Accidental Tech Podcast

677: I Accept the Battery Cost

February 5, 2026

Key Takeaways Copied to clipboard!

  • The debate over unit systems continues, with one host staunchly defending Fahrenheit for ambient temperature measurement against Celsius. 
  • Cloudflare Access utilizes a Zero Trust architecture, which is deemed too heavyweight for personal mini-apps compared to simpler authentication methods like OIDC providers such as Pocket ID or TSIDP. 
  • The state of consumer MicroLED technology is currently characterized by extremely high cost ($55,000 for a 157-inch model), complex installation requirements, and surprisingly high idle power consumption (600 watts when 'off'). 
  • Marco is undertaking a 'year of efficiency' experiment, leading him to switch from Safari to Chrome as his default desktop browser due to perceived speed and better integration with tools like OnePassword, despite the known battery cost on Apple Silicon Macs. 
  • Google's Gemini AI is showing increasing sophistication, particularly in integrating with Google Maps to generate actionable routes, suggesting Google is well-positioned to leverage its existing data properties for future AI advancements. 
  • Apple's release of Xcode 26.3 with agentic coding support is seen as a necessary, albeit late, move to integrate AI development tools, which is already contributing to a rapid increase in iOS app submissions, potentially overwhelming App Review. 
  • The discontinuation of the Tesla Model S and Model X signals the end of an era where Tesla focused on high-end luxury EVs, as the company has shifted its focus entirely toward high-volume, lower-margin vehicles like the Model 3 and Y. 
  • Tesla's stated goal of achieving widespread autonomous driving is contradicted by its own data, which shows its robo-taxis crash at a rate three times worse than human drivers, contrasting sharply with Waymo's significantly safer performance metrics. 
  • The speaker expresses a fundamental distrust in Tesla and Elon Musk's leadership due to perceived recklessness in product development (e.g., removing LiDAR, questionable safety practices) and political actions, making them unwilling to trust the company with their life in an autonomous vehicle. 

Segments

Icepocalypse Follow-up and Unit Debate
Copied to clipboard!
(00:00:00)
  • Key Takeaway: One speaker’s children returned to school after an extended closure due to an ‘Icepocalypse,’ leading to a renewed, passionate defense of the Fahrenheit temperature scale.
  • Summary: School resumed after an extended closure caused by ice, prompting a discussion on unit conversions. The speaker strongly asserted that Fahrenheit is superior for ambient air temperature because it relates directly to human perception of hotness, unlike Celsius. They dismissed Celsius as ’trash’ for ambient temperature, though conceded its use in cooking.
Cloudflare Authentication Alternatives
Copied to clipboard!
(00:04:36)
  • Key Takeaway: Cloudflare Access implements a Zero Trust architecture, which requires a Zero Trust plan and is generally too complex for securing small personal applications.
  • Summary: John explored alternatives for authenticating his personal apps, including Cloudflare Access, which enforces Zero Trust principles where no user or device is trusted by default. OpenID Connect (OIDC) was suggested as another enterprise-friendly standard, with Pocket ID and TSIDP offered as lightweight, self-hostable OIDC providers for simpler needs.
Per-Pixel Lighting vs. Dimming Zones
Copied to clipboard!
(00:10:35)
  • Key Takeaway: While high zone counts in LCDs (like 5,000 per inch on MacBooks) mitigate blooming, per-pixel lighting control is ultimately superior because dynamic backlights can still cause issues, especially with high-contrast content like starfields.
  • Summary: The discussion referenced David Schaub’s point that display technology only needs to meet human perception thresholds, suggesting high zone counts might suffice. However, worst-case scenarios, like starfields on black backgrounds, still stress dynamic backlights by forcing the entire backlight region on. Per-pixel lighting remains the ideal solution, avoiding computational complexity and lag associated with dynamic backlight management.
MicroLED TV Reality Check
Copied to clipboard!
(00:12:26)
  • Key Takeaway: The current state of large MicroLED TVs involves extremely high costs ($55,000 for a 157-inch model), modular panel construction, and significant power consumption, even when the display is off.
  • Summary: A review of a $55,000 MicroLED TV revealed it is built from 27-inch modular panels requiring complex installation, including 18 Cat6 wires and three dedicated power drops. Performance is excellent (800 nits full-field brightness, minimal blooming), but the unit is not full 4K and requires an additional $1,000 processor to function as a standard TV with HDMI inputs. The system consumes 1,200 watts at full brightness and an astonishing 600 watts when the screen is off.
AI Ethics and Industrial Revolution Analogy
Copied to clipboard!
(00:25:20)
  • Key Takeaway: The ethical and legal issues surrounding AI training data theft and societal impact remain entirely unresolved, mirroring the long, harmful transition period of the Industrial Revolution.
  • Summary: Despite using tools like Claude Code, the hosts affirmed that unresolved issues like data compensation, job displacement, and environmental impact persist, comparing the AI revolution to the Industrial Revolution’s initial period of severe worker exploitation and pollution. Staying in the technology business necessitates engaging with AI while simultaneously fighting for fair rules and sustainable practices, as sitting out means abandoning the industry’s future.
Renewable Energy Progress and Vision Pro Content
Copied to clipboard!
(00:40:51)
  • Key Takeaway: Technological progress in renewables is significant, exemplified by a recent Irish wind farm upgrade where one new turbine produces more power than all 21 older turbines combined.
  • Summary: A video from Technology Connections highlights that global renewable energy adoption is strong, despite US struggles, showing significant efficiency gains in wind power technology. Separately, the immersive Crufts documentary on Apple Vision Pro was criticized for its heavy use of rapid cuts and narrow vertical field of view, making it feel more like 3D content than true immersive media.
Marco’s Efficiency Experiment
Copied to clipboard!
(00:51:56)
  • Key Takeaway: Marco initiated a personal theme for the year focused on identifying and addressing accumulated inefficiencies and friction points in his life, rather than ruthlessly cutting them.
  • Summary: Marco announced his focus for the year is efficiency, aiming to become aware of long-tolerated friction points in his life. This involves considering existing inefficiencies and deciding whether to implement changes, contrasting with the hosts’ initial perception that Marco already minimizes such tolerance.
Marco’s Efficiency Experiment
Copied to clipboard!
(00:51:56)
  • Key Takeaway: Marco initiated a ‘year of efficiency’ focus to eliminate personal friction, leading him to re-evaluate long-held technology choices.
  • Summary: Marco’s efficiency focus prompted him to examine inefficiencies in his life, including the hosting stack of Overcast and financial planning management. A major point of friction identified was the poor integration between OnePassword and Safari, prompting a review of password manager allegiance.
AI Chatbot Usage Comparison
Copied to clipboard!
(00:55:53)
  • Key Takeaway: Google Gemini is showing factual accuracy advantages over ChatGPT, suggesting its model sophistication is slightly ahead in certain areas.
  • Summary: Marco has significantly increased his use of AI chatbots like ChatGPT for small queries, replacing traditional web searches. He observed that Google Gemini is often more accurate when answering factual questions compared to ChatGPT. Despite Gemini’s perceived lead in model sophistication, it still possesses its own quirks and annoyances.
Google’s AI Ecosystem Advantage
Copied to clipboard!
(00:57:25)
  • Key Takeaway: Google’s existing infrastructure—including cloud, custom chips, and data sources like Maps—positions them strongly for the future of integrated AI services.
  • Summary: Google is considered well-positioned for the AI future because they control their own models, cloud infrastructure, and custom chips, unlike competitors like OpenAI. Their integration of Google Maps into Gemini allows for powerful, actionable requests, such as generating custom walking routes that can be immediately sent to the phone.
Consolidating Mobile Map Apps
Copied to clipboard!
(00:59:14)
  • Key Takeaway: Marco consolidated his use of three separate mapping applications (Waze, Google Maps, Apple Maps) into exclusively Google Maps for efficiency.
  • Summary: Marco previously bounced between Waze for traffic navigation, Google Maps for business data, and Apple Maps for walking directions. He found Apple Maps’ lock screen takeover feature significantly slowed down phone unlocking, and Google Maps offered superior business data (hours, ratings, photos) compared to Apple Maps. Google Maps is now his sole mapping application due to its speed and comprehensive features.
Switching Search Engines and Browsers
Copied to clipboard!
(01:04:29)
  • Key Takeaway: Driven by efficiency and the poor quality of current web search results, Marco switched from DuckDuckGo/Kagi back to Google Search and is testing Chrome as his default browser.
  • Summary: Marco abandoned DuckDuckGo due to his reliance on its bang shortcuts, which he now replicates via extensions or by switching search engines. He found Kagi good for web search but realized he was using it less for search and more for Amazon shortcuts. The poor quality of Google’s traditional search results, exacerbated by bad AI summaries, is offset by the superior speed and functionality of Chrome, especially with OnePassword integration.
Chrome Performance vs. Safari
Copied to clipboard!
(01:12:30)
  • Key Takeaway: Switching from Safari to Chrome on macOS provided a dramatic, baseline performance increase, comparable to the shift from Intel to Apple Silicon.
  • Summary: Marco reports that everything in Chrome—opening tabs, loading websites—is significantly faster than in Safari, despite Safari traditionally being better for power consumption. Chrome’s superior handling of form autofill and OnePassword integration has saved him significant time previously wasted on manual data entry. He accepts the increased battery cost as a worthwhile trade-off for the regained time and reduced friction.
AI Impact on Software Development
Copied to clipboard!
(01:22:29)
  • Key Takeaway: The immediate integration of agentic coding into Xcode 26.3 RC signals that AI is rapidly reshaping software development, leading to an explosion in app submissions.
  • Summary: Apple released agentic coding support quickly, allowing agents like Claude and Codex to interact with Xcode documentation and previews, indicating they recognize the urgency of AI integration. John used these agents to solve his Tahoe icon issue via a complex workaround, demonstrating their utility for ‘rubber ducking’ complex problems. The resulting ease of app creation is already causing a sharp increase in monthly iOS app submissions, potentially straining App Review.
Future of Coding and AI Skills
Copied to clipboard!
(01:33:04)
  • Key Takeaway: Coding agents are improving rapidly because code is tractable and testable, unlike general knowledge LLMs, making proficiency in using these agents a critical, evolving skill for developers.
  • Summary: Coding agents are currently on a steep upward curve of improvement, unlike general-purpose chatbots, because code execution allows for immediate, objective testing and iteration. Mastering these agents is becoming a necessary skill, similar to learning Photoshop, where deep knowledge yields significant professional advantage. Ignoring this technological shift risks being left behind, similar to ignoring the internet revolution in the late 1990s.
Model S/X Discontinuation Rationale
Copied to clipboard!
(01:46:59)
  • Key Takeaway: Tesla ended the Model S/X production because the aging platform failed to keep pace with competitors, despite initial luxury dominance.
  • Summary: The Model S, once the best EV, stopped keeping up with competition, leading to its end despite minor tweaks over the years. The decision to end production is also attributed to factors beyond product stagnation, notably Elon Musk’s influence. The original Model S successfully forced the industry toward electrification, achieving its initial goal.
Tesla’s Shift to Volume and Decontenting
Copied to clipboard!
(01:50:16)
  • Key Takeaway: Tesla prioritized volume and lower prices with the Model 3/Y, leading to the decontenting and simplification of the Model S interior.
  • Summary: As EVs became mainstream, Tesla focused on volume, stripping down cars to lower prices, which negatively impacted the luxury feel of the S and X. The current Model S is considered significantly less luxurious than earlier versions due to these cost-cutting measures. This strategy contrasts with the initial high-end positioning of the Model S, which was priced to compete with the Mercedes S-Class.
Tesla’s Recent Financial Decline
Copied to clipboard!
(01:55:23)
  • Key Takeaway: Tesla reported a significant 61% drop in profits year-over-year for Q4 2025 and lost its title as the world’s top-selling EV maker to BYD.
  • Summary: In the quarter ending December 2025, Tesla saw a 3% revenue decrease and a 61% profit decrease compared to Q4 2024. Tesla sold 1.6 million vehicles in 2025, an 8.5% decrease year-over-year, allowing China’s BYD to take the top EV sales spot. Elon Musk stated that long-term, only autonomous vehicles will be made, excluding the next-generation Roadster.
Roadster Announcement Timeline
Copied to clipboard!
(01:55:01)
  • Key Takeaway: The next-generation Tesla Roadster was announced eight years prior in November 2017, and it remains unreleased.
  • Summary: The next-generation Roadster was announced in November 2017, meaning a deposit could have been placed eight years ago for a car that still does not exist. Tesla showed a moving prototype at the announcement, but it was clearly not the final production model. This vehicle is expected to be an expensive luxury car if it ever ships.
Tesla FSD vs. Waymo Safety Data
Copied to clipboard!
(01:56:52)
  • Key Takeaway: Tesla’s robo-taxis crash three times more often than human drivers, while Waymo maintains a safety record significantly better than human averages.
  • Summary: Tesla’s autonomous vehicles crash approximately once every 55,000 miles, compared to one crash every 500,000 miles for human drivers, even with a safety monitor present. Waymo’s data shows 91% fewer serious injury crashes compared to human drivers on the same roads. Waymo’s cautious, slow rollout contrasts with Tesla’s approach of overpromising and hiding accident details.