some things to say this week :)

WebinarTV Secretly Scraped Zoom Meetings of Anonymous Recovery Programs WebinarTV scraped and shared 12 steps-based anonymous meetings for people recovering from addiction and other private support groups.

WebinarTV Secretly Scraped Zoom Meetings of Anonymous Recovery Programs www.404media.co/webinartv-se...

-> rapidly climbing the ranks of the sleaziest "businesses" of all time

We're in dire need of fresh approaches to enterprise AI and details on how they work, says @jon.diginomica.com So he caught up with @zoho.com 's Raju Vegesna and Ram Ramamoorthy to discuss privacy, data sovereignty, and customer control over "intelligence" in our latest podcast: bit.ly/3QtV0Xa

yeah let's do this....

well, your concern is reality, I'm going live on a piece on that in a couple hours. AI it often good at outlier detection, so that's a good one, but there are other useful HR things. Execution of payroll is easy, allowing users to adjust their payroll preferences on the fly isn't....

yes - the cost of outliers is a major factor in AI use case selection and design......payroll is one of the worst agentic AI use cases.... :) but, you could design an self-service AI bot that included a deterministic payroll execution option (e.g. switch from monthly to weekly or an advance etc).

Microsoft's AI in its own terms: "use Copilot at your own risk" The Copilot terms of use, updated last October, draw clear limits around what the software is meant to do. The document states Copilot is for entertainment purposes...

Microsoft's AI in its own terms: "use Copilot at your own risk" www.techspot.com/news/111949-...

so - we're embedding risky tools everywhere, we'll leave it to you to determine that risk.

-> err, doesn't risk factor into any reasonable assessment of value as well?

Zooming Out: WebinarTV’s Rampant Scraping of Online Meetings Zooming Out: WebinarTV’s Rampant Scraping of Online Meetings

Zooming Out: WebinarTV’s Rampant Scraping of Online Meetings cyberalberta.ca/zooming-out-...

-> There are so many unethical business practices wrapped up in this lawsuit-waiting-to-happen it's hard to know where to begin...

Paul McCartney Banned From Reddit After Promoting Himself in Paul McCartney Subreddit Reddit blamed a technical glitch for the removal of the living legend’s concert footage.

Paul McCartney Banned From Reddit After Promoting Himself in Paul McCartney Subreddit www.404media.co/paul-mccartn...

-> I'm a fan of strong community moderation but this does seem extreme lol

Getting real about agentic AI projects - what's working, and what's not? The AI project lessons are boiling over - but so is the agentic AI keynote hyperbole. How will customers find their way? Grab your beverage of choice, and have a look at my spicy review of what's work...

indeed....I bang that particular drum a lot :) see #4 in the what works section: re - use case design and human escalation :) diginomica.com/getting-real...

agreed, we must wade into it with discernment but not cling to our old thinking :) thx for providing a spike of thoughtful probing on the right issues, that is refreshing. gotta run for now but sure we'll do it again.

yes - edge cases are always an issue, this architecture does include the ability to look at so-called reasoning traces/what went wrong but IMO managing user expectations and use case design (escalation and evaluation) are both needed to handle outliers. See the example re: "turnover" in piece

p.s. you and I are interested in same questions, I have learned I can't boil the ocean in one blog post or event but I'll continue to tackle the kind of points you've raised as the year chaotically unfolds :)

yes... as per the article the domain specific finance models are more cost effective and more financially literate, had some discussions on cost, as noted this keeps costs significantly lower but as you know AI model costs are a moving target, and degree of usage impacts profit margins

2/2 I didn't get to it here but most popular of the finance agents so far is (I'm told) Financial Intelligence agent which essentially surfaces sourced/audited data for users rather than having to generate reports or look up data. It's still early days for these better more contextual architectures

That will be something to document more as year unfolds re: customer adoption. I don't like either of the use cases you cited though. The goal here was to explore why a disciplined approach to finance data for agents works far better/more economically than off the shelf LLM agents architecturally.

yep, it's that time again....

BuzzFeed Nearing Bankruptcy After Disastrous Turn Toward AI Three years after its pivot to AI, the writing is on the wall for BuzzFeed. The company said there's "substantial doubt" it can keep going.

BuzzFeed Nearing Bankruptcy After Disastrous Turn Toward AI futurism.com/artificial-i...

"The brutal reality check seemingly hasn’t put Peretti off from pursuing AI, though."

-> hey, if you keep pivoting you may actually land on something people might care about lol

Why OpenAI really shut down Sora | TechCrunch OpenAI's decision last week to shut down Sora, its AI video-generation tool, just six months after releasing it to the public raised immediate suspicions. The app had invited users to upload their…

Why OpenAI really shut down Sora techcrunch.com/2026/03/29/w...

"While a whole team inside OpenAI was focused on making Sora work, Anthropic was quietly winning over the software engineers and enterprises that drive revenue."

-> ouch....

Walmart fires OpenAI in playbook-changing move www.thestreet.com/retail/walma...

"OpenAI underestimated how difficult the enablement of transactions was going to be"

-> executives continue to struggle with the limits of probablistic tech versus their fantasies of amazing user experiences

the SaaS value squeeze, headcounts versus new business models, alternatives to big AI - that's enough for a video

Okay, I was running a bit hot... but, I also boiled down six months of research.

with extra spices....

going up :)

Dancing humanoid robot causes chaos at busy California restaurant A dancing humanoid robot at a Haidilao restaurant in Cupertino caused chaos after getting too close to diners and smashing plates.

Dancing humanoid robot causes chaos at busy California restaurant interestingengineering.com/ai-robotics/...

-> Styx predicted this! #mrroboto

(via @iC)

What privacy? Meta's smart glasses are filming unwitting naked people Meta's Ray-Ban smart glasses are a privacy nightmare, with footage of naked people, sensitive information, and violent acts captured and seen by Meta's AI and an army of employees.

What privacy? Meta's smart glasses are filming unwitting naked people

appleinsider.com/articles/26/...

-> everything is awesome.... maybe a techno-optimist can chime in here

OpenAI to nