fast. [01:45] We can get into some of those strategies [01:47] for this personal software as well [01:49] because I do think there's a lot of like [01:50] really neat things that I' I've gained [01:52] by by doing some of this. So, uh, first [01:54] and foremost, what is personal software? [01:56] Why might you care about it? What what [01:58] is it? And it really is hypersp specific [02:01] software that is tuned to your needs. [02:04] Like, I have a need in my life. [02:06] Therefore, I'm going to put an AI agent [02:09] on crushing out a quick project. I might [02:12] not even care about the quality of the [02:14] code. This isn't for public consumption. [02:17] Uh, this could be just for you. And [02:20] there is something about the fact that [02:24] other people aren't using it and it's [02:26] not exposed to the worldwide internet or [02:28] possibly not exposed to the worldwide [02:30] internet. [02:31] >> Hopefully not. [02:32] >> Yes. that makes the security or the slop [02:37] nature of the code more freeing. I was [02:40] definitely in the camp of I need to read [02:42] every line of code that AI produces for [02:44] me forever until I started working on [02:48] personal software that like let's face [02:51] it, if it's not getting off my network, [02:54] what the hell do I care if it's exposing [02:55] API keys or something? It's just on my [02:57] computer, right? So, as long as it's not [03:00] deleting my files or or ruining my file [03:02] system, the slop nature of the code, I [03:04] like I don't care about the slop nature [03:06] of this code in particular. As long as [03:08] it's functioning fine, [03:09] >> especially if you build an app where you [03:11] want to use API keys directly in the [03:14] browser and make the requests directly [03:16] from the browser, that's typically not a [03:18] good idea at all. You shouldn't be [03:19] putting your API keys client side. But [03:22] if you're especially just in prototyping [03:24] stage, you're like, I don't feel like [03:25] building some sort of like proxy to send [03:28] it to the back end and get the response [03:30] and hot potato it back is you can just [03:32] yolo send those directly from the client [03:35] which again let me specify this not a [03:37] good idea in general in general [03:39] >> but if you are just building something [03:41] that is only for you and probably should [03:44] also disable [03:45] >> like Chrome extensions as well. You got [03:47] to think about like what else might be [03:49] reaching in there. [03:50] >> Yeah. What else might be? You're you're [03:52] totally right on on that aspect of [03:53] things. Uh it depends on where you're [03:55] running, how sandboxed it is, and [03:57] whatever. Uh I'm not even talking about [03:58] like API keys, though. I'm just like [04:00] just generally like talking like code [04:02] quality. I don't give a about the [04:03] code quality. If uh nobody else is using [04:06] it. If somebody else is using it, that [04:08] that story changes completely for me. [04:10] I'll tell you that. I care somewhat [04:12] about the code quality because I've I've [04:13] certainly built lots of stuff way too [04:15] fast and it becomes an absolute mess [04:18] where I simply just scrap the entire [04:20] project. But we have an episode coming [04:22] up in the future of like like how do you [04:25] do this stuff so it actually produces [04:27] code that you you would want to be able [04:29] to debug and work on if if it comes to [04:32] the absolute worst case where you have [04:34] to write some code yourself. [04:36] >> Yeah. And I will say that AI is not [04:38] writing all of this code for me. I am I [04:40] have Z open. I'm modifying. I'm writing [04:43] code. I'm chang especially CSS still, [04:45] man. Uh I have some tips there for my [04:47] CSS stuff. But yeah, so let's talk about [04:50] like the idea of like AI and uh personal [04:53] assistant style or personal agents and [04:56] building this kind of stuff because what [04:58] I started was Claudebot. I started a [05:01] Claudebot which is kind of a personal [05:03] assistant uh that is an agent that runs [05:06] on your machine. It just runs on a bun [05:09] process for me. And there's a a cloud [05:12] cloudbot gateway that really handles [05:14] kind of the connections and things like [05:16] that. Like you can connect it to Discord [05:18] and talk to it through a Discord bot, [05:19] which I'm not doing. [05:20] >> Let's explain what this is because this [05:22] has popped in the last couple weeks. [05:24] Claude C L A WD like a [05:29] >> what? Like a lobster. This has this has [05:31] very little to do with like claw code [05:34] and and things. It it does use that, but [05:36] this is a a totally separate project [05:38] that [05:39] >> allows you to create your own personal [05:41] assistant. [05:41] >> Yeah, it it is really it it is a [05:43] personal assistant. And and I personally [05:46] have uh not understood [05:50] why like a a personal agent or assistant [05:53] in AI would be super useful. I couldn't [05:55] think of any use cases for it until I [05:57] started using this. And I do want to [06:01] stress with Claudebot, some of this is, [06:04] you know, if you're sending your data [06:05] anthropic or whatever, you want to be [06:07] careful about what you're connecting to [06:09] this where. But the cool thing about [06:11] Claudebot is they have connections to [06:13] WhatsApp, Telegram, Discord, or [06:15] whatever. So, you don't have to use [06:17] their chat UI, which I don't like at [06:18] all. I ended up building my own even, or [06:20] the AI built one for itself. Or you can [06:23] use WhatsApp, Telegram, or any of that [06:25] stuff to interface with it. So, you [06:26] don't even have to have a chat UI for [06:28] it. You can just do it through Discord [06:30] and then there's all these kind of [06:31] extensions. So, I connected mine to like [06:32] Home Assistant. I connected it to my [06:35] calendar and things like that. I [06:37] connected it I'll talk about some of the [06:39] Home Assistant stuff cuz I was crazy. It [06:41] It's done a lot of cool stuff with my [06:42] Home Assistant. There's a lot of neat [06:43] connections you could make and it's up [06:45] to your own discretion as to what kind [06:47] of sensitive information you want to [06:49] give it cuz there are extensions to [06:51] connect it to your email inbox and then [06:53] it could send emails as you and stuff [06:55] like that. So there is like a degree of [06:58] what you want to uh you know curate with [07:02] this thing. [07:02] >> But the whole idea is that you have this [07:04] bot you you can connect obviously you [07:07] connect to chat to talk back and forth [07:09] with it but then you also connect it to [07:11] like tools right like you said you can [07:13] connect it to GitHub and your email [07:15] iMessage your home assistant you can [07:17] have it control your computer run a [07:19] browser all of these types of things. [07:20] >> Yeah and it has a browser so it can look [07:23] things up. You could have it spawn [07:25] coding processes and things like that. I [07:27] have mine connected to my beads so it [07:30] can project manage things. Yeah. [07:32] >> People are probably thinking like, okay, [07:33] like what what's the use of that? is [07:35] that that now that you just have a [07:36] natural language interface via a text [07:40] message so you can do anything like [07:42] you're you're away from your computer [07:45] and you could ask it to code a project [07:48] or um go through your your iMes [07:52] like filter through your email inbox and [07:54] reply to things for you, right? [07:56] >> Yeah. Yeah. And I I even wrote a custom [07:59] skill for it to connect to tweak which [08:01] is my to-do list. And so every morning I [08:05] it has cron baked in too, west. So every [08:07] morning at 8:00 I get a message from my [08:10] personal assistant with the items in my [08:13] to-do list and then it will like start [08:16] to even work on some of them. So like if [08:18] I put like I need to do something and it [08:21] like I need to research something or I [08:23] can do this, it can start to give me [08:24] links or create documents for that or [08:27] like uh help that process along and it [08:30] can notice if like something stayed on [08:31] my to-do list for several days and be [08:33] like you should really do this, you [08:35] know, just giving me that that [08:36] encouragement which is like kind of [08:37] ridiculous, but the amount of things you [08:39] can connect to it are really great. [08:40] where I found uh one of the the cool use [08:43] cases that I found with it is in Home [08:45] Assistant, which um yes, you might think [08:47] natural language, okay, I can tell it to [08:49] turn off and off my lights, whatever, [08:51] right? But the coolest thing I was able [08:53] to do was say, given what you know about [08:55] all the devices on my home assistant, [08:57] what are some automations that I'm not [08:59] doing that you think I could have? And [09:00] it gave me a whole bunch of ideas. One [09:02] of which was that I have an air quality [09:04] sensor in my office and [09:07] >> I don't have I just have it here and I [09:08] never check the data on it or anything [09:10] and it was like during business hours [09:14] you know 9 to5 Monday through Friday [09:16] I'll check your air quality four times a [09:18] day and if anything is off I'll send you [09:20] a message and the message is in a a home [09:24] assistant prompt like a home assistant [09:26] alert. So that cron job is basically [09:28] just saying check the levels. Is there [09:31] anything off on these levels? Uh if not [09:33] and and yesterday at 12:00 it was like [09:35] your CO2 in your office is four times [09:38] higher than it should be. You should [09:39] open a window. And that led me to [09:41] realize my office isn't venting [09:43] correctly. The mini spit split doesn't [09:45] vent. There's nothing venting air in [09:47] this office. I'm just uh breathing in [09:49] the same air all day. So like that means [09:51] I got to figure out venting or I got to [09:53] just crack a window. I had no idea that [09:55] that was going on. So, uh just that that [09:58] alone was really great. And another [10:00] thing, I was having a a an automation in [10:02] my home assistant that wasn't uh [10:04] triggering correctly or reliably. And I [10:07] said, "Can you investigate this?" And it [10:09] did. And it fixed the dang thing and [10:11] gave me some suggestions. And I was [10:12] like, "Sure, do them." And it modified [10:14] my dang automations on Home Assistant. [10:16] So [10:17] >> man, the awesome [10:18] >> this LLM stuff is going to make the home [10:21] automation stuff way easier because [10:24] >> so much of home automation is people [10:26] just don't do anything with it because [10:28] it it's it's a whole it's a full-time [10:30] job. You having to like make all these [10:32] scenes and make sure they turn on and [10:34] then if something stops working and you [10:36] know it doesn't work and then people [10:39] >> manually turn off the lights. It's [10:41] really frustrating and unless you're [10:43] willing to invest a whole bunch of time [10:45] in this type of thing, it's that's the [10:48] reason why a lot of people don't do it. [10:49] So, I I got Gemini access on my Google [10:53] Home. Man, it is so much better. I love [10:56] it. Like, there's there's one little [10:58] command that I say every night, which is [11:00] turn every light off except for my [11:02] wife's bed lamp. [11:03] >> And I previously would like go and turn [11:06] every room off and then turn hers back [11:08] on cuz she like goes to bed a little [11:10] later. And now I can just like say very [11:12] obtuse things like turn every light off [11:14] except this one and turn the white noise [11:17] on and it just like figures it out. It's [11:18] a little slower. You can tell that it's [11:20] like sending it to an LLM to process, [11:23] but I don't necessarily care because [11:25] it's it's so much better. And I've been [11:27] so frustrated with my Google Home over [11:29] the last year or so and it's finally [11:31] getting better. [11:32] >> Yeah, that that actually that rips. I [11:35] had to write an automation that if it [11:36] detects presence from when we're not [11:38] home to when we're home because home [11:40] assistant can do that. If it detects [11:42] presence in the home, like if you're [11:44] arriving home, it's going to turn on [11:46] lights automatically before you get into [11:48] the house. It's just like, oh, [11:50] >> okay. I wouldn't have thought of that. [11:52] You did it. Okay. It's not motion [11:53] sensoring. Whatever. It's interesting. [11:55] So, Claudebot has been really cool. It [11:58] does this gateway. You can connect it to [12:00] various LLMs. Uh the idea for me here [12:03] eventually, right now I'm I'm having him [12:04] connected to Claude uh Opus. Claude is [12:07] an anthropic Claude, but the idea here [12:09] will eventually be for me to get a beefy [12:11] computer like a Mac Studio. Um CJ [12:15] actually just got a little uh integrated [12:17] uh GPU PC. that would wouldn't be a bad [12:20] idea to get either, but I think I want [12:22] to get a a really beefy Mac Studio and [12:24] then have that run a much better local [12:27] LLM than what I can run on my MacBook [12:29] Pro and then do everything through that [12:31] instead of going off to anthropic [12:33] because