Side Project Spotlight

Hosted by Philly CocoaHeads

Technology

Website RSS feed

Episodes

113

Latest episode

Jun 2026

Language

About the show

Side Project Spotlight is a podcast for app builders, documenting the process of producing real apps for the Apple App Store using Swift, SwiftUI, and other technologies.

Listen to episodes

60 recent

June 9, 2026Episode 11345 min

#113: WWDC26 — No, Seriously, Siri Works Now

Recorded thirty minutes after the WWDC26 State of the Union keynote ended, The Trio delivers a hot-take reaction to everything Apple just announced. Steve makes his boldest claim yet: 2026 is the year the "Universal UI" era begins, anchored by a Siri demo that appeared to actually work in real time. Xcode quietly Sherlocked the Codex app, SwiftUI got reorderable containers and (finally) AsyncImage caching, and Aaron spotted some very suspicious folding phone tea leaves in the new Simulator replacement.## Chapters00:08 Introductions 01:31 Reviewing The Trio's "Universal UI" Concept 02:39 Comparison of "AI" Apps: Siri, Claude, Codex, ChatGPT 05:43 Multimodal Prompts & Private Cloud Compute 07:26 Foundation Model Device Requirements 09:50 Dynamic Profiles and Custom Model Configurations 13:41 Xcode 27 Sherlocked the Codex App 15:42 Xcode and Developer Tool Evolution 21:47 SwiftUI Updates: Reordering and AsyncImage Cache 26:14 A Grab Bag of Random Stuff 28:23 App Actions and Siri Integration 35:04 No Apple Claw? 39:09 Swift Compiler Unable to Type Check Error 40:37 Final Impressions 42:02 Folding Phone Tea Leaves 43:07 Snow Leopard Speed Improvements 43:53 Wrap Up & One More Thing... 45:50 Tag ## Show Notes- Steve declares 2026 the start of the "Universal UI era," with a live Siri demo that actually worked as his primary evidence.- Aaron clocked the demo as mostly staring at a loading spinner; Steve argues Apple had to prove the on-device inference wasn't faked this time.- The Foundation Models framework supports dynamic profiles: configurable system prompts, temperatures, and thinking budgets per scenario within a single app.- Xcode 27 ships an agentic coding UI seemingly inspired by the Codex app, prompting Kotaro to ask point-blank: "Are you saying they Sherlocked Codex?"- SwiftUI finally has a reorderable container, which The Trio immediately wants in Bento Fit after a previous attempt even an "AI" agent couldn't pull off.- AsyncImage gets a built-in cache after years of third-party workarounds; Steve suspects some intern with an unlimited Claude Code budget finally got it done.- App Actions now supports natural language invocation without requiring specific phrases or app name mentions, though exact limits remain fuzzy.- Aaron flags resizable iOS windows (previously iPad-only) and an arbitrary-aspect-ratio Simulator replacement as very suspicious folding phone tea leaves.- Kotaro closes on Snow Leopard-style speed wins across the board, including 80% faster AirDrop, because speed is still a feature worth shipping.## Links**One More Thing**Cleo Family: https://www.cleofamily.app/track**PhillyCocoa:** https://phillycocoa.orgIntro music: "When I Hit the Floor", © 2021 Lorne Behrman. Used with permission of the artist.

