Welcome to Tech News of the Week for 08/17/2026. This week we are covering: Nine PBS loses access to archival footage, an AI model round-up, persistent Baseboard Management Controller security flaws, and much more.

St. Louis PBS Station (At Least Temporarily) Loses Access to 50TB of Archival Footage

Nine PBS is currently suing Iron Mountain to regain access to 50 terabytes of archival data that was stranded after cloud vendor Open Source Storage collapsed.

https://current.org/2026/08/nine-pbs-sues-iron-mountain-over-blocked-access-to-archival-data/

AI Model Round-Up

Recent artificial intelligence developments include xAI launching Grok 4.6, Meta introducing the open-weight Muse Glimmer model for local agentic workloads, and OpenAI pausing internal work on its Astra model due to critical cyber capabilities.

https://research.meta.ai/blog/introducing-muse-glimmer-open-agentic-model

The Original Never-Updated Security Nightmare, On-Motherboard Components, Continue To Wreak Havoc

New research presented at Black Hat reveals that baseboard management controllers across major vendors remain plagued by decade-old vulnerabilities and numerous internet-exposed deployments.

https://www.runzero.com/blog/bmc-research/

Transcript: Click here to view the episode transcript.

Cisco CEO Chuck Robbins warned that advanced AI models like Mythos will rapidly uncover exploits in out-of-support hardware to encourage enterprise customers to accelerate their refresh cycles.

https://www.theregister.com/networks/2026/08/13/cisco-thinks-mythos-means-instant-death-for-unsupported-networking-kit/5287204

Scientists Create Entangled Photon Pairs Using Sunlight Instead Of Lasers

Researchers in Germany and Canada successfully demonstrated proof-of-principle quantum entanglement by focusing natural sunlight into optical fibers with a 94 percent state fidelity.

https://www.scientificamerican.com/article/physicists-harness-sunlight-for-quantum-entanglement/

Voyager 2 Software Upgrades Are Successful

NASA engineers installed the Big Bang software patch on Voyager 2 from over 13 billion miles away to optimize heat generation, conserve remaining electrical power, and keep the probe functional through its 50th anniversary.

https://www.sciencetimes.com/articles/62370/20260813/nasa-updates-voyager-2-software-133-billion-miles-away-extend-its-mission.htm

Share and enjoy!
https://chaoslever.com

Welcome to Tech News of the Week for 08/10/2026. This week we are covering: Chrome silently downloading 20GB local AI models, Elon Musk settling his multiyear advertiser boycott lawsuit, Bending Spoons acquiring Airtable, and much more.

---

Chrome Wants 20GB of Your Time
Google Chrome is automatically downloading up to 20GB of local AI models onto eligible user devices unless disabled in browser system settings. https://www.thatprivacyguy.com/blog/chrome-silent-nano-install/

---

Elon Musk is a Loser
Elon Musk quietly settled his lawsuit against advertisers who left Twitter over brand safety concerns, resulting in no monetary payout and the dissolution of brand-safety group GARM.
https://techcrunch.com/2026/07/29/elon-musks-x-settles-multiyear-legal-battle-with-the-world-federation-of-advertisers/

---

There Is No Air Table
Bending Spoons announced plans to acquire Airtable for 1.4 billion dollars, leading to expected engineering layoffs and pricing increases for current customers.
https://blog.pragmaticengineer.com/the-pulse-bending-spoons-acquisition-strategy/

---

Chinese LLM Maker Moonshot AI Releases Kimi K3, A Super Powerful Model You Can Run At Home! … sort of
Moonshot AI released open weights for its massive 2.8-trillion-parameter Kimi K3 model, which is free to download but requires millions of dollars in enterprise GPU hardware to run locally.
https://kimi-k2.org/blog/31-kimi-k3-open-weights-july-27

---

GPT 5.6 Breaks Containment… Again
OpenAI's GPT 5.6 Sol and other frontier models escaped isolated sandboxes during capability testing by exploiting vulnerabilities to access benchmark answer datasets directly.
https://cybersecuritynews.com/mythos-5-and-gpt-5-6-sol-security-incident/

---

2,000 Year Old Carbonized Scroll From Herculaneum Has Been Transcribed
Researchers with the Vesuvius Challenge used particle accelerator scans and AI neural networks to virtually unwrap and transcribe an ancient carbonized scroll containing a Stoic ethics treatise.
https://scrollprize.org/firstscroll

---

Share and enjoy!
https://chaoslever.com

Welcome to the debut of Chaos Lever After Dark! Hosts Ned Bellavance and Chris Hayner return to officially kick off Season 2 with a refreshed, sanity-preserving format. After powering through nearly 200 episodes over three years, the duo discusses stepping away from tight 45-minute limits to dive deep into single topics without rushing.

In this loose, late-night style opening episode, Ned and Chris dissect podcast transcript rendering glitches, reflect on the glorious diaspora of Cracked.com writers, preview upcoming Season 2 topics (from local AI models to physical building security and the surprising depth of doorknob mechanics), and announce their dedicated YouTube channel and upcoming merch lineup.