the moment that you do that and [12:35] the data is not leaving your computer, [12:37] then you can give it more sensitive [12:40] information, maybe even like financial [12:42] information and it can do, you know, [12:44] help you with your money tracking or [12:45] stuff like that or or you know, even [12:47] email, you know, maybe give it read [12:49] access. [12:50] >> There's one thing I have to do every [12:52] four times a year and that's like go [12:53] into every bank account that I own and [12:55] export it all and everybody tells me to [12:58] use plaid or whatever, but it's like I'm [13:00] not paying for that and [13:01] >> yeah, some of my bank accounts doesn't [13:03] work and like the second I can automate [13:06] that entire thing easily. I probably [13:08] could automate it right now with a whole [13:09] bunch of puppeteer and clicking buttons, [13:11] but it feels too [13:13] >> dangerous and like like fragile like it [13:16] could break when they move a button or [13:18] whatever. But as soon as I can hook it [13:19] up to something like this, especially [13:21] with a local LLM, [13:23] >> I I I think we're probably a bit of a [13:25] ways away before we can get like decent [13:28] quality LLMs running locally without [13:31] having to spend like what, four grand on [13:33] some hardware. [13:34] >> Oh, I'd probably be spending more than [13:36] four grand. I'd probably be getting into [13:37] six territory if you're looking at the [13:39] Mac Studio. Uh but like then again that [13:42] becomes there's part of this personal [13:44] software thing that has kind of flipped [13:45] in my head about how I'm running my dev [13:48] process in general. So I'll talk about [13:49] that right now which is that I've moved [13:52] everything to a tail scale network [13:54] throughout my house which has become [13:56] really really awesome. So I have my [13:58] phone, I have my uh Mac mini, I have my [14:02] computer, my tablet all connected to [14:04] scale and tail scale is basically allows [14:07] you to have like a personal tunnel [14:09] throughout your entire network. And so [14:11] what it is is it becomes the VPN on your [14:13] computer, your phone, whatever. And when [14:16] I want to access things on my Mac Mini, [14:19] it is via a tail scale IP address. And [14:24] that means that like anything that's [14:26] running on my Mac Mini via its ports can [14:29] just be accessible via that IP uh [14:32] without having to turn on uh any sort of [14:34] like global file sharing or stuff like [14:36] that. It's just being done through tail [14:37] scale. The way I would have like done [14:40] some of this before, like if I had a [14:42] site running on something, I might throw [14:44] it on a Cloudflare tunnel and then [14:46] you're putting that out to the wider [14:47] internet and you could protect it with [14:49] like a zero trust on that thing and have [14:51] it be like a a password that sends you [14:53] an email. But what's so cool about [14:55] Tailscale is that like it's so seamless [14:58] on my phone, my computer or whatever. [15:00] And then all of these things just I I [15:02] can access my dev sites really easily. [15:04] So I have like a vit dev server for a UI [15:08] running on my Mac Mini and then I tell [15:11] tail scale to serve it as HTTPS because [15:14] if I want HTTPS and all that stuff uh I [15:17] want a SSL certificate and then tail [15:20] scale makes an actual domain for it with [15:23] machineame.tailnet.ts [15:24] TS, whatever, and then a port on [15:26] whatever port you've assigned it to. And [15:28] now I have a secure location on a dev [15:31] server on my own network that I can [15:33] access from any device. I can test it or [15:35] use it on my phone, my computer, or [15:37] whatever. And it took 2 seconds. And it [15:39] costs free.99. I didn't pay a scent to [15:42] be able to do that. And then on your [15:44] phone, what's the process of setting [15:46] that up? Do you put like a VPN on your [15:48] phone then? [15:49] >> Yeah, you really just install the Tail [15:51] Scale app and you turn it on. You log [15:54] in, which I log in via GitHub. You turn [15:56] it on, but it basically just shows all [15:58] the devices that are connected through [16:00] your T. They don't have to be on the [16:02] same Wi-Fi network, which is awesome. If [16:04] I'm out of my Wi-Fi network, I'm on uh, [16:07] you know, 5G, I'm uh, I'm I'm out and [16:10] about, I still get access to all of my [16:12] devices easily through the TailscaleVPN. [16:16] Mhm. [16:16] >> And then I was typically running like a [16:18] MolvadVPN [16:20] uh on my stuff before, but there's [16:21] actually a Mulvad add-on to this. So all [16:25] outward traffic to the wider internet [16:27] goes through a Mulvad VPN through the [16:30] tail scale app itself. [16:32] >> What is Molvad? That's [16:34] >> it's just a normal VPN. It's just like a [16:36] you know NordVPN, like one of those [16:38] types of providers. Yeah. [16:40] >> Okay. We talked about this on our like [16:42] Synology server show, but I'll I'll [16:44] quickly say how I'm doing this on my [16:45] myself is I'm running the Cloudflare [16:47] tunnels on my Synology. And then what I [16:50] can do is I can selectively opt into [16:53] what like you are doing with tail [16:55] scales. You can opt into which machine [16:57] on the network and which port you would [16:58] like to then exposed to the internet. [17:01] But then you have the problem of like [17:03] you've just exposed your local computer [17:05] or local server to the global internet, [17:07] right? So then you have to then put some [17:09] O in front of that. And in my case, I [17:11] just use GitHub. I sign in with GitHub [17:15] to be able to access any of my stuff and [17:18] it works super well. Um, it's also free [17:20] as well. But if there's something on my [17:24] local network, like for example, if I'm [17:26] away from the house and I don't feel [17:27] like setting up like a like a forwarding [17:29] or or port or something like that, I can [17:31] also use UniFi Teleport. So, this is [17:34] Teleport runs on on my router, my UniFi [17:38] device. And what that does is you just [17:40] on your phone, you flip on the app or on [17:41] your desktop, you flip on the app, and [17:43] it makes a VPN right back to your home [17:45] network. And then you can access [17:47] anything on your local home network, [17:49] which is good because I have a like a 3D [17:51] printer I was running some custom [17:52] firmware on and I needed to be able to [17:55] look at the camera and I forgot to [17:57] expose it to the global internet so I [17:58] could look at it when I was out away [18:00] from home. Uh, and I just jumped on [18:02] UniFi teleport and it it did