May 25, 2026Episode 11242 min

#112: WWDC26 — Xcode Pro

With WWDC26 just around the corner, The Trio burns off some Google I/O feelings before getting to the speculation they actually care about. Steve breaks down why AI-slopified search might be quietly destroying the Web's economic model, Aaron wonders what any of these agentic tools actually do right now, and Kotaro lays out a case for why this might finally be the year Siri stops being a punchline. Also: Xcode Pro is coming. Probably. Apple, please don't.## Chapters00:00 Introductions 00:52 Google I/O — "AI" All the Things 09:12 Google I/O — Killing Search with "AI" Slop 13:32 Google I/O — "Agentic Commerce" is Coming For Your Business 15:09 WWDC26 Speculations — Dynamic Widgets 17:58 WWDC26 Speculations — New Siri (Finally) 20:36 WWDC26 Speculations — AppleClaw & Xcode 26:27 WWDC26 Speculations — Folding Phone Tea Leaves 28:24 WWDC26 Speculations — HomeOS 31:11 WWDC26 Speculations — Out of the Box 37:32 WWDC26 Speculations — New Foundation Model 41:49 Outro & One More Thing... 42:51 Tag## Show Notes- Google I/O 2026 went all-in on AI, with Kotaro noting the near-total absence of Kotlin, Jetpack, and Flutter talks at this year's developer sessions.- Gemini 3.5 Flash launched as the new "affordable" model and turned out to be significantly more expensive than its predecessor.- Steve takes apart Google's AI search redesign, arguing it quietly destroys the economic model that funds the Web, including Google's own ad revenue.- Aaron's deadpan verdict on Google's AI search demos: "the only thing it showed was it generating these giant slop docs. Who wants to read those?"- Kotaro and Aaron speculate that WWDC26 could bring more dynamic, context-aware widgets, ones that are smarter about timing and context than the current static rectangles.- The Trio agrees the headline WWDC feature is a Siri that actually understands intent, with Shortcuts workflow building as a hopeful bonus.- Kotaro floats an "AppleClaw" style personal assistant via iMessage, letting developers submit agentic tasks to an Xcode Cloud backend.- The folding iPhone question comes up: does it run iPadOS, iOS, or something in between, and how will apps scale across the form factor?- Steve expects Apple to plant HomeOS seeds at WWDC, APIs and features that will only make full sense once a HomePod-with-a-screen arrives later in the year.- The Trio caps the WWDC wishlist by accidentally inventing Xcode Pro, Apple's inevitable premium developer subscription tier.## Links**Google I/O 2026**Google I/O 2026: https://io.google/2026/Everything Announced at Google I/O 2026 in 13 Minutes: https://youtu.be/qCfARlv74jQ | 100 Things We Announced at I/O 2026: https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/**WWDC26**WWDC26: https://developer.apple.com/wwdc26/**One More Thing**AppJawn LLC: https://appjawn.comApps: Clipdish, Mio Vino, Minimalist Meditation Timer**PhillyCocoa:** https://phillycocoa.orgIntro music: "When I Hit the Floor", © 2021 Lorne Behrman. Used with permission of the artist.