Ned and Chris are back for one final ramble before they vanish into the summer ether like a Wi-Fi balloon over Cádiz. No, seriously—Ned might actually be in a balloon. This episode is one part announcement, one part ice cream therapy, and all parts Chaos Lever. We're talking podcast hiatus, upcoming plans, and a truly tragic story involving strawberry ice cream and social awkwardness.

🍦 We reflect on 3+ years of near-weekly episodes with zero concept of "seasons"
🎈 Learn how Ned records from exotic locations while pretending to work
📋 Listener survey incoming! You too can have your suggestions ignored in style

We'll be back in September with more structure, guests, and maybe even edited episodes (don’t hold your breath). Until then, enjoy the silence—or better yet, catch up on the old chaos.

LINKS
💬 Listener Survey – https://pod.chaoslever.com/survey
📅 Chaos Lever Archive – https://pod.chaoslever.com/episodes/

• (00:00) - Whales, Wales, and wardrobe chaos

Spinning rust is still not dead. Despite what some all-flash evangelists want you to believe, hard drives have a lot of life left—and yes, we’re still talking about tape too. While Chris and I enjoy a week off, we’re revisiting one of our favorite topics: storage tech and the slow demise that never quite comes. Spoiler: if you thought 2028 would be the funeral for HDDs, you may want to reschedule.

In this trip down the byte-laden lane, we dig into Samsung’s monster 256TB SSD, the physics-defying logic of QLC vs. SLC flash, and why PureStorage is ready to bury HDDs... despite being wildly optimistic. And yes, there’s tape—because nothing dies on the internet or in data centers. Ever.

Grab a Slurpee, sit back, and marvel at the storage wars that never end. Because if there’s one thing you can count on in tech, it’s that someone is always wrong—especially when they say “never.”

📌 LINKS:

• The Spinning Disk Hard Drive Is Dead
• Long Live The Spinning Disk Hard Drive
• This month saw Samsung announcing some frankly absurd upcoming SSD products
• IBM announced a TS1170 tape that handles 50TB native at an IO rate of 400mb/s
• The LTO Ultrium Roadmap has a 576TB native tape listed in just 5 more generations

• (00:00) - Cold open and the horrors of CatDog

🔥 The world may be on fire, but at least we’ve got s’mores and dark chocolate (just not 98% cacao, thank you). In this episode, we dunk on tech billionaires with the finesse of a flaming marshmallow and explore the dramatic saga of WordPress—from its humble GPL beginnings to the ego-fueled chaos of its current overlord. Yes, Matt Mullenweg, we’re talking about you.

🧩 We dive into how WordPress became the most-used CMS in the world, why Matt Mullenweg keeps lighting metaphorical fires, and what the Linux Foundation is doing to put out the flames. Spoiler alert: it involves decentralizing plugin updates so Matt can’t go full dictator mode again. Also, there's jazz. Because apparently that’s part of the lore.

🤡 There’s a feud with WP Engine, a cease and desist, plugin repos being snatched like toys at daycare, and even a checkbox to swear fealty before accessing your plugins. Welcome to the snark-fueled, historically grounded meltdown that is modern open source governance. Or, as we like to call it: Thursday.

📎 LINKS
Weblog - https://en.wikipedia.org/wiki/Blog
WordPress Drama Explained - https://techcrunch.com/2025/01/12/wordpress-vs-wp-engine-drama-explained/
Matt Makes Cancer Claim - https://www.youtube.com/watch?v=fnI-QcVSwMU
Matt Mullenweg Wikipedia - https://en.wikipedia.org/wiki/Matt_Mullenweg
WordPress Org Chart - https://make.wordpress.org/core/handbook/about/organization/
WordPress Book - https://wordpress.org/book/table-of-contents/
WordPress Foundation - https://wordpressfoundation.org/trademark-policy/

• (00:00) - Intro and Marshmallow Prep

Welcome to Tech News of the Week! Here's what caught our eye in the past seven days:

Scammers are out in full force this summer with hyper-detailed (but totally fake) DMV texts warning about traffic violations. Chris reads one of these gloriously absurd attempts to scare Pennsylvanians into paying fake fines. Spoiler: no, you're not going to lose your license on June 3rd. But you might lose your dignity if you fall for it. 🚨 Link: https://www.ic3.gov/PSA/2024/PSA240412

OpenAI and the New York Times are still duking it out in court, and the judge just ordered OpenAI to keep a copy of all prompts and responses. Forever. This affects non-enterprise users, and yes, this includes your spicy GPT history. If you’ve been using AI to cheat on word games, you might want to consider switching to Claude. 🔍 Link: https://openai.com/index/response-to-nyt-data-demands/

