Skip to content

Maybe it's all just an elaborate April Fool's joke?

The Roundup is what happened in CX and tech, and what was nonsense about it. All Roundup issues →

News from Around Supportlandia

Hey friends! Happy to be back on schedule this week. I’m still trying to catch up on this and work and life, though, so I’ve got a content-heavy issue for you. But hey, I’m not being a perfectionist about it, so I think that’s growth!

Speaking of catching up, I’m sure y’all noted the absence of Bad Job Bingo on Monday; it should be back next week. Probably. I hope.

Oracle betrays its name and its employees

On March 31 around 6am, somewhere between 20,000 and 30,000 Oracle employees — roughly 18% of the company's global workforce — received an email from "Oracle Leadership” saying their role had been eliminated, effective immediately, followed by an immediate system lockout.

If you’re thinking a company named Oracle would be wise and trustworthy enough to treat its employees with some care…oh, who am I kidding, no one’s thinking that. This is Larry Ellison’s Oracle, after all. Of course he has no qualms using the brutality of RIFs to line his company’s pockets, as former 25-year Oracle veteran Dan Williams put it. But perhaps Tim Davidson said it most simply: loyalty is dead.

The Anthropic Variety Hour Week

Meanwhile, Anthropic saw a variety of news items break this week, none of them particularly good.

News item one: Anthropic is reportedly eyeing an IPO as soon as October 2026, potentially raising more than $60 billion, which of course made me speculate on Bluesky about some of their recent moves:

Techmeme @techmeme.com

Steph Lundberg on Bluesky

News item two, arguably the biggest story: While preparing to go public as the company that cares about safety, Anthropic accidentally leaked its proprietary Claude Code source code to the world. When releasing version 2.1.88 of its Claude Code npm package, a developer left in a source map file that — true to its name — led enterprising users directly to Claude’s source code. One of those enterprising users uploaded the codebase to a public GitHub repository, which, as of 9pm MT tonight, has 163,000 stars and 101,000 forks.

Image of K-2SO from Andor saying to a logo of Claude: "Congratulations! You are being open sourced. Please do not resist."

Anthropic responded with a sweeping copyright takedown request for over 8,000 copies (which it eventually reduced to 96 copies). Regardless, as Futurism reports, this is extremely ironic given Anthropic’s insatiable appetite for the intellectual property of others:

Back when Anthropic was still a nascent splinter group formed from former OpenAI researchers, for instance, it needed access to a wealth of high quality training data to build its Claude AI model.

To do that, it first relied on digital books. But it didn’t pay for them or choose only to use ones in the public domain. Instead, it downloaded millions of pirated volumes from the online “shadow library” LibGen. While LibGen doesn’t position itself as a pirate website, Anthropic also downloaded books from a similar hub literally called “Pirate Library Mirror.” (Anthropic cofounder Ben Mann was ebullient about the site’s launch: “just in time!!!” he wrote in a message to employees, along with a link to the site.)

The practice was unearthed in a lawsuit brought by a group of authors against Anthropic, which ended in a $1.5 billion settlement after a judge deemed the use of the pirated books to be illegal.

Anthropic also scanned and destroyed millions of used physical books in a secret initiative called Project Panama. The process involved cutting the pages out of the volumes using higher powered machinery, which once scanned were tossed out and recycled.

News item three: Anthropic admitted this week that users are hitting usage limits "way faster than expected." A subscriber to the $200/month plan reported using Claude a grand total of 12 days out of 30, while a user on the $100/month plan said they “used up Max 5 in 1 hour of working.”

Anthropic has tried to give users usage credits to make up for the trouble, but even that is not going smoothly, something I’ve experienced myself. Having seen that credit was being offered, I logged into my Claude account, found a white button inside a blank square, clicked it, and saw nothing but an empty message with a gift box animation that appeared and then immediately disappeared before I could take a screenshot (so I have no idea if I got a credit or not). Others can see the credit but cannot claim it thanks to an error.

And in something that’s perhaps even more ironic than Anthropic suddenly being an IP warrior: I’ve seen a report from a CX pro in one of my Slack communities that Anthropic’s AI customer support agent (powered by Intercom’s Fin) just isn’t AI agenting at all.

If you’re up for an excellent long-form read, Ed Zitron wrote a piece this week that ties all of this together and draws some uncomfortable parallels between AI’s current trajectory and the subprime mortgage crisis. It might ruin your weekend, but at least you’ll be informed?

Speaking of AI support agents not agenting…

CNBC reported this week on the rocky relationship between consumers and AI customer service. According to a Qualtrics study cited in the story, “nearly one in five consumers who have used AI for customer service saw no benefit from the experience, […] a failure rate almost four times higher than for AI use in general.”

The complaints from customers are familiar: endless loops, limited functionality, strict adherence to policy without recourse, and the general sense that the bot exists to exhaust you into giving up on pursuing help rather than offering any real assistance.

"Too many companies are deploying AI to cut costs, not solve problems," said study author Isabelle Zdatny of Qualtrics, "and customers can tell the difference."