May 11, 2026Episode 1111 hr 9 min

#111: A Bazooka of Syntax

Steve finally fixed phillycocoa.org, and the journey from broken CircleCI pipelines and hijacked S3 buckets to a blazing-fast Cloudflare Pages site took one Side Project Saturday and an embarrassing number of Codex tokens. Then The Trio turns to the AI hype machine, and they're tired: tired of opaque token costs, tired of reviewing generated code that complicates everything it touches, and tired of an industry that mistakes syntax speed for software engineering. Fred Brooks called it in 1986, and The Trio is calling it now.## Chapters00:00 Introductions01:47 The Journey of Updating the Website06:38 Challenges with CircleCI and S3 Buckets09:23 Exploring Cloudflare Pages11:14 Navigating Cloudflare's User Interface14:22 Setting Up Automatic Deployments17:35 Managing DNS and SSL with Cloudflare23:07 LLM Development Fatigue26:15 Navigating Concerns and Costs in AI Usage29:11 LLMs are No Silver Bullet31:57 The Exhaustion of Code Review and Architectural Decisions36:25 Token Management and Cost Awareness in AI Tools40:07 The Economics of AI and Software Development42:45 The Hype vs. Reality of AI Tools46:34 Future Prospects of LLMs and Universal UI50:16 The Future of Edge Computing with LLMs53:08 The Evolution of Software Development and AI Integration54:17 AI in Sci-Fi: Myths vs. Reality57:54 The Challenges of Local Models and Hardware Limitations01:03:21 Outro & Upcoming Event01:09:21 Tag## Show Notes- Steve spent Side Project Saturday migrating phillycocoa.org from a broken CircleCI/S3 setup to Cloudflare Pages, burning his entire weekly Codex token budget in about three hours.- Cloudflare Pages handles Hugo builds automatically and manages SSL and CDN without manual config, all on a free tier that's plenty for the site.- Cloudflare's UI hides the Pages "Get Started" link below giant worker buttons, which Kotaro calls "the weirdest dark pattern."- Steve argues that syntax generation was never the real bottleneck in software engineering, citing Fred Brooks' 1986 essay "No Silver Bullet."- Aaron is worn out from reviewing AI-generated code and still having to make every architectural decision himself.- LLM costs are nearly impossible to forecast: a single prompt can burn a significant chunk of your plan, depending on model, tool calls, and context.- The Trio sees firms rushing to adopt LLM tooling before the ROI math makes sense, driven by hype rather than evidence.- ThePrimeagen's recent take on the shifting AI economy lines up with what Steve sees at work: token-based billing is starting to expose the real cost.- The Trio agrees local models running on personal hardware are the interesting long-term play, but RAM shortages make even basic setups expensive.- Kotaro closes with a dad joke: he thought his LLM skills landed him his current job, but it turns out...## Links**PhillyCocoa.org Update**Website: https://phillycocoa.org**Articles & Essays**"Let's talk about LLMs" by James Bennett: https://www.b-list.org/weblog/2026/apr/09/llms/"No Silver Bullet" by Fred Brooks: https://www.cs.unc.edu/techreports/86-020.pdf**Videos**"The AI economy is about to change" by ThePrimeagen: https://youtu.be/_Q-e_nczWqM**One More Thing**"Beyond the Simulator: Perspectives on Modern App Development": https://luma.com/i00ll61z**PhillyCocoa:** https://phillycocoa.orgIntro music: "When I Hit the Floor", © 2021 Lorne Behrman. Used with permission of the artist.

April 27, 2026Episode 1101 hr 0 min

#110: So Long, Tim Apple, and Thanks for All the Fish!

Tim Cook is stepping down in September, and The Trio has plenty of thoughts on what the Ternus era means for Apple. Kotaro dives into his embedded systems rabbit hole (Raspberry Pis, ESP32s, and a Godot refresher), while Steve sounds the AI hype alarm, comparing the current frenzy to NFTs and the Metaverse, complete with a shoe company that somehow pivoted to GPU data centers on a $50M budget. Steve's monitor saga drags on, the SpaceX/Cursor "announcement of an announcement" gets the skepticism it deserves, and The Trio wraps up with details on the May 14 IRL meetup in Philly.## Chapters00:00 Introductions05:54 Kotaro's Side Project Adventures08:29 Diving into Hardware and Embedded Systems11:17 Raspberry Pi Adventures and Microcontrollers14:02 Creating AI Projects with Raspberry Pi18:19 Exploring DIY Devices and Learning in Tech23:21 Game Development and Learning Curves24:16 AI Tools and Programming Challenges26:55 The AI Hype Update and Economic Realities36:57 Balancing AI Use in Software Development39:53 The Hype Cycle of AI and Media44:32 So Long, Time Apple, and Thanks for All the Fish!53:07 The Future of Apple in the Ternus Era56:43 Steve's Monitor Watch Update58:57 Wrap Up01:00:37 Tag## Show Notes- Tim Cook announced his retirement as Apple CEO, effective September, with hardware chief John Ternus set to take the helm.- The Trio agrees Cook grew Apple into the world's most valuable company, and the MacBook Neo might just be his most quintessential product.- Ternus is seen as more of an engineer/visionary, and Steve is cautiously hoping he'll bring more Jobs-era decisiveness to Apple's product direction.- Kotaro is deep in embedded systems this year, learning Raspberry Pi 5s and ESP32 microcontrollers the hard way (wrong cables, wrong GPIO boards, all of it).- He's built a basic AI chatbot device (think DIY Rabbit R1, hooked to Google Gemini) and is eyeing a 5-inch touchscreen home automation kiosk.- TRMNL, the E Ink dashboard device, comes up as a goal Kotaro is working toward, though the large version is sold out.- GitHub Copilot paused new signups, dropped Opus from Pro plans, and started rationing usage, which Steve reads as AI's economic reality finally catching up.- Steve puts AI hype at NFT/Metaverse levels: a shoe company pivoted to GPU data centers, and SpaceX "announced" it has the option to buy Cursor for $60B without actually buying anything.- Steve's XDR monitor watch continues: he watched a glowing review, still can't justify the price, but is eyeing the nano-texture option for his glare-heavy room.- The Trio closes with news of a PhillyCocoa IRL meetup on May 14 at the Vanguard building, featuring Kotaro on Metal shaders.## Links**Hardware & Devices**TRMNL: https://trmnl.com/ | Rabbit R1: https://www.rabbit.tech/rabbit-r1**Snazzy Labs TRMNL Review**Watch: https://www.youtube.com/watch?v=YWw5NKUx40o**AI Hype Update**We are near peak hype (Primeagen): https://www.youtube.com/watch?v=rAREqdtUN48SpaceX/Cursor ($60B): https://www.reuters.com/technology/spacex-says-it-has-option-acquire-startup-cursor-60-billion-2026-04-21/**One More Thing**IRL Meetup RSVP (May 14): https://luma.com/i00ll61z**PhillyCocoa:** http://phillycocoa.orgIntro music: "When I Hit the Floor", © 2021 Lorne Behrman. Used with permission of the artist.

