Microsoft Gives Up On The Copilot Key - WAN Show May 22, 2026

Episode Summary

Executive Summary: The episode centered on big tech and platform shifts: Microsoft softening Copilot pressure, Google’s sprawling AI keynote, Windows driver fixes, Plex’s steep price hike pushing Jellyfin adoption, security embarrassments at CISA, and broader debates about AI, jobs, and software quality. The hosts mixed skepticism and optimism, especially favoring user control, local hosting, and companies that actually improve core functionality.

Main Topics: Microsoft backs off Copilot enforcement (Priority: 5/5): Microsoft will let Windows 11 users remap the Copilot key, and the hosts framed this as a long-overdue concession to user control and customizability after years of forcing AI features into workflows. Google I/O and the AI saturation problem (Priority: 5/5): The hosts dissected Google’s keynote, covering Gemini 3.5, video generation, AI agents, and Workspace integration, while questioning whether the demos are truly practical or just another wave of AI overreach. Windows driver quality and battery-life bugs (Priority: 4/5): Microsoft admitted its old driver testing was too narrow and let bad third-party drivers harm battery life, sleep, and thermal behavior. The hosts welcomed the fix as a major quality-of-life win for Windows users. Plex price hikes and the rise of Jellyfin (Priority: 5/5): Plex Pass lifetime pricing rose sharply, prompting discussion of better alternatives like Jellyfin, Emby, and Kodi, plus frustration that Plex has drifted away from its core media-server users. Security and government incompetence (Priority: 4/5): A public GitHub repo tied to a CISA contractor exposed passwords, keys, and tokens for months, underscoring the hosts’ concern about weak operational security, staffing cuts, and human error. AI, labor, and the changing power balance (Priority: 4/5): The hosts debated whether AI threatens junior jobs, whether developers are being replaced by ‘idea people,’ and who should bear moral responsibility when working on AI or other destructive-but-lucrative tech. Local LLMs and practical self-hosting (Priority: 4/5): They explored what people actually use locally hosted LLMs for: home automation, document search, media tooling, private companionship, task automation, and other privacy-sensitive or high-volume workloads.

Key Arguments: Users want agency; AI features are more acceptable when they are optional and rebindable, not shoved into every workflow. Google’s AI demos look impressive but often rely on idealized conditions; practicality and trustworthiness remain the big barrier. Microsoft’s driver-testing changes matter because ‘stable but secretly bad’ drivers can still destroy battery life and sleep behavior. Plex is losing goodwill because it keeps worsening core media playback while pushing adjacent services users don’t care about. Local LLMs are valuable not because they beat cloud models at everything, but because they can handle private, repetitive, or expensive tasks without sending data to a third party. The hosts think a lot of AI hype is powered by capital pressure and hype cycles, not by current user demand or product maturity. People doing very well-compensated work on destructive or morally questionable systems still have personal moral responsibility, even if survival pressures can change the line for others.

Data Points: Copilot key remapping: Coming in a future Windows 11 update - Microsoft is allowing users to remap the Copilot key instead of forcing its AI function Gemini 3.5 Flash availability: Available now - Google’s new model family began rolling out with Flash Gemini 3.5 Pro launch: Next month - The higher-end model was announced as forthcoming Google I/O interaction ratio: About 0.05% thumbs-up/view ratio - The hosts cited extremely low engagement relative to views Google I/O views: 6.8 million - They discussed the keynote’s unusually high view count Google AI pricing: AI Pro plan reduced from $250/month to $100/month - Google lowered the price while tightening free usage caps Free AI usage limits: Weekly hard caps plus a smaller rolling five-hour cap - Google shifted free usage from prompt-count based to compute-based limits Anthropic revenue/profitability: Projecting first profitable quarter this quarter - Discussed as a possible first profitable pure-AI company Anthropic to xAI/SpaceX capacity deal: $1.25 billion per month for three years - Renting data-center capacity was used to illustrate the AI infrastructure boom Windows driver testing change: Expanded beyond crash checks to include power/thermal efficiency - Microsoft is fixing hidden battery drain and sleep issues caused by bad drivers Plex Pass lifetime price increase: $250 to $750 - The lifetime pass tripled in price Plex price increase timing: July 1 - The new lifetime pricing date was announced ahead of time CISA leak duration: Roughly 6 months - The contractor’s public repository remained exposed for months CISA leak size: 844 MB - The repo contained production infrastructure material and credentials CISA repo scope: Dozens of systems and multiple key types - Included AWS, Azure, Kubernetes, Argo CD, Terraform, GitHub tokens, and SAML certs CISA staff loss: Nearly one-third - Mentioned alongside the agency’s reduced capacity CISA proposed budget cut: Around $707 million - Used to explain worsening cybersecurity readiness Samsung strike risk: 48,000 workers - Nearly the entire walkout scale if the deal had failed Samsung strike window: May 21 to June 7 - Scheduled strike dates before the tentative agreement Samsung bonus pool: 12% of operating profit - Tentative agreement for semiconductor division bonuses Average Samsung bonus: $338,000 per chip employee - Company estimate under the tentative deal Plex alternatives: Jellyfin, Emby, Kodi - Discussed as replacements or complements to Plex FTC deepfake rule enforcement: 48 hours - Platforms must remove flagged content within 48 hours FTC deepfake civil penalty: Approximately $53,000 per violation - Warning letters were sent to major tech firms OpenAI/Musk lawsuit damages sought: Up to $134 billion - Musk’s claim in the OpenAI case Samsung chip strike estimated impact: $14–20.8 billion - Potential loss from an 18-day strike Floatplane T-shirt: Exclusive drop for members - Merch announcement during the show Cables inventory: 57,000 units - LTT Store expected cable stock in 6–12 weeks

Pivotal Quotes: "Is it starting to feel like Microsoft is getting the message that users do not want to have this stuff crammed down their throats?" — Linus: Discussing Microsoft allowing Copilot key remapping "The thing you showed us, this isn't what we want." — Luke: On community backlash to Nvidia’s AI-heavy game demo and concerns about AI-generated slop "If you actually want to add value, read the documentation, create a patch too, and add some real value on top of what the AI did." — Linus Torvalds: Referenced while discussing AI-generated security bug reports overwhelming Linux mailing lists

Implications: Expect more pressure on tech companies to respect user choice, improve core reliability, and price services transparently. AI adoption is not rejected outright, but trust, privacy, and utility will determine which tools survive.