The WAN Show — Microsoft Finally Admits AI Sucks (Feb 6, 2026)

Podcast: The WAN Show (Linus Tech Tips)
Hosts: Linus Sebastian, Luke Lafreniere
Date: February 7, 2026

Main Theme & Purpose

This WAN Show episode delivers a no-holds-barred discussion of Microsoft’s tumultuous relationship with AI in Windows, its impact on user trust, and the company’s course correction following a slew of unpopular features and buggy releases. Linus and Luke also take the opportunity to lampoon tech industry trends—dark patterns, unreliable software, privacy issues, and the broader implications of unchecked AI and monopolistic practices. The show, consistent with WAN’s style, veers into rants, personal stories, and playful tangents about Linux, gaming, loot boxes, cable myths, and more.

Key Discussion Points & Insights

1. Microsoft’s AI Rollback & Trust Problem

[03:11–10:23]

• Microsoft, acknowledging years of negative user feedback and high-profile Windows 11 debacles, is pulling back on aggressive AI features (like Copilot and AI buttons everywhere).
• Pavan Davuluri (Windows President) has admitted the need for user-meaningful improvements:
  Linus [03:11]: “Microsoft is trying to earn back user trust by dialing back AI features no one wanted. That’s right. They’re finally acknowledging what users have been screaming about for years.”
• A series of buggy updates this January (shutdown bugs, OneDrive/Dropbox crashes, boot failures) worsened user perception.
• Tom Warren (The Verge) noted a disappearance of longtime Windows fans and declining consumer revenue for Microsoft's Windows/Xbox division.
• Linus & Luke reflect on the internal disconnect at Microsoft and empathize (somewhat) with the scale of making changes at a massive organization.
• Luke [07:42]: “I’m going to try to have some faith that they’re going to try to do some good.”

2. Dark Patterns and Bloatware in Windows

[10:23–14:56]

• Hosts rail against invasive software: login requirements for MS Paint and Clipchamp, OneDrive being forced in setup and saves, browser hijacks in Teams, and non-standard UX.
• Linus’ viral tweet about the “Sign in to Paint” button struck a nerve. Linus [11:39]: “Microsoft, if at any point you’re looking back going, when did we go wrong? This was it. And that seemed to really resonate with people.”
• Discussion of “dark patterns” and how difficult it has become to set up a new Windows PC without constantly being nudged into Microsoft’s ecosystem.

3. Monopoly Behavior & Antitrust

[14:56–16:59]

• Linus questions why Microsoft (and others) haven’t been reined in for modern monopoly practices, referencing both the 90s antitrust case and today’s worse behaviors.
• Brief, comedic digression on Lina Khan (FTC Chair) and regulatory hopes.

4. Windows Recall Feature Backlash & Internal Reckoning

[17:08–19:05]

• Linus reports on the internal failure of “Recall”—Windows’ privacy-intrusive screenshot aggregator—following social media outcry.
• Management got "roasted" on social media for boasting about making Windows an "agentic OS". Linus [17:34]: “[Davaluri] tweeted that Windows would become an agentic OS. You sure you still like this guy? And got absolutely roasted.”

5. Linux & Alternative OS Ecosystem Gaining Steam

[10:23–28:33]

• Linux gaming and alternative OS (ChromeOS, macOS, iPadOS) are improving. Linus outlines a new “Linux Challenge” involving multiple machines.
• Discussion on how alternative platforms are now legitimate threats thanks to Microsoft’s blunders and the rise of tech-competent kids using Chromebooks and iPads.

6. AI Prompt Injection (Physical World Attacks)

[141:13–143:41]

• Recent research at UC Santa Cruz and Johns Hopkins demonstrates prompt injection attacks against real-world AI/robotic systems using physical signs (e.g., causing vehicles/drones to ignore pedestrians or land in unsafe places).
• The ease and cross-language efficacy of these attacks is called “terrifying”—it’s the classic prompt injection exploit, but with real-world risk.

7. Grok AI & AI-generated Harmful Content

[33:15–40:59]

• Elon Musk’s X/Twitter faces raids and legal trouble in France (child sexual abuse material), and Grok AI is under investigation in the UK for unconsensual image generation.
• Linus and Luke debate the difficulty of regulating AI and the urgency of proactive takedown when harms are revealed (referencing the infamous Microsoft Tay bot as a positive example of shutting down a bad AI fast). Luke [36:05]: “Having some form of error or unintended functionality is going to happen... but not immediately disavowing it and turning it off while you fix it... that's what tells us the intent.”