April 13, 2026Episode 10951 min

#109: Instant Wrong Answer

The Trio tests Google's new Gemma 4 model running locally on an iPhone and gets an instant wrong answer, which sets the tone for a sprawling conversation about AI hype, Anthropic's Mythos release, and the social media panic Steve dubs "mythos psychosis." Along the way, token maxing emerges as the new worst productivity metric, Kotaro reminds developers they could just do things themselves, and Steve makes a compelling case for building your own push notification system on Apple platforms. Plus, the never-ending monitor saga inches closer to a verdict.## Chapters00:00 Introductions00:37 "AI" Hype Bubble Updates02:39 Mythos and Myth Making05:24 Local AI Models and Their Performance12:57 Mission Impossible: Mythos Psychosis17:09 Responsible Token Maxing23:42 Local Models and Personal Assistants26:56 You Could Just Do It Yourself30:02 Modern Push Notifications37:25 Building Your Own Push Notification System43:14 The Value of Push Notifications in Apps47:03 Steve's Monitor Upgrade Update50:34 Wrap Up51:30 Tag## Show Notes- Anthropic's Mythos, Google's Gemma 4, and Meta's Muse Spark all dropped within days, kicking off The Trio's AI hype bubble update segment.- Mythos is reportedly 5x more expensive than Opus, and Anthropic says it's too dangerous to release publicly, so only a handful of partners have access.- Steve demos Gemma 4's 2B parameter model on his iPhone via Google's AI Edge Gallery app: it's blazing fast, but instantly wrong about how many Rs are in "raspberry."- Aaron points out small local models like Gemma are meant to be fine-tuned for specific tasks, not used as general chatbots.- Steve coins "mythos psychosis" for the social media meltdown over AI capabilities, with influencers posting videos titled things like "the end of software."- Token maxing is apparently real at some companies, where managers measure dev productivity by token usage; The Trio helpfully workshops a strategy of generating and deleting code to keep your codebase in perfect equilibrium.- Kotaro's advice for devs reaching for AI on simple tasks: you're a good developer, just do it yourself.- Steve dives into modern Apple push notifications, highlighting the Push Notifications Console, broadcast push for live activities, and how much easier setup is with SwiftUI.- Building your own push server is just HTTP/2 calls to Apple's API, and Apple doesn't charge for sending notifications at indie scale.- The monitor saga continues: Aaron and the LLMs keep pushing Steve toward the Pro Display XDR, which now has a VESA mount discount making the price slightly less painful.## Links**Modern Push Notifications**Apple Push Notifications: https://developer.apple.com/notifications/**One More Thing**Slopes: https://getslopes.com**PhillyCocoa:** http://phillycocoa.orgIntro music: "When I Hit the Floor", © 2021 Lorne Behrman. Used with permission of the artist.

March 30, 2026Episode 10844 min

#108: AppleClaw

iOS 26.4 just dropped and Steve is thrilled that the keyboard finally works again. The Trio digs into the new Music app concerts feature (powered by Bands in Town, probably), which leads Steve into a passionate case for the local music scene, Guinness floats, and why authenticity matters more than ever in the age of AI-generated slop. From there, Kotaro floats a wild idea: what if Apple built their own version of OpenClaw using iMessage and their own hardware? Steve points out Apple already has the pieces in place with App Intents and Shortcuts, and the WWDC speculation spirals into distilled Gemini models, local inference on M5 hardware, and Marco Arment's absurd 48 Mac mini data center rack. Steve also reports back from his Apple Store recon mission on the Studio Display vs. the XDR, and Aaron keeps egging him toward the expensive one.## Chapters00:00 Introductions & OS Updates05:48 The Local Music Scene and Its Importance08:35 Authenticity in Music and Art11:40 AI and Its Impact on Creativity14:33 WWDC26 and "AppleClaw?"27:04 Exploring AI Model Parameters and Storage Needs28:31 The Future of Apple "AI" Services30:28 Local vs Cloud Inference: The Power Struggle32:50 Steve's Monitor Update43:13 Wrap-Up43:32 One More Thing...44:48 Tag## Show Notes- iOS 26.4 is out and Steve says the iPhone keyboard actually works now, which is apparently the highlight of the whole release.- The Music app's new concerts feature surfaces local shows based on your listening history, with ticket links through Bands in Town.- Steve makes the case for local music: cheaper shows, interesting venues, accessible artists, and the guarantee that you're not listening to AI-generated slop.- The Trio agrees AI art works as a stock photo replacement but loses something the moment you know it's generated.- Kotaro pitches "AppleClaw," the idea that Apple could build an OpenClaw-style agent using iMessage and their own hardware.- Steve thinks Apple is well positioned since they already have App Intents, Shortcuts, and a Gemini backend they can distill into local models.- Marco Arment apparently has 45+ Mac minis in a data center rack for transcoding podcasts, and yes, he rents actual data center space for them.- The M5 chips can handle useful local inference on 30B parameter models, and Apple's power efficiency gives them an edge over GPU rigs that melt cables (looking at you, PewDiePie).- Steve visited the Apple Store and confirms the XDR display has the best HDR he's ever seen, but he can't unsee the fuzziness of nanotexture.- The monitor decision is down to a glossy Studio Display or the BenQ MA Series, with Aaron lobbying hard for the XDR.## Links**Apple**iOS 26.4: Available now on all Apple platforms**AI & Agents**Welcome to Gas Town: https://steve-yegge.medium.com/welcome-to-gas-town-4f25ee16dd04#55: The "Universal" UI: https://podcast.phillycocoa.org/episodes/55-the-universal-ui**Apps**AppJawn LLC Apps: https://appjawn.com/#apps**One More Thing**SwiftUI Architecture Book by Mohammad Azam: https://azamsharp.school/swiftui-architecture-book.html**PhillyCocoa:** http://phillycocoa.orgIntro music: "When I Hit the Floor", © 2021 Lorne Behrman. Used with permission of the artist.

March 16, 2026Episode 1071 hr 2 min

#107: Buttery Smooth Text

The Trio declares this an AI-free episode and dives into Apple's latest hardware announcements. Kotaro, Steve, and Aaron break down the new MacBook Neo, its surprisingly capable A18 chip, aluminum build, colorful design, and its potential to expand the Mac market at just $600. Then Steve takes The Trio on a deep dive into his increasingly desperate search for the perfect external monitor, weighing the refreshed Studio Display, the new Studio Display XDR, BenQ's upcoming 5K Mac monitor, Dell's 4K Thunderbolt hub display, and BenQ's programmer-focused 3:2 aspect ratio monitors. What follows is a lively debate over 120Hz refresh rates, macOS 4K scaling quirks, nano-texture vs. glossy glass, multi-monitor setups, MacBook Pro pricing strategies, and whether Kotaro and Aaron can convince Steve to just pick one already.## Chapters00:00 Introductions01:48 MacBook Neo18:31 The Studio Display (2026)22:17 The Studio Display XDR25:26 Understanding 120 Hertz Displays27:49 Evaluating Cost vs. Performance in Monitors29:24 Comparing Alternatives to the Studio Display31:27 The BenQ Monitor: A Viable Contender33:06 Dell 4K and macOS Scaling36:43 BenQ Programmer Series Monitors42:28 Multiple Monitors vs. One High-End Display43:55 Navigating MacBook Pro Configurations45:34 Understanding RAM and Pricing Strategies48:11 Choosing the Right Display for Your Needs53:33 A Digression About Apple Care Prices55:53 A Monitor Intervention01:00:30 Wrap-Up01:00:58 One More Thing...01:02:24 Tag## Show Notes- Apple announces new hardware including the MacBook Neo, refreshed Studio Display, and the new Studio Display XDR- The MacBook Neo starts at $600 ($500 education), features an A18 chip, aluminum unibody, colorful options, and is capable enough for 4K video editing in Final Cut- Touch ID is a $100 add-on: The Trio agree it's worth the upgrade- The refreshed Studio Display gains Thunderbolt 5 and daisy-chaining support but remains 60Hz- The Studio Display XDR starts at $3,299 with mini-LED backlighting, 120Hz ProMotion, and the stand included- Steve's monitor search considers: Studio Display ($1,500 edu), Studio Display XDR ($3,200 edu), BenQ 5K Mac monitor (~$1,000), Dell 4K Thunderbolt hub (~$800), and BenQ RD280UG programmer monitor (~$700)- Discussion of macOS 4K scaling issues: macOS renders at 5K and downsamples to 4K, which can cause artifacts- The BenQ programmer monitor features a 3:2 aspect ratio (28"), 120Hz, dark/light mode presets, and a halo backlight, but weaker color reproduction- Nano-texture vs. glossy glass: nano-texture reduces glare but can appear slightly fuzzy; standard glass is easier to clean- Apple's MacBook Pro pricing now ties higher RAM options to higher chip tiers, effectively bundling price increases- AppleCare One costs more per additional device ($6/mo) than standalone AppleCare Plus ($5/mo): a pricing quirk Steve finds baffling- BentoFit, The Trio's health kit dashboard app, gets a plug: download it at bentofit.app- Kotaro asks if this is the "Steve intervention podcast" as the monitor debate spirals. Aaron's rational choice? Buy the XDR and keep it for 10 years. Steve remains unconvinced. Stay tuned.## Links**MacBook Neo Reviews**John Gruber (Daring Fireball): https://daringfireball.net/2026/03/the_macbook_neoTyler Stalman: https://www.youtube.com/watch?v=d-VOt9559GkMarques Brownlee (MKBHD): https://www.youtube.com/watch?v=iGeXGdYE7UELinus Tech Tips: https://www.youtube.com/watch?v=sSheV0FEYYU**Displays**ArtIsRight on nano-texture displays: https://www.youtube.com/watch?v=xzEmuA53LLE**One More Thing**AppJawn LLC: https://appjawn.comApps: Clipdish, Mio Vino, Minimalist Meditation Timer**PhillyCocoa:** http://phillycocoa.orgIntro music: "When I Hit the Floor", © 2021 Lorne Behrman. Used with permission of the artist.