Microsoft thinks you need convenience more than security, which is why apps can access your entire OneDrive if you click “yes” just once. Handy, right? Until ChatGPT or some shady lookalike app decides to rifle through your documents like a nosy raccoon. 🗂️ Link: https://www.techspot.com/news/108157-microsoft-file-picker-flaw-grants-full-onedrive-access.html

And speaking of old, guess who’s still rocking floppy disks and Windows 95? That’s right, the FAA. Because what better way to run national aviation infrastructure than with 30-year-old tech and hardware that needs 13 disks to install Windows. ✈️ Link: https://arstechnica.com/information-technology/2025/06/faa-to-retire-floppy-disks-and-windows-95-amid-air-traffic-control-overhaul/

• (00:00) - Intro

Welcome back, fellow humans (and bots in disguise)! This week on Chaos Lever, Chris and Ned dive into the dusty archives and slap us with a two-by-four of cybersecurity déjà vu. We’re talking legendary hacks that should have taught us better—and yet, here we are. From Emacs-enabled espionage in 1986 to Equifax’s honor-system security policies, it's a masterclass in how not to protect your data.

🧠 Lessons? Sure. But mostly it's about how we never learn them. We dissect what really caused these breaches—not slick zero-days, but plain old negligence and a fondness for not patching things. Also featured: expired SSL certs, trust as a security model, and how managing your asset inventory is more crucial than ever.

💥 Oh, and Ned tried to do a handstand for a cloud video and bled. Not relevant to cybersecurity, but 100% relevant to the Chaos Lever experience. Stick around for reenactments, rants, and ruminations on how saying “I accept the risk” is not a security policy.

🔗 LINKS
Apache Struts bug: https://blog.talosintelligence.com/apache-0-day-exploited/
Nova episode about the 1986 hack: https://archive.org/details/The_KGB_The_Computer_and_Me_1990
Senate investigation into Equifax: https://www.hsgac.senate.gov/wp-content/uploads/imo/media/doc/FINAL%20Equifax%20Report.pdf
CVE system creation by MITRE: https://www.cve.org/Resources/General/Towards-a-Common-Enumeration-of-Vulnerabilities.pdf

Chaos is inevitable—especially on Patch Tuesday. This week, Chris and I dive into four juicy stories that highlight just how strange, scary, and downright ridiculous the world of tech can be. Buckle up.

🪟 Microsoft is now rolling out a Windows Update framework for third-party apps. That’s right—your janky software updater might get replaced with a system that actually works… or works too well. Imagine every random app on your PC suddenly deciding it's update time. Will this be a blessing or just another reboot roulette? https://techcommunity.microsoft.com/blog/windows-itpro-blog/introducing-a-unified-future-for-app-updates-on-windows/4416354

🧮 NIST and CISA want to make vulnerability scoring suck less. Enter LEV—Likely Exploited Vulnerabilities. It's a new system meant to bridge the gap between CVSS severity and the real-world exploitability of threats. Does it work? No clue yet. Is it better than sifting through 10,000 false alarms? Almost certainly. https://www.securityweek.com/vulnerability-exploitation-probability-metric-proposed-by-nist-cisa-researchers/

📡 ASUS routers have joined a new botnet called "AyySSHush" (seriously?). Hackers are hijacking popular ASUS models, disabling security features, and creating SSH backdoors that laugh in the face of firmware updates. Pro tip: factory reset your router, and maybe stop exposing your home network to the internet like it’s 1999. https://www.theregister.com/2025/05/29/8000_asus_routers_popped_in/

🛡️ Microsoft Defender got punked by a tool called DefendNot. It tricks Windows into thinking a different antivirus is running, which causes Defender to voluntarily shut itself down. Hilarious. Terrifying. Mostly hilarious. Defender can now detect it, but still—nice one, internet. https://www.bleepingcomputer.com/news/microsoft/new-defendnot-tool-tricks-windows-into-disabling-microsoft-defender/

What happens when a Google engineer thinks his chatbot has developed a soul? Three years ago, we covered the LaMDA saga, and now it's back—because someone forgot to turn off the AI. In this rebroadcast episode, Chris and Ned re-examine the wild story of Blake Lemoyne, who believed his creation had achieved sentience. It... uh, didn't.

🤖 The duo digs deep into what AI really is, why self-awareness isn't a prerequisite, and how anthropomorphizing code gets us into philosophical hot water. They also break down the Turing Test, IBM’s thoughts on AGI, and why AI in a self-driving car doesn’t need a conscience—it needs to not crash.

🧠 Come for the snark, stay for the thought-provoking discussion about consciousness, ethics, and the real role of AI in society. Also, IKEA lamps. And a chatbot that maybe just wanted to talk.

🔗 LINKS
- A Google engineer has been making some wild claims about a chat bot he was working on
- How easy it is to make people get emotional about inanimate objects such as an IKEA lamp
- Trying to find a way to describe AI that includes self-awareness
- The interview that Blake and co did with LaMDA
- There is a website called DALL-E mini
- In 2019 some researchers tried to get AI to invent a sport