that. [18:05] >> Yeah. Well, that's cool. Uh, and so you [18:07] have to have a UniFi router to do that. [18:09] >> Yeah. Yeah, you got to have Unifi. So, [18:10] it's probably not most people um it's [18:13] probably better to use something like [18:15] like Tailscale or a Cloudflare Tunnels [18:18] um to do. [18:18] >> Yeah. I I kind of avoided the Tailscale [18:20] thing because I don't know why. I was [18:22] like, "Oh, I want to expose it to the [18:24] internet." And then through this stuff, [18:25] I'm like, "No, wait. I don't I don't [18:27] really want to do that. I want to have [18:28] it all just on my devices." And as long [18:30] as I'm running tail scale on those, it's [18:32] all good. [18:32] >> Yeah. You want your your devices to go [18:34] back home, not have your home devices [18:37] exposed to the global internet. Right. [18:39] >> Exactly. Yeah. Exactly. Especially for [18:41] this stuff. Now, there's some things [18:43] >> that that's probably different like uh [18:46] my audio bookshelf or uh my Plex like if [18:50] I want to give like my brother control [18:52] or access to those libraries as well. Um [18:55] then, you know, I I don't want him to [18:58] connect to my tail scale network. So [18:59] those make more sense for Cloudflare [19:01] tunnels or to to expose that port via [19:04] that, but I'm not going to I'm not going [19:05] to open the ports. Okay. So let's talk [19:07] about like privacy stuff because you do [19:09] need to make sure that whatever you're [19:11] sending to the LLMs is still something [19:13] you want to be sending to a third party. [19:16] Um again, I I've experimented a bit here [19:18] with local LLMs. I have Olama running. [19:22] Uh I haven't found anything that I love [19:24] yet. Uh I think my computer needs to be [19:26] a bit beefier. Now, the thing about [19:28] Claudebot is is that the assistant part [19:30] of things doesn't need to be that good. [19:35] It needs to be good, but not like not [19:38] Opus 4.5 good like cuz you could have [19:41] that assistant spawning Opus agents to [19:44] work on code stuff or you could just [19:46] work on it via your your normal cloud [19:48] code or whatever. So, the actual [19:49] personal assistant really doesn't need [19:51] to be that good. It just needs to have a [19:53] decent amount of context. So, when I [19:55] have a better computer, I will run, I [19:58] think, all of the personal assistant [19:59] stuff on Olama and be totally fine with [20:01] that. Another thing I did is I I've been [20:03] getting into other local processes for [20:05] like processing like text to speech. I I [20:08] use one called uh Coco. Cocooro Cocoor. [20:13] >> Yes, I use Coco. And the way I have it [20:16] set up is through something called MLX [20:17] audio. It's a Mac specific thing for Mac [20:20] silicon. uh basically just runs as a [20:23] server on my computer. Then my Mac Mini [20:26] hits that endpoint through the tail [20:28] scale address, generates the audio as an [20:31] MP3, sends it back to the Mac Mini where [20:33] it's stored and then served via a [20:36] spelkit app. So what I was able to do [20:38] with my chat assistant, my chat UI, was [20:41] build a really quick texttospech speaker [20:43] button. I click the button and it reads [20:45] any message. That also enabled some [20:47] other neat ideas which I'll talk about a [20:48] little bit. But you can run text to [20:50] speech entirely on your local machine. [20:52] And what I found is especially with uh [20:54] this Cocoaro, the voice quality is super [20:58] good. Um I started with Piper and it was [21:02] like very like 1990s kind of computer [21:04] voice. But with Coco, man, the the [21:08] quality is is way good enough that I [21:10] could have this read me an audio. I [21:12] could have it generate an audio book, [21:13] honestly. And it would And you can you [21:16] can run it in the browser with [21:17] Transformers.js as well. I've used it in [21:19] the browser before. You have to use a [21:21] slightly smaller model and it's it's not [21:23] as good, but it's still [21:25] >> extremely impressive of how good it is. [21:28] You know, we've gone from like we had [21:30] like Bonsy Buddy in like 95 and then we [21:33] went to like 2021 with very similar [21:36] quality and then it's just in the last [21:38] couple years it's gotten so much better. [21:40] And if you want to see all of the errors [21:42] in your application, you'll want to [21:43] check out Sentry at centry.io/sintax. [21:48] You don't want a production application [21:50] out there that well, you have no [21:51] visibility into in case something is [21:54] blowing up and you might not even know [21:55] it. So head on over to centry.io/s [21:58] syntax. Again, we've been using this [22:00] tool for a long time and it totally [22:02] rules. All right. [22:04] Yeah. So, for me, uh, I'd like to move [22:07] as much of this stuff locally as [22:08] possible. And Kokuro runs totally fine [22:11] next to anything, uh, running on my my [22:14] laptop. And since it's all on the same [22:15] network, it's just communicating back [22:17] and forth via the tail scale, uh, URL. [22:20] So, really nice there. Let's talk about [22:22] app ideas. I' I've kind of dropped some [22:24] of the app ideas or stuff that I've been [22:26] working on or or talking about through [22:28] here, but like some of the app ideas. [22:30] So, we talked a lot about giving this [22:31] like personal assistant stuff, like what [22:33] can a personal assistant do, but what [22:35] about like I have my personal assistant [22:37] sending off agents to build personal [22:39] software? And I did that through a like [22:42] a a mothership app, which um is [22:45] basically I built a chat UI that I can [22:48] chat, I can spawn new agents, I can uh [22:50] work with my beads and projects, and [22:52] then I can add on to that with [22:53] individual personal software things. So, [22:56] one thing that Courtney and I have been [22:57] talking about having forever is that [22:59] anytime she asks me, "What do I want for [23:00] dinner this week?" I say, "I have no [23:02] clue. I'm I can't My brain can't think [23:04] of a single thing you've ever made in [23:06] your entire life." Uh, and so to augment [23:09] that, I'm going to be taking photos of [23:12] the meals she makes when she makes them, [23:14] and it autocatategorizes them. It autoc [23:16] cataloges them, and it puts them into a [23:19] bank. So I could say, "Give me one of [23:22] each style of food or or give me food [23:25] ideas." And it can give me food ideas [23:27] through that bank. So I have meal [23:28] tracking in there now. And it's just [23:30] building this database. Now, how that's [23:31] working is it's just taking the photo [23:34] from my camera. And since it's all [23:35] running locally, it's just throwing in [23:37] my file system of the app. And then for [23:40] database stuff, let me tell you, the [23:43] number one database for this stuff is [23:46] just a stinking JSON file. Man, I I I [23:50] get instant saves. I get all this stuff. [23:53] I don't have to log in. I don't user [23:55] accounts or user access. It's just one [23:57] person. But it's just writing it to a [24:00] JSON file every single time you even or [24:02] type a character. It is awesome that it [24:05] just does that. And then now like it [24:08] it's trivial for the LLM to read through [24:10] that JSON file. Or what you can even do [24:13] is I built in slash commands to my uh [24:16] chat UI that can actually just run [24:18] script. So it could pull uh certain ones [24:22] with JavaScript instead of having to do [24:24] an LLM call for that. Yeah, that we [24:27] talked about that a couple episodes ago [24:28] where I was like, I just want a brain [24:31] that I can throw things into. And I [24:34] surprisingly I didn't get any any [24:36] suggestions from the audience of like I [24:38] just want something where if I have [24:41] something that I'm remembering, you [24:42] know, like the the lug nuts on the [24:45] tractor at the cottage are this weird [24:47] 5/16 fine thread and I just want [24:50] somewhere to put that or I measuring [24:52] windows in our house or measuring a gap [24:54] where we need to find like a a dresser [24:57] or something in our in our house like [24:58] like all of these little things where I [25:02] just need to remember. I always think [25:03] like where should I put that? Do I put [25:04] that in an Apple note? A lot of people [25:06] email it to themselves, you know, and [25:09] like I just want to be able to very [25:12] similar to how you use like a reminder, [25:14] just be able to to speak that into some [25:17] sort of brain and have it come back. I [25:20] found this thing called Sublime, which [25:22] is unfortunately not Sublime Text. [25:24] Sublime.app. I haven't tried it yet, but [25:26] it seems to be [25:29] that, but also a great way to like [25:31] manage inspiration, bookmarks, you know, [25:35] like almost like a like a Pinterest [25:37] where you would dump things into. [25:40] >> Dude, my my chat UI could build that [25:42] into the chat UI in an afternoon. That's [25:44] what's so crazy about this. [25:46] >> And then you don't have to do some [25:48] thirdparty service. You can have it do [25:49] exactly what you want. So, the meal [25:51] tracking thing was great. Another thing [25:52] I did was fitness tracking and insights. [25:55] Uh right now I'm just uploading a [25:57] screenshot of my tonal workouts. It's [25:59] doing OCR on that text and it's uh [26:02] cataloging that workout. And then if I [26:05] give it my goals, it can come back and [26:08] say, "Hey, I noticed you're doing a lot [26:10] of traps and you your your traps are [26:12] fine." [26:13] >> Yeah. [26:13] >> So like like cut the trap workout and do [26:16] something else here. Like just to be [26:17] able to get that insight over time. like [26:19] Apple Health as well, you know, you [26:21] could hook it into Apple Health and push [26:23] and pull data. I think a lot about [26:25] >> that with software that I pay for yearly [26:29] or monthly [26:31] >> that I don't use all of the features. [26:33] So, um, macro factor [26:36] >> probably one of the best apps in the [26:38] world for tracking calories and all of [26:42] that like health and fitness. you know, [26:43] if you've used like Daily Burn in the in [26:45] the past, um, MacroFactor, Scott [26:47] recommended it what, probably two or [26:49] three years ago, and I've been paying [26:50] like like 200 bucks a year for that [26:53] thing. And I mostly just use it for [26:56] tracking what I'm eating. And there's [26:58] some nice features, right? You can scan [26:59] a barcode, and you can change how like [27:02] if you're having one and a half servings [27:03] of something. And they also have this [27:05] new AI input where you can take a photo [27:07] and describe what it is and it will [27:08] break it into. But I I just keep looking [27:11] at that being like, [27:13] >> "This can't be that hard." You know, [27:17] couple graphs of where you're at. [27:19] >> There's got to be APIs for all of these [27:22] foods um available somewhere, you know? [27:24] I'm I'm sure that exists. [27:26] >> So, like 200 bucks a year, that's quite [27:28] a bit. And same with like um I use [27:30] Strong Lifts um which is like like a [27:32] like a workout tracking app. And I look [27:36] at that every time. It's like 60 bucks a [27:38] year, which is not too bad. Like I don't [27:40] mind paying. I like that the developer [27:41] is kind of independent and whatnot, but [27:43] I often think like [27:45] >> I'm just pressing a button every second. [27:49] >> Yeah. It's tracking how much weight [27:51] there is, you know, and uh there's an [27:54] Apple Watch integration which is an [27:55] absolute must. So I'd have to build [27:57] something like that. I don't know if 60 [27:58] bucks is worth [28:00] >> uh not use moving away from that. But [28:02] there's a lot of little things where I'm [28:04] just getting nickeled and dime every [28:05] single year. Every single I'm paying for [28:08] it. I'm thinking, I wonder if I could [28:09] just build this myself. [28:11] >> I know. I know. I was thinking about [28:13] that and then some of the stuff I just [28:14] like I just threw a sub age on and say, [28:16] "Hey, can you throw this together?" And [28:18] then like 15 minutes later it's like [28:19] done. I'm like, "What?" [28:21] >> Mhm. [28:21] >> What? [28:22] >> To be clear, sometimes the answer is [28:24] absolutely not. And it just will kick [28:26] out a whole bunch of garbage and you you [28:28] wasted a bunch of your time. We're not [28:29] we're not being those guys that are just [28:31] like you can just think anything into [28:34] existence, you know, like [28:36] >> sometimes it just duplicates all of my [28:38] CSS, you know, like we got bigger [28:40] problems as well, but I I think we're on [28:42] a good path right now. [28:44] >> We'll we'll talk about this in the using [28:45] AI effectively. But I have good bones, I [28:48] have a good system, I have good files, [28:51] uh good agents to work on this stuff, [28:53] and I use my CSS framework graffiti, [28:56] which has good AI documentation. And so [28:58] like the agent says do not add a new [29:01] class un unless like this UI element [29:04] does not exist already in graffiti if [29:06] like do not like all those things and it [29:08] actually has been really adhering well [29:10] to that. So uh that that's been like [29:12] positive in terms of keeping the [29:14] structure. That said, I'm sure some of [29:15] these files are monstrous. Another thing [29:16] I added in here is like journaling where [29:19] like it will custom prompt me every day [29:21] at like 1:00 and like ask me a question [29:24] where I can like respond to it. It will [29:27] be able to pull out and tag things and [29:29] like mark things as wisdom. It has like [29:31] 200 prompts that I haven't yet to see [29:33] because I haven't I just told this to do [29:35] this last night. And then it's going to [29:37] do it via audio. So I'm can just push [29:40] record in the app and speak and then [29:42] it's going to do that speech to text and [29:44] then autocategorization cataloging and [29:46] then putting them in as markdown files [29:48] into a folder. This is a good one uh [29:50] because it has a little bit of a [29:52] background about like my concussion and [29:54] my recovery. I had it build a personalg [29:58] guided meditation and using the text to [30:01] speech. The voice is so good it actually [30:04] puts in ambient noise beyond it. So, [30:07] it's using ffmpeg to layer in ambient [30:09] noise with the text to speech and it is [30:12] as good as headsp space or anything, but [30:14] it's saying like Scott like recovery is [30:18] a journey. Like like it's talking to me [30:21] exactly about my issue and my like [30:24] health thing with my concussion and it's [30:27] talking to me directly. It's it's like [30:30] man it I've never wanted to do [30:31] meditation and I'm so like now I have [30:34] guided meditations that are like really [30:35] tuned to my exact situation [30:39] and man that was really really like kind [30:42] of an eye openening thing for like wow [30:44] we can do this that's crazy. [30:45] >> Yeah it's I'm very impressed with all of [30:48] the stuff that you can do. I built [30:50] something where my kids have this like [30:52] yoto and I needed to like upload a whole [30:55] bunch of MP3s that I had um and then set [30:58] icons for each of them and set some [30:59] album art for for them and like this [31:01] like they have a really nice UI for [31:04] doing it. But it's just like slow um [31:07] being able to do it. You know, I wanted [31:08] to upload like 20 of them at once. So I [31:10] just boop boop just figured out the [31:13] reverse engineered the API and built a [31:15] little script to be able to find them in [31:18] the thing. AI is so good at doing [31:21] fuzziness um where things might not be [31:24] named exactly as you would expect it and [31:27] it's so good at just being able to like [31:29] look at the directories. Oh, here's the [31:32] one that you're looking for and here's [31:34] what the names of the actual files where [31:36] this one had a dash in it. This one [31:38] didn't have a space in it. You know, [31:39] like so much of my like script writing [31:42] over the past 10 years has been so built [31:44] on rigidness and then maybe writing some [31:47] like reaxes to make it a little bit more [31:49] flexible and the like fuzziness of data [31:52] and data being [31:54] >> a little bit weird has been so much [31:56] better with all these AI agents. There's [31:59] something really powerful about software [32:02] that like you wouldn't have the patience [32:04] or time to write, but like now that you [32:08] can throw an agent on it and have it do [32:10] it in no time, it unlocks a bunch of [32:13] access to like like little helpful [32:15] things that you just straight up [32:16] wouldn't have. I for me, I've just been [32:18] thinking about like what can I augment [32:20] about the the friction in my life here? [32:23] um whether that's reminding me about [32:25] to-dos or or again like helping my [32:28] concussion recovery, etc. So, that's [32:30] where my brain's been with a lot of this [32:32] stuff lately. [32:32] >> I was doing my Christmas lights uh this [32:35] year and I went all out on building um [32:37] window frames. So, all of the windows in [32:39] our house, both of the garage doors and [32:42] the entrance to our house, I built out [32:44] of PVC pipe. I built these um window [32:47] frames that have every about 3 in [32:51] >> um I drilled holes in them and you poked [32:53] these little LED lights in and then our [32:55] house was just like lit up from the sky. [32:57] It was it was amazing. And in order in [33:01] order to do this, I needed to make sure [33:04] that the spacing of the LEDs was exactly [33:09] the same. Um because I wanted I you [33:12] drive by people's houses and you can [33:13] tell that they they don't exactly have [33:15] the LEDs in the corners of the windows [33:18] and what I wanted was just like CSS [33:21] grid. I wanted exact even spacing all [33:23] the way across. And if you just most [33:25] people in the industry just say ah 3 in [33:27] is about good. But the thing is is that [33:29] like over a run of I don't know 20 30 [33:33] lights if you're off even one or two [33:36] millimeters or half a millimeter over 20 [33:39] lights it it really adds up and then you [33:41] have a you have an LED that's too close [33:43] to the edge or not far or too far away [33:46] from the edge and it just looks weird. [33:48] So what I wanted is is to plan exactly [33:51] not 3 in but however many millimeters um [33:54] it was going to be. So, ideally 3 in. [33:57] And then what it did is it figured out [33:59] if it needed to bump that value up or [34:01] down a little bit in order to get even [34:03] spacing. So, I figured that out. And [34:05] then it also provided me with a cut [34:08] sheet, meaning that I could cut the PVC [34:11] pipe exactly. So, it told me exactly how [34:14] long each piece of PVC pipe should go. I [34:17] told it I I bought it in 10ft lengths. [34:19] So, what it did is it figured out the [34:21] minimum waste, meaning that like, okay, [34:23] well, like you take the long piece from [34:25] this window and the short piece from [34:27] this window, and those add up to like [34:29] 98% of the actual thing. So, you're not [34:31] just cutting them and ending up with a [34:33] bunch of pieces that you can't use. It [34:35] thought about it all all the way ahead [34:36] of ahead of it. Um, and then what else [34:39] did it do? One of ours had an arch in [34:41] there. So, I I had it figure out I gave [34:44] it all the measurements of um how tall [34:47] the middle of the arch was and how how [34:49] tall the sides of the arch were. And it [34:50] figured out how much of a degree I [34:53] needed to angle the connectors, which [34:56] was really good. And then what else did [34:58] I do with it? Oh, I had it I had it [35:00] output a 3D printable jig [35:02] >> so that I could just put the PVC in, [35:05] drill a hole, move it forward, it would [35:07] click into place, drill a hole that was [35:10] not perfect, so I ended up modeling [35:12] something myself in Fusion, but it did a [35:14] pretty good job at at making a little [35:16] jig that I used for the first couple [35:18] ones. [35:18] >> Can I say I accidentally killed two [35:20] hotends on my 3D printer in week or so? [35:24] >> Doing what, dude? I don't there's I'm [35:26] trying to do the gridfinity base where [35:28] it does a PLA and then PET G and then P [35:31] whatever. It stacks them. [35:32] >> Oh yeah. Stacking. Yeah. [35:34] >> And in that process of switching the two [35:36] materials. I've gotten two clogs so bad [35:39] that a hot push doesn't work. Extraction [35:42] doesn't work. I got to take apart the [35:44] whole thing. I had to take apart the [35:45] gears cuz the gears were stripping the [35:47] filament so I couldn't even pull it out. [35:48] and two hotends. And I know, granted, [35:50] they're like 35 bucks or something, but [35:52] like two hotends dead in a in in like [35:55] three prints, maybe. I I was so [35:57] frustrated when the second one died. I [35:58] was like, "This is [36:00] >> Oh, this is going to kill me." And and [36:01] I'm trying to do the hot push where you [36:03] get the the thing, you hold it, and [36:04] you're pushing and it's like billion [36:06] degrees and you don't want to do a hot [36:08] pull. [36:09] >> There's not enough in it's like right. [36:11] It's flush to the top of the thing. You [36:12] couldn't do a hot pull. [36:14] >> I could I did a hot pull. I have the [36:15] same one as you. I had um [36:17] >> No, I'm saying like the uh the amount of [36:19] filament coming out of the top of the [36:20] hot end is flush with the top of the hot [36:23] end. So, there's nowhere to even reach [36:24] in to pull. [36:26] >> So, [36:26] >> Oh, yeah. And since you can't push it [36:28] out, you can't [36:29] >> Oh, I see. Yeah. [36:31] >> Yeah. I had to take the whole thing [36:32] apart. I had like a thousand screws and [36:34] everything laid out. I I like got really [36:36] familiar with the [36:38] >> I once melted the end of some of the [36:40] filament and stuck it on a little piece [36:42] and let it cool. And then you heat up [36:45] then you heat up the nozzle and then you [36:48] then you do a pull. I had a little fleck [36:50] of as a marble filament. I had a fleck [36:52] of the marble stuck in the thing and [36:55] that [36:56] >> that killed it as well. [36:57] >> Yeah, I could still try that. I don't [36:59] know if it's just the filament that I [37:00] bought. It was the same filament that [37:02] killed it both times. So I'm like throw [37:04] that filament. [37:04] >> Probably a bad Yeah, bad batch. Anyways, [37:07] we should do an updated 3D printing [37:08] episode because we're coming up on a [37:10] year now. [37:11] >> What? U or maybe not the year. I'm [37:14] coming up on a year on mine and all the [37:16] comments on our 3D printing episode were [37:19] >> see it in talk to us in a year. It's [37:21] going to be collecting dust. [37:23] >> I could use a second one. [37:24] >> Honestly, I have a second one and I'm [37:26] thinking about like hm I could use a [37:27] bigger one. [37:29] >> Yeah. Right. Whatever. Yeah. Whatever [37:30] the next bamboo one they come out of [37:32] that's bigger. Yeah. I know. I just like [37:33] built my whole workshop over the [37:34] whatever. And [37:35] >> looks good. Yeah. [37:37] >> Thank you, [37:37] >> man. Yeah. That that's Anyways, that's [37:39] another show. What else? Dealinders is [37:41] something I've written several times [37:43] over. I've used this to find like like [37:46] filament is is a really good example [37:48] where like sometimes filament will go [37:50] extremely cheap, especially on like [37:52] Amazon returns. You know, people get [37:55] something on Amazon, it doesn't work, [37:56] they return it and then they put it [37:58] sometimes it's like marked down like [38:00] half off. Um, so I bought a little bot [38:02] that would just scrape the Amazon API [38:04] and then alert me if something pops up, [38:07] things like that. I've done that many [38:08] times over the years and it is so much [38:11] better when you can use like a proxy man [38:14] and just give AI the actual output be [38:16] like this is what a request looks like [38:18] you know here's the curl command here's [38:20] the actual return values and then it can [38:23] write types for all the return data and [38:24] then go ahead and re-replicate all of [38:27] those requests. [38:28] >> Yeah. Yeah. It's a wild world. [38:30] >> Yes. I'm I'm going to install this [38:32] cloudbot. I've been been holding off cuz [38:35] everybody's talking about it and [38:36] everyone's all excited about it and [38:38] >> I was like, "Yeah, I I probably should [38:39] have spent a little bit of time on that. [38:41] I think it'd be fun, but I got to figure [38:42] out this this like but the personal [38:44] brain thing. I'm going to figure this [38:45] out cuz this is high on my list of [38:48] things that I wish I had." [38:50] >> Yeah, it's amazing what the uh you can [38:53] do with just markdown and JSON um if [38:56] this isn't like public facing software. [38:58] So, uh, that low friction of it has been [39:00] one of the most revoly things for me is [39:02] like, oh, right. When you're just saving [39:04] everything to a JSON file, everything [39:07] gets way more simple, right? So, uh, [39:12] that's all I got. Uh, Wes, do you have [39:13] anything else? [39:14] >> That's it. [39:15] >> Cool. We want to do sick picks and then [39:17] get out of here. [39:18] >> Yeah. [39:19] >> Cool. Uh, sick pics. Do you have any? [39:22] >> Yeah, I'm going to sick pick a Roku. So, [39:25] we are mostly in our house, we are a [39:29] >> a Google TV family for the most part. [39:32] Meaning that we have a lot of these like [39:33] >> Do you have a yard sign? [39:34] >> Chcast. Sorry. [39:36] >> In this house. Do you have a yard sign [39:37] in this house? [39:38] >> In this house, we are a chcast family. [39:41] >> No, like we have a bunch of these like [39:43] >> Cloudflare. [39:44] >> Yeah, Cloudflare. We use only Cloudflare [39:47] and Chrome [39:48] >> in this house. No, I have a bunch of [39:49] these like chcasts that you plug in, [39:52] right? And of fantastic and amazing. And [39:54] of course, because they're fantastic and [39:55] amazing, it seems like Google's going to [39:57] phase them out. But I we recently put a [40:01] TV in our gym. [40:03] >> And the problem with the TV in the gym [40:05] was that it's if you're on the treadmill [40:07] or if you're like blasting something, [40:09] it's it's kind of loud for everybody [40:11] else in the house. So, I was trying to [40:12] figure