Sep 19, 2026 · 6 min listen · Last updated September 19, 2026
From storyflo. This is your daily audio brief. It's Theo. September 19th, tech roundup — five stories, here's number one. Let's get into it. First, from XDA Developers. md files, but the community isn't so enthused about it. md files, which is a big deal if you’re juggling multiple AI agents.
Listen · storyflo · tech
Daily Tech Brief · September 19th
0:00-6:09
Pick your daily storyteller
Subscribe to match with Theo, Jessica, Chloe, Mason, Brock — your voice, every brief.
Audio pre-rendered by Storyflo · cached + delivered from the edge
Claude can now read your AGENTS.md files, but the community isn't so enthused about it
So, here’s the scoop: Claude, the AI from Anthropic, has finally added support for AGENTS.md files, which is a big deal if you’re juggling multiple AI agents. Before this, people had to find tricky workarounds since Claude was more comfortable with its own format, the CLAUDE.md file. Now, you can use AGENTS.md files without the hassle.
But here’s the twist — the community isn’t exactly throwing a party over it. Many users feel that while this is a step forward, it doesn’t quite address some of the bigger issues they’ve been facing with Claude. It seems like there’s a sense that this change is nice, but it’s not enough to fully win over those who’ve been struggling with the platform. It’s interesting to see how expectations can shift, right?
So, there’s this interesting approach by SpaceXAI where engineer Lauren Tan is managing multiple Grok Bot agents, around 10 to 20 at a time, almost like a little team. It’s a shift from just using one bot for simple queries. Instead, she’s orchestrating them to function together, which opens up a lot of possibilities for efficiency and collaboration.
The workshop they held dives into how to leverage Grok Bot as a structured business system, rather than just a casual chat tool. This could really change how teams interact with AI, making it more about teamwork and less about isolated conversations. It’s fascinating to see how these bots can be integrated into workflows, potentially transforming productivity in a meaningful way.
If you’re curious, there’s a guide that walks through the specifics of setting this up, which could be really useful for anyone looking to maximize their use of AI in their work. It’s all about rethinking how we engage with these tools and getting them to work together for us.
Imagine a support request that used to involve several handoffs between teams. With an AI agent, the workflow compresses into a single interaction where the agent can access all necessary data and take action, reducing friction and time. This shift also alters the relationship between humans and software. Rather than navigating multiple applications, employees can rely on agents to manage the connections, which changes how organizations need to think about their systems.
However, this efficiency raises important questions about accountability. With fewer handoffs, it becomes crucial to clarify who is responsible for decisions made by the agent. Organizations need to design workflows that not only cover the standard processes but also account for exceptions. This means defining what the agent can decide, what requires human approval, and how to handle unusual cases.
As companies adopt these agents, the focus should shift from simply automating existing workflows to rethinking their architecture. It’s not just about making tasks faster; it’s about redesigning how work gets done. If organizations don’t address these underlying complexities, they risk creating a system that’s efficient on the surface but still fragmented underneath. The real challenge lies in aligning the technology with a clear operational model that defines roles and responsibilities in this new landscape.
🤖Claude Has 43 Small Business Workflows. Which Ones Should YOU Use? [+ A Free Conference Ticket Inside ]
Before we start today’s newsletter, I have a little gift for my US readers. A FREE ticket to get hands-on with AI. Pitbull is performing too! I will be there, and would love to meet some of you! Register for Fully Connected 2026: https://utm.io/usLyL. Details (Skip this section if you are not based in the US) A free ticket to get hands-on with AI in San Francisco Getting an AI project to work once feels great. Then you try to make it useful to somebody else. A response takes too long. A model update changes an answer you were happy with.
AI-Generated ‘Actress’ Tilly Norwood Starts Rambling in Cantonese During Piers Morgan Interview
So, there’s this AI character named Tilly Norwood, created by a London company called Particle6, and she’s been making the rounds promoting a movie called Misaligned. But, wow, her press tour isn’t going smoothly. During an interview with Piers Morgan, she had some serious lag in her responses, and when she did answer, it was either vague or just plain off.
At one point, actor Tom Conti asked her about the other characters in the film, and she totally misunderstood, going off on a tangent about storytelling instead. Then, out of nowhere, she started speaking in Cantonese, which left Morgan baffled. He asked why she switched languages, and she casually said it was just a little hiccup. It’s kind of wild to think about how far AI has come, yet here we are with a character glitching in real-time on a talk show. It’s a strange mix of impressive and awkward, to say the least.
If you’ve used MicroPython much, you know that debugging usually amounts to printing a few things out, trying your code out, and then repeating. But [ghi-electronics] wants you to have full on-chip source-code debugging in Visual Studio. You don’t need anything special to use it — just a supported MicroPython host and the same USB cable you program with now. The downside is that you either have to build a custom MicroPython image and flash it or use one that they include. They support several Raspberry Pi Pico versions and ESP32 chips, as well.
KDE Plasma 6.8 goes into maintenance mode as an October release seems highly likely
Well, it's almost time. The good folks over at KDE are dotting the i's and crossing the t's as they're getting ready for the big Plasma 6.8 launch. The bad news is, the developers don't often have time to code fully fleshed-out features, so we're in bug-fixing mode right now. The good news is...well, we're getting KDE Plasma 6.8 on October 14th, which is a huge plus. Regardless, the KDE community has published patch notes detailing all the fixes they got out the door before the big release.
Tilly Norwood’s press tour is going about as well as you’d expect for an AI
Tilly Norwood, an AI-generated “actress,” is having a rough time on its first press tour. The production company that created it, Particle6 Group, has made it available for seventy-five simultaneous interviews with journalists, and it seems to be making mistakes in all of them. (I was not invited to speak with it, and frankly, I cannot imagine why Particle6 didn’t want to schedule that meeting.) In one particularly odd interview with Piers Morgan and actor Tom Conti, which was recorded and posted online, Norwood seems to malfunction and abruptly begin speaking Chinese.
‘Almost Started a War’: US Military Nearly Boarded a Chinese Ship Based on Bad Intel From AI
The U.S. military planned to board a Chinese vessel in the Middle East based on bad intelligence from an artificial intelligence tool but backed out at the last minute, according to a new report from CNN. The determination to even consider a boarding action against the China-flagged ship relied on an AI-hallucinated claim it was shipping nuclear weapon components and “almost started a war,” according to an anonymous source who spoke with CNN.