8. Supply Chain Software Attacks: Notepad++ Hack

[85:53–89:59]

• Notepad++'s update server was compromised for six months by state-sponsored attackers, who targeted infrastructure and government organizations.
• The incident raises alarms about the lax security of open-source update mechanisms and broader blind trust in software updates.

9. Gambling, Loot Boxes & Regulation

[120:30–127:00]

• Austrian Supreme Court rules FIFA Ultimate Team loot boxes are not gambling, claiming player skill is involved—Linus, Luke, and Dan strongly disagree, equating them with slot machines and other pure games of chance.
• Extended debate on why government regulation lags behind tech, and their frustration that gambling has been normalized in games and sports.

10. Cables, Audiophile Myths, and Product Choices

[144:00–147:41]

• Discussion of a blind test showing people can't distinguish audio signals run through wire, banana, or mud—debunking the high-end “audiophile cables” myth.
• Dan [145:25]: “The cables that I make here… They are good forever. They’re not because they sound better.”

11. Company Transparency: LMG Salary & Investment Data

[189:03–204:31]

• Linus details LMG’s actual spending on employee wages and growth over the years, presenting charts to counter social media misinformation.
• The hosts discuss why full financial transparency isn’t possible and the challenges of reporting (and trusting) failure rates for hardware—with anecdotes about how companies, incentives, and even trusted third parties can sometimes skew or misrepresent stats.

12. Listener Q&A Highlights

[58:55+ interspersed]

• Honest advice about screen time for kids (“never go Cocomelon”), retro kids' shows, and challenges of keeping tech misconceptions straight (e.g., intranet vs. internet speeds, cable nomenclature, tech support stories).
• Life updates, workflow rants, and musings about balancing work and family.

Notable Quotes & Memorable Moments

• Linus [03:11]:
  “Microsoft is trying to earn back user trust by dialing back AI features no one wanted… Why it took this long, we don’t know.”
• Luke [07:42]:
  “I’m going to try to have some faith that they’re going to try to do some good.”
• Linus [11:39]:
  “Microsoft, if at any point you’re looking back going, when did we go wrong? This was it.”
• Linus [17:42]:
  “[Davaluri] tweeted that Windows would become an agentic OS. You sure you still like this guy? And got absolutely roasted.”
• Luke [36:05]:
  “Having some form of error or unintended functionality is going to happen... but not immediately disavowing it and turning it off while you fix it... that's what tells us the intent.”
• Dan [145:25]:
  “The cables that I make here… They are good forever. They’re not because they sound better.”
• Linus [189:03]:
  “Sunlight is the best disinfectant. So I actually asked accounting to put together some additional numbers... LMG continues, has and continues to invest in people.”

Comedic/Meta Moments:

• Linus’ “Death Star” Costco tag explanation and tech support stories.
• The “Calm Gobbler” checkout message, and ensuing laughter [223:03+].
• Bird anatomy lesson during T-shirt design debate [136:25+].
• “Every fart’s a gamble.”
• Braces storylines:
  Luke [237:04]: “Do you guys… They're yellow, bro. Get a different color.”

Timestamps for Major Segments

• 03:11 – Microsoft Admits AI Mistakes, User Backlash
• 10:23 – Windows Bloatware, Dark Patterns, Setup Experiments
• 17:08 – Internal Failures, Recall, “Agentic OS” Backlash
• 20:48 – Windows, Linux, and the User Exodus
• 85:53 – Notepad++ Hack, Software Update Trust
• 120:30 – Austrian Court on Loot Boxes, Gambling Discussion
• 141:13 – Prompt Injection in Real-World AI
• 144:00 – Audio Cables and Audiophile Snake Oil
• 189:03 – Salary Transparency, LMG Spending
• 223:03 – Listener: 'Calm Gobbler' Checkout Message
• 236:52 – Linus Braces, Orthodontic Tangents

Tone & Style

Conversational, irreverent, and unfiltered—Linus and Luke roast, rant, and relate on all things tech, with snark, exasperation, and the occcasional earnestness about the tech industry’s trajectory.
Frequent diversions and inside jokes keep things lively, but the tech analysis remains sharp and, at its best, insightful for listeners and viewers who want both entertainment and critique.

Final Word

This episode exemplifies WAN Show’s appeal—a blend of topical tech news, irreverent commentary, and honest industry perspectives. Whether it’s Windows' AI follies, the future of open software, or the eternal comfort of banana-powered audio, Linus and Luke’s chemistry and candor make for an engaging, memorable listen—especially for anyone frustrated by the state of modern computing.