I think we all can.1

And Now for Some Good News

Again, there was a ton of career wins this week; the great news below barely scratches the surface of the announcements I saw on LinkedIn. Have hope! I know I’m trying.

Read, Watch, and Listen

LinkedIn gets fooled (and its own temporary section)

It's April Fool's Day week, and (for once) LinkedIn did not disappoint.

First up, my absolute fave: Deon Nicholas — who you may remember from last issue's footnotes, fresh off the Forethought/Zendesk acquisition — announced he was leaving tech to become a full-time rapper-singer-songwriter. Complete with music video. The commitment to the bit was as complete as the song was a bop.

Bonnie Dilber hurt my feelings.

Finally, Benjamin Devey signed on to work to find his entire team had changed their names and profile pictures in Slack to impersonate him. A Slack all about the Benjamins! (Sorry, that was terrible.)

Read

Even astronauts need tech support.

Stephen Driggs let an AI he named "Agent LinkedInSlop5" take over his feed for the day, with predictably hilarious results. Given what LinkedIn feeds look like on a normal day, it's possible no one noticed.

Fernando Duarte wrote about what to do when your AI starts failing.

Ryan Wang of Assembled wrote in Forbes about how AI's real-world impact on CX in 2026 looks far more nuanced than leaders expected.

Kateryna Babenko wrote about how to make your LinkedIn feed usable again. (You guys' LinkedIn feeds were usable???)

James Baldwin wrote about the work Zapier put into its V2 AI Fluency Rubric. Worth a read for anyone thinking about AI adoption on their own teams.

Kristi Faltorusso wrote about why you should stop trying to keep customers who don't want to be kept.

Renate Munson wrote about why support starts before the sale — the moment a potential customer first learns about your product.

Walter Cabrera wrote about why low survey response volume becomes a CX problem in itself.

Permission Slips interviewed Jae Washington on thinking (and living) outside the box.

Kirsten Penaloza asked what's more dangerous: upward feedback or bad leadership?

Tom Sweeny introduced his Support Measurement Stack — a three-layer framework for organizing support data into a chain of evidence that connects activity to outcomes to business value.

Watch

Harald's Curious Corner talked with Clea Mahoney, Dan Braithwaite, Melissa Kruminas, and Kristine Kukich about customer education metrics that actually impact revenue.

Listen

Ben Foden's podcast The CX Files rebranded this week to CX Heroes. New name, same great show.

Rebels of SaaS (hosted by Dannah Vaughan) wrapped up Season 3 with Kate Neal talking about how to AI-proof your Customer Success career.

Upcoming Events

Community Week Online 2026
April 7-10 at 3pm BST. Hosted by Led by Community. Get tickets here.

AI Customer Success Summit
April 8 at 10am ET. Online event hosted by Customer Success Collective. Register here.

AI Lovable Hackathon: Community Tools Edition
April 14 at 2pm BST. Hosted by Led by Community and Lovable. Register here.

The Blueprint for Scaling AI Support: Live in Austin
April 14 at 2:30pm CT in Austin, TX. Hosted by Intercom. Register here.

Chief Customer Officer Summit Silicon Valley
April 15 at DoubleTree by Hilton Hotel, San Jose, CA. Hosted by Customer Success Collective. Request invite here.

Shaping Your Story Through Your CV
April 16 at 4pm BST. Hosted by Led by Community, feat. Todd Nilson. Register here.

Phoenix CX Gathering
April 16 at 5pm MT at Pitch Scottsdale, Scottsdale, AZ. Hosted by Donna Drehmann and Vicki Brackett, feat. Nate Brown.

Road to Summit Tour: Community IRL in LA
April 16 at 5:30pm PT at Santa Monica Brew Works, Santa Monica, CA. Hosted by CMX by Bevy. Register here.

Customer Success Meetup - New York
April 21 at 5:30pm ET in New York City, NY. Hosted by Customer Success Meetup. Register here.

Is Your Contact Center Actually Delivering Value?
April 23 at 3:30pm ET. Hosted by Deepak Selvaratnam, feat. Lisa Diehl, Brian Jeppesen, Matt Woody, and Jason Mercer-Pottinger. Register here.

2026 CX Leaders Advance
April 27-29 at the Omni King Edward in Toronto, ON, CAN. Hosted by CXPA. Register here.

Community as a Catalyst for Growth
April 29 at 4pm BST. Hosted by Led by Community, feat. Candace Grobler. Register here.

CMX Summit 2026
April 30-May 1 at Fox Theatre, Redwood City, CA. Hosted by CMX by Bevy. Get tickets here.

kill the imposter syndrome in your head because not only is there someone out there doing it worse than you, they're also using chat gpt to do it
  1. Hat tip to Jake Finch for putting this on my radar.

Comments

Before you post

Disagree with me all you like — that's half the point. Don't come after other readers, don't post anyone's private details, don't pitch anything, and don't post anything you didn't write yourself. I delete what breaks that and I don't argue about it. Full version →

Something off in this thread? Send me a note and say which post it's on.