out like how do I get the TV [40:15] sound through my AirPods without having [40:18] to pair them with the TV because like [40:21] that's a pain in the butt. Pairing [40:24] Bluetooth. I have this pair of [40:26] headphones which I have paired to my [40:28] phone and my computer and like switching [40:31] between them absolutely awful. And [40:34] apparently if you have an Apple TV [40:35] switching AirPods is is a little easier. [40:37] But [40:38] >> I was I was trying to figure out like [40:40] how do I [40:41] >> get the audio through my phone and and [40:45] chcast doesn't do that. And somebody on [40:46] Twitter says if you get a Roku, [40:49] >> you can watch stuff on the Roku but then [40:51] use the Roku app to pipe the audio [40:53] through your phone which then in turn if [40:56] you got headphones on, it goes right [40:57] through that. And it works amazingly [41:01] well. It works so well. There's no lag [41:03] in it. It sounds really good. And now [41:05] you can just like watch something on the [41:07] Roku on the TV, but then have all of the [41:09] audio come through your phone or or [41:12] through your your AirPods or whatever [41:14] Bluetooth headphones you you're working [41:16] on. And I was amazed at how well it [41:18] worked. Plus, like the Roku UI, very [41:22] like very responsive, very nice and [41:24] fast. It's like 25 bucks or something [41:28] for a a 4K stick. It's a good good [41:31] product. We have a Roku TV at the [41:33] cottage, but I've never like bought one [41:35] of their sticks before and I I might be [41:37] converted to a Roku. [41:40] >> Yeah, I just I do the same thing with [41:43] the chcasts everywhere. But yeah, what's [41:45] it that it feels like that that isn't [41:47] one of those products that Google's [41:48] like, "This is really good. We're just [41:49] going to ignore it forever." Which is [41:51] drives me nuts because some of them are [41:53] just get like frustrating like uh I [41:55] can't even Yeah. I can't get mine to [41:56] connect to my speaker re reliably or [41:59] anything like that. So, yeah. Yeah, it's [42:02] Chcast. I'm going to s pick a iPhone [42:05] app. Um, if you're an iPhone user, you [42:08] might know that the file app sucks [42:11] dearly. And every single file explorer [42:14] app that you find on the App Store also [42:17] sucks because I don't know what it is [42:20] about file apps that people make, but [42:24] every one of them is like, you don't [42:25] just want to access your files. Let me [42:27] tell you what you want. You want a [42:28] browser. You want a VPN. You want to [42:32] connect to Google Drive and all these [42:34] things and do a cloud file syncing. I [42:37] just want something that is Finder like [42:39] a file system manager that's up to date. [42:42] I can connect to my servers, SSH, [42:45] whatever, and just get access to my dang [42:47] files on my own network. Um, so I [42:50] actually was uh complaining about this [42:51] on Twitter as one does and I was given [42:55] the suggestion of the app from Booger [42:59] Butt Cheek on uh Twitter. Booger Booger [43:03] Butt Tweet uh Booger Butt Cheek [43:07] suggested FTP Manager and it's a nice [43:11] simple little app, nice little files [43:13] app. does everything I need it to do and [43:16] nothing more and is free. Works super [43:19] well. Very minimal. So, shout out Booger [43:22] Butt Cheek for that suggestion. [43:24] >> That accesses your just your local [43:26] files. It's has nothing. You're not even [43:28] using the FTP feature. [43:30] >> I It's called FTP manager because it can [43:33] FTP, but I just have my my uh local [43:36] things connected via SSH on here and [43:39] they just show up as sources and then [43:42] that's it. I use it for the files on my [43:44] phone and then the SSH into my machines [43:47] on my [43:48] >> Oh, so you can access your computer's [43:50] file system as well. [43:52] >> Nice and easy. The fi you can do that in [43:54] the files app, but I found something [43:56] very frustrating. It caches the life out [43:58] of those files. I was like looking at a [44:01] file and then looking at my computer and [44:03] just being like, where why is this [44:04] different? There's no way to clear the [44:06] cache. There's like no uh sync setting [44:08] cuz it's supposed to be just accessing [44:10] it. And I'm just like, why why are you [44:12] doing this button? [44:14] >> Yeah. No. Uh, they they direct you to [44:17] preview the preview app that they now [44:19] have on the phone. They direct you to [44:20] that to open all photos. So, I have a [44:22] screenshot in my files app. I click on [44:25] that screenshot and it opens it in the [44:27] preview app and then that is like it's [44:30] like don't just open the just show me [44:32] what the the screenshot looks like. [44:34] Yeah. Drive me nuts. [44:37] >> Oh, that's great. I'm going to check [44:39] that out. That's that's a problem I have [44:41] is like I want sometimes I want to [44:42] access my computer from the go and if [44:46] that's the case I [44:48] >> I usually use the Dropbox app actually [44:50] because most things that I care about [44:52] I'll I'll throw in a Dropbox app or if [44:54] it's like code I usually have it on on [44:56] GitHub or whatever but I'm not a dual [44:58] computer guy. I have my laptop so I [45:00] usually have it with me. Yeah, this tail [45:02] scale stuff has changed my tune on that. [45:04] I think for my next setup, I'm going to [45:06] get a big old beefy Mac Studio there. [45:08] And then I'm going to get a cheapo Mac [45:10] Air, MacBook Air, and take that with me [45:12] everywhere and just do everything [45:14] between the two because running the dev [45:16] server on that through tail scale, you [45:18] still get the hot reloading and all that [45:20] stuff going on. It just the processing [45:23] is being done on a different machine. [45:25] >> Thin client I'm in. I I want the I often [45:28] every time I take a flight, I think this [45:30] I love the 16inch MacBook Pro, but it [45:32] you can't work on it on a plane. It's [45:35] It's way too It's too big. It's just [45:36] like a little the 15-in you could just [45:39] open it enough to work on a plane, [45:41] >> but the 16 ines is too big [45:43] >> or planes are getting smaller. I think [45:44] that's the same that's happening as [45:46] well. [45:46] >> Oh, yeah. Yeah. All right. All right. [45:49] Cool. Well, thanks for watching. Hit [45:51] subscribe. All that good stuff. Let us [45:52] know. Are you building any cool personal [45:54] software? Did you get inspired by any of [45:55] this? Let us know what you are building, [45:56] what you want to build, or do you think [45:58] all this is stupid and you hate it? [46:00] Peace. [46:01] >> Peace.