March 2, 2026Episode 1061 hr 0 min

#106: No Flow State

The Trio kick off with the rumors about Apple's March 4th event and the possible return of a budget 12-inch MacBook on an A18 chip, which leads to a very poorly researched price analysis and a pitch for a MacBook sock accessory. Steve and Aaron talk about how agentic-assisted coding at work has been mentally exhausting and how they miss actually writing code. The conversation covers why LLMs are rough for greenfield projects, what "vibe coding" actually means (and why they're not doing it), the Alex Hillman episode follow-up, and Steve's experiment running different models against Bento Fit to produce slop PRs that Kotaro then spent an hour reviewing for some reason. Steve also crashes out about the state of the industry and public perception of AI. It's a lot.## Chapters- 00:00 Introductions- 01:51 Rumors and Speculations on New Macs- 10:43 The Impact of Pricing on Apple's Product Strategy- 11:53 The Developer Perspective on a New MacBook- 17:36 Comparing MacBooks and iPads in Today's Market- 18:36 The MacBook Sock- 21:12 Mac App Renaissance- 22:26 Follow-Up: Alex Hillman Episode- 25:12 Agentic Coding Flow States- 29:50 Balancing Traditional and AI-Assisted Development- 33:26 Navigating the Challenges of Greenfield Projects- 38:00 The Dilemma of AI in Coding- 41:48 Navigating Agentic Coding and Professional Ethics- 43:50 The Reality of Code Maintenance- 44:36 Public Perception of AI and Software- 47:41 Steve Crashes Out About the Industry- 51:24 Bento Fit Slop PRs- 58:37 Wrap-Up- 59:37 One More Thing...- 01:00:50 Tag## Show Notes- Apple "Experience" event March 4th, rumored budget MacBook with A18, ~12 inch, fun colors, maybe $699–$799- Updated Studio Display and touchscreen MacBooks also rumored- People buying $600 Mac Minis for OpenClaw setups- Mac app renaissance? More Mac apps being submitted, possibly thanks to LLMs making AppKit less painful- Alex Hillman episode follow-up: 219 views, 5 likes, watch hours up 46,639%- Agentic coding fatigue: Steve and Aaron are tired. No flow state. Just planning, reviewing, iterating.- Greenfield projects with LLMs produce average code. Better to write some bespoke code first and give the robot examples.- "We're not vibe coding." Steve proposes "agentic-assisted" as the term. The acronym is AA, which... maybe not great.- Code is a liability. 1,000 lines a day is not a good metric.- People outside the bubble mostly know ChatGPT, don't pay for it, and hate it- Steve ran three slop PRs on Bento Fit with different models as an experiment. Kotaro reviewed one for an hour anyway.- Bento Fit's $4.34 in tip revenue resulted in a $10 tax bill- OpenCode now has a $10/month plan for open source models## Links**Bento Fit**Website: https://bentofit.app**Tools & Services Mentioned**OpenClaw: https://openclaw.ai | OpenCode: https://opencode.ai | T3 Chat: https://t3.chat | Codex CLI: https://openai.com/codex | Claude Code: https://claude.com/product/claude-code**One More Thing**AppJawn LLC: https://appjawn.comApps: Clipdish, Mio Vino, Minimalist Meditation Timer**PhillyCocoa:** http://phillycocoa.orgIntro music: "When I Hit the Floor", © 2021 Lorne Behrman. Used with permission of the artist.

