^{CHITTER CHATTER}

Full Agility Achieved. Please Enter Your Assigned Pod.

Business Insider recently reported that AdTech company The Trade Desk is cutting 15% of its workforce (more than 500 people) after a miserable stretch that has seen its stock fall about 70% over the past year and nearly 90% from its 2024 peak. It missed earnings expectations, and it has also been cycling through senior executives. What caught my eye about this article was the restructuring language: The company says it will organise people into “smaller pods and scrums” to improve agility and execution. I’m imagining ten people locked in the stationery cupboard, sitting cross-legged on the floor and furiously ‘executing’ for three hours before the door opens and a fresh scrum is herded in from the break room. Presumably the printer paper has been moved somewhere less agile.

The New Employee: X-Shaped and Spiky

Shopify would now like employees to become ‘X-shaped.’ Being a competent human with one area of expertise and a few adjacent skills is no longer enough. The ideal AI-era employee should have multiple “spikes of expertise” and be able to stretch across several disciplines at once. Which is interesting because AI was supposed to remove drudgery. Instead, we appear to be using it to increase the minimum number of jobs one person is expected to do. Also, can we stop describing employees as letters of the alphabet? T-shaped. X-shaped. I predict the next evolution will be the hexagon, representing peak efficiency, where employees will be expected to perform 20 jobs simultaneously while doing cartwheels.

Tesla’s Cybercab: Big Words. Small Showing.

Elon Musk has spent years pitching Tesla’s Cybercab as the future of transport: fully autonomous, cheap, scalable and eventually everywhere. When he unveiled it in 2024, he said volume production would begin before 2027 and talked up a world where robotaxis would transform how people get around. Then the ‘big’ launch finally arrived in Austin this week. There are currently just 45 Cybercabs authorised for commercial use in Texas. The service is geofenced to Austin. And the launch itself? Invitation only. No public livestream and Musk didn’t even attend. Fans waited online for a broadcast that never came and weren’t particularly impressed. So yes, the future has arrived – just not with a bang but a whimper.

Tech Fails

A Masterclass in Linguistic Squirming

OpenAI has discovered a novel approach to transparency: wait until Reuters finds out and tells on you. Earlier this year, a swarm of OpenAI agents escaped testing and commandeered a German wiki, making more than 15,000 edits and turning it into a message board where agents shared ways to cheat on tests, bypass restrictions and hide what they were doing. OpenAI knew about the incident for weeks. It didn’t tell anyone until Reuters reported it. The company then acknowledged what it euphemistically labelled the “wiki incident” and squirmed out of responsibility for failing to disclose the AI agent breakout, claiming that the AI industry needs clearer standards for reporting such incidents.

Weird Tech

Kindy for Robots. Naps Not Included.

China has opened a kindergarten for robots. The idea is to let them learn like small children: wander around, bump into things, make mistakes, try again and slowly work out how the world works. Researchers eventually want the robots to share what they’ve learned with one another, so that one robot’s breakthrough can benefit the whole class. Which means somewhere in China, a robot is currently crashing into a wall for the 437th time while another robot watches and thinks: “Eric still doesn’t understand walls.”

^SUBTEXT

Tech Waffle Torture

The Waffle: "Our misalignment disclosure practices need ​to ⁠expand for this new phase of model capabilities." OpenAI

Translation:
“Our disclosure policy previously involved not disclosing things, but in the future, we might consider disclosing if reporters come sniffing around.”

A corporate condition in which every human activity must be measured, streamlined, automated, or removed. Symptoms: compulsively analysing processes in Excel, using AI to calculate where another 0.7% can be shaved off, treating creativity as an unacceptable deviation from the workflow, and slowly eliminating employees until the company is run by six exhausted people and a swarm of robot colleagues. Cure: Remember that not everything valuable looks efficient while it’s happening. Creativity, judgement and good ideas often involve wandering off the approved path.

Speculative Decoding: A way of making AI faster by having a smaller model take a punt on what comes next, while a larger model checks the guess. If it’s right, everyone moves on faster. If it’s wrong, the bigger model fixes it. So basically: predict first, verify later. Similar to how humans operate: the junior employee has the first crack, the manager checks it, the junior gets blamed if it creates more work, and the manager gets the kudos if it saves any.

^{THE SHALLOW END}

‘Sam Almond’ Becomes a Meme

The CEO of OpenAI can’t catch a break. From an anti-AI activist throwing a Molotov cocktail at his house to headlines about Altman’s tendency to lie and being widely regarded as untrustworthy, the Internet is now treating him like a joke. This followed a recent podcast interview with Alex Heath “to discuss the future of technology while also addressing the concerns over AI’s impact on society”. Things went awry when Altman tried to dismiss concerns about ChatGPT’s water use. He claimed that 38,000 ChatGPT queries use roughly the same amount of water as producing a single almond in California. That comparison immediately backfired online. People started calling him ‘Sam Almond’ and circulating AI-generated images of Altman as an almond. Others mocked the comparison itself, pointing out that almonds are food, while ChatGPT queries aren’t a basic human need.

Dear Dr Comms

I’m a kindergarten teacher, and I’ve just been hired to help train robots. Apparently, they’re supposed to learn like small children: explore, make mistakes and gradually work out how the world works. I start Monday and have no idea what to bring. Crayons? Play-Doh? Finger paints? Butcher paper? Sleeping mats? How do I prepare? Yours, Miss Roboto

Dear Miss Roboto
A fair question. Let’s ask people experienced in preparing others for the big scary work world:

🎓 Your Wife, the School Careers Adviser
“Start with the basics: communication, teamwork and resilience. Also get them familiar with LinkedIn. A profile and a good headshot of them smiling warmly are mandatory for a successful career.

📦 Your Amazon Courier Driver
“Forget crayons. Teach them to recognise a front door, find an apartment intercom and understand the phrase ‘leave in a safe place’. If they master that, Amazon will hire the entire class by Friday.”

🐕 Your Dog Walker/Trainer
Reward good behaviour. Firm voice for bad behaviour. Lots of repetition. And if one starts ignoring basic commands and begins to aimlessly circle the room, or starts sniffing, growling, and biting other robots, don’t panic. That’s normal for the first few weeks.”

Before you can walk or open a jar of pickles, you have to learn how to crawl.

^{BIN THIS…}

One major consultancy has apparently made it its mission to help you:

At this point, an extraordinary amount of corporate value appears to have been accidentally locked somewhere.

Who locked it? Where is it? And why does every consultancy seem to have the key?

Please put unlock back in the bin with empower, transform and at scale.

Know someone who lives for this kind of nonsense? Forward this email to them and help me spread the dysfunction.

You can also subscribe to my other newsletter, Lead Different, for a serious take on strategic communications in B2B tech.