The Byte & Barrel Bulletin, November 2026: Wall Street Wants Receipts, AI Edits Memory, and the War on Ownership Rolls On
Welcome to the November edition of The Byte & Barrel Bulletin. This month, Wall Street stopped applauding and started demanding receipts from the AI boom, a Danish whiskey earned the first unanimous thumbs down in recent memory, and a device designed to operate your keyboard and mouse “like a digital coworker” landed on our holiday gift list, which tells you exactly how weird this shopping season is going to be.
The pattern across these episodes is familiar: convenience keeps arriving faster than accountability, and the people building the future keep forgetting to ask the rest of us whether we want it.
Photo by Gustavo Fring via Pexels.
The November edition opens. Photo by Gustavo Fring via Pexels.
EPISODE 319, When AI Edits Your Memory
An AI summary can do more than get a detail wrong. Research discussed on the show found that misleading AI-generated summaries can introduce details that never happened and distort what people later remember, even when they know the source is AI. The episode described memory accuracy dropping nearly in half in one example. That figure belongs to the study discussed on the show, not a universal rule for every AI summary or every person.
The concern is still worth taking seriously: if someone has already been nudged by a machine’s account of an event, asking that person to verify the machine may not be much of a safeguard. The human “in the loop” can become a rubber stamp that feels like verification.
When a machine becomes the tool, who is liable? Photo by Karen Zhao via Pexels.
A robot used to harass a teenager. Beyond the immediate harm, the legal question is unsettled: when a machine becomes the tool, who is responsible, the person operating it, the company that made it, or someone else? The answer may depend on how the machine was controlled and what the people involved did. “The robot did it” is not an accountability plan.
Supermarket refrigeration is critical infrastructure, too. A cyberattack disrupting refrigeration at European supermarkets showed how everyday systems can carry real-world consequences. Infrastructure is not only power grids and banks; it also includes the systems that help keep food safe on ordinary grocery shelves.
The episode also worked through listener-submitted scams: fake invoices, dubious investment pitches, and a LinkedIn phishing trick that can hide malicious links inside legitimate-looking notifications. Slow down before clicking. Verify unexpected requests through a channel you choose yourself, not the phone number or link in the message.
EPISODE 320, The Halloween Special: Nick Espinosa on the Real Scary Stuff
Cybersecurity expert Nick Espinosa of Security Fanatics joined us for the Halloween episode. The scariest part was not any one breach. It was the structural problem underneath them: more systems connected to more tools, while responsibility and oversight remain stubbornly human.
Three headlines set the table. Carbon X lets researchers ask questions about genetic code in plain English on a local laptop. A lawsuit involving McDonald’s could test the use of AI pricing recommendations across thousands of locations. And modular battlefield drones built with 3D-printed components and off-the-shelf parts show how quickly the reach of AI-enabled technology is widening. None of these stories means every use is harmful. They do mean the consequences can spread well beyond the screen.
The ownership problem doesn’t stop at the device. Espinosa discussed the possibility of self-replicating AI botnets moving across devices their owners may have forgotten they even had. Add platform influence and AI search summaries shaping what voters see, and the election-integrity picture gets murky fast. In the voting-machine discussion, one principle stood out: security testing needs ethical disclosure rules, not secrecy that leaves the people responsible for fixing a vulnerability in the dark.
Espinosa’s blunt takeaway is worth repeating: the biggest threat may be “gradual disempowerment”. Convenience turning into dependence, with humans reduced to the approval button for automated life. We also covered the Capcom breach and its box-office fallout, and the anniversary of the “War of the Worlds” broadcast: misinformation did not start with the internet. It just got faster.
EPISODE 321: Wall Street Wants Its Money
Wall Street wants proof the AI boom pays. Photo via Pexels.
The biggest signal of the month came from the market. After years of technology companies promising that AI would transform everything, investors started asking a tougher question: where’s the money?
Intel, Alphabet, Amazon, and Reddit all saw their stocks stumble amid concerns about AI spending. Episode 321 cited nearly $800 billion expected to be spent on AI projects next year. The exact estimates vary by source and what spending is counted, but the broader point is clear: building the infrastructure is expensive, and investors want evidence that the investment can produce returns.
The skeptical reading is not that AI has failed. It is that a technology can work and still struggle to become a profitable business. Those are different problems, and only one of them has a balance sheet attached.
The consequences reach ordinary buyers, too. Demand for AI infrastructure is putting pressure on memory and hardware costs, while Microsoft is nudging some owners to trade in dusty Xboxes. At theme parks, guests appear more open to robots handling directions and logistics than creating the moments that feel magical. Good instinct. Hold onto it.
Our Top 10 AI Holiday Gifts countdown made the season stranger still: chore-focused robots, stair-climbing vacuums, physical AI control pads, local AI “lunchbox” computers, conversation-summarizing wearables, and smart glasses that raise serious privacy questions. The most unsettling entry was a device designed to watch how you use your computer, then operate your keyboard and mouse like a digital coworker. A gift that watches you work. Wrap it carefully.
Mike the AI Guy closed with Google’s next move: pushing ads into AI answers. If a helpful summary becomes a sponsored maze, treat the answer like an advertisement, not a fact.
Photo courtesy of TechTime Radio.
FROM THE BLOG: Three Recent Stories Worth Your Attention
Our recent blog coverage follows the same theme: “no damage was done” is a claim, not a finding.
- The $6.3 Billion Hack Started With a Job Interview. It Ended With the Company’s Own AI. The attack chain ran through the target’s own AI tools, turning company automation into part of the attack surface.
Practical takeaway: Inventory every AI agent and tool with external access, restrict what each can reach, and treat its credentials like production credentials, because they are. - The Fake Claude Code Extension Was Live for 11 Hours. The Backdoor Is Still on Some Machines. A malicious extension slipped into a trusted developer channel. Removing it is not the same as cleaning up every machine it may have touched.
Practical takeaway: For a compromised developer tool, assume persistence, rotate credentials it could access, and audit machines, not just repositories. - When AI Finds a Flaw, Who’s Responsible for Fixing It Before Someone Exploits It? Vulnerability discovery can be automated; remediation still depends on people, calendars, and competing priorities.
Practical takeaway: Shorten the gap between discovery and repair, and assign ownership before a disclosure forces the issue.
Defense is still running the slow lane. Photo via Pexels.
The common thread is not that every AI system is malicious. It is that automation speeds up both sides of the equation, while defense is still running the slow lane.
THE POUR: The Byte & Barrel Pour
This month’s tastings, as always, with Mark Gregoire and the panel:
Episode 319: Stauning Rye Whiskey, Denmark. Made from locally grown Danish rye and barley, floor-malted at the distillery, double-distilled in a small direct-fired copper pot still, and matured in new American oak. It’s three to four years old, 96 proof, roughly a 60/40 rye-and-barley blend, and about $50.
Verdict: Thumbs down from Nathan, Mike, and Peyton. A rare unanimous rejection.
Episode 320. Halloween pour. Mark’s pick was a non-age-stated Scotch single malt: 92 proof and made from 100% organic Scottish malted barley.
Verdict: Mike gave it a thumbs up. Nathan didn’t love it, but conceded the point. A split decision that leaned friendly.
Episode 321. Advent-calendar showdown. The show compared whiskey advent-calendar bottles, including a five-to-seven-year bourbon at about $60 and New Riff Malted Rye from Newport, Kentucky: a straight rye aged at least six years, priced at $62. The standout was First West small-batch bourbon, roughly $60 to $62.
Verdict: Nathan and Mike both went thumbs up on the First West. Nathan called it the best of the bunch and a candidate to win the whole calendar.
Drink responsibly. No summary, human or AI, can make that decision for you.
LISTEN TO TECHTIME RADIO
Listen Live at 3:00pm Daily on iTalkRadioLive.com.
Or subscribe wherever you listen: Apple Podcasts, Spotify, YouTube, Amazon Music, and iHeartRadio.
Visit TechTimeRadio.com to join the fan list, find recent episodes, and keep up with the blog. If a gadget promises to think for you this holiday season, maybe let it start with something small, like the thermostat.
Keep questioning, keep your permissions narrow, and keep your whiskey glass safely away from the keyboard.
Nathan Mumm
Mike Gorday