February 16, 2026Episode 1051 hr 22 min

#105: How Alex Hillman Built an AI Assistant with Claude Code

In this episode, Alex Hillman, co-founder of Philadelphia's legendary coworking space Indy Hall, takes us through his journey building a sophisticated AI executive assistant using Claude Code. What started as a simple terminal experiment in October 2025 has evolved into a full production system that autonomously manages network diagnostics, email workflows, relationship tracking, and newsletter automation. Alex shares the technical architecture, real-world stories of AI-powered problem solving, cost insights, and his thoughtful approach to building trust with AI while maintaining strong ethical guardrails.## Chapters- 00:00 Coming Up...- 02:01 Introductions- 03:57 The Origins of PhillyCocoa and Indie Hall- 06:12 The Evolution of AI and Personal Assistants- 07:35 Building a Personal Assistant with Claude Code- 10:26 The Architecture of the Personal Assistant- 14:04 Creating a Web App Interface for the Assistant- 16:10 Using Tailscale for Secure Access- 19:01 Mitigating Risks with AI Autonomy- 29:24 Backup Protocols and Data Management- 31:23 Emergent Behavior in AI Systems- 34:10 Flow State and Productivity in Programming- 37:56 Understanding AI Behavior and User Education- 39:45 Cost Management in AI Development- 45:37 Building Trust with AI Systems- 53:53 Navigating Trust in Skill Utilization- 55:23 Technical Applications for Non-Developers- 01:00:17 Innovative Personal and Business Management- 01:09:03 Transforming Workflows with AI- 01:12:56 Ethics and Responsibility in AI Usage- 01:18:25 Community Building Through Meetups- 01:21:55 Tag## Highlights**Architecture:** Claude Code headless via CLI with WebSocket communication, Docker on Hetzner VPS, Tailscale networking, hourly snapshots, git hooks for destructive commands, multi-layered security.**Real Use Cases:**- Network monitoring that diagnosed an overheating router fan from a screenshot- Email sorted by "easiest to hardest" instead of chronological- Date night tracking with restaurant and wine pairing suggestions- Organized 51 wine bottles via photos into ASCII grid layout- Newsletter reduced from 4 hours to 30 minutes while preserving human writing**Costs:** $20/month plan lasted 20 minutes. Now at $200/month. One Thanksgiving week hit $1,500 in overages during heavy development.**Philosophy:** "Modest YOLO" approach—autonomous but controlled. AI enhances human work, doesn't replace it. The system can modify itself: type "add a button," refresh, it works.**Open Source:**- **Kuato**: Session search for Claude Code- **Smaug**: Twitter bookmark archiver with AI analysis- **Andy Timeline**: Auto-generated weekly narrative of the AI's evolution## Event**Big Philly Meetup Mashup** - March 15, 2026Hackathon for Philadelphia's tech and creative communities. Theme: "Good Neighbors." Sponsored by Supabase.https://indyhall.org/goodneighbors/## Links**Alex Hillman**YouTube: https://www.youtube.com/@AlexHillman | Website: https://dangerouslyawesome.com | GitHub: https://github.com/alexknowshtml**Open Source Projects**Kuato: https://github.com/alexknowshtml/kuato | Smaug: https://github.com/alexknowshtml/smaug | Andy Timeline: https://github.com/alexknowshtml/andy-timeline**Tools & Resources**Indy Hall: https://indyhall.org | Claude Code: https://claude.com/product/claude-code | OpenClaw: https://openclaw.ai | Brian Casel: https://www.youtube.com/@briancasel | Termius: https://termius.com | Point-Free: https://www.pointfree.co/the-way**PhillyCocoa:** http://phillycocoa.orgIntro music: "When I Hit the Floor", © 2021 Lorne Behrman. Used with permission of the artist.

February 2, 2026Episode 10448 min

#104: Beyond the Simulator IRL!

The Trio meet to discuss two of the talks (ours!) from the recent in-person PhillyCocoa meetup called "Beyond the Simulator: Perspectives on Modern App Development" that took place on January 29 at the Vanguard offices in Philadelphia. This pod was recorded before the event due to scheduling, but we go into detail on what you missed now that it is the future (insert Spaceballs joke here)! Kotaro talks about Liquid Glass and what it means for modern UI/UX while Steve goes into some detail about how to effectively get started using tools like Codex CLI or Claud Code for app development. Be sure to check out PhillyCocoa.org for a link to join our Slack and follow us on Luma so you know when our next in-person and virtual events are scheduled: https://luma.com/phillycocoa.## Show Notes- Introductions- IRL Meetup Follow-up (recorded before the meetup!) - Kotaro’s Liquid Glass talk - Steve’s Spec. Plan. Ship. “AI” assisted dev talk- Wrap-Up- One More Thing... - Monthly Zoom Call Meeting in February - Follow us on Luma: https://luma.com/phillycocoa## Chapters00:00 Introductions02:33 Beyond the Simulator IRL Event03:49 Kotaro's Talk: Liquid Glass and Modern UI/UX Trends11:25 Liquid Glass Encourages Gesture-Based Interactions16:18 Branding Challenges in Liquid Glass UI19:31 Steve's Talk: Spec. Plan. Ship21:47 The Four I Workflow: Intent, Interact, Increment, Iterate28:59 Continuously Iterate on Your System33:17 Steve's Tips for Getting Started41:51 Best Practices for Using AI in Development46:20 Wrap-Up46:38 One More Thing...47:59 TagIntro music: "When I Hit the Floor", © 2021 Lorne Behrman. Used with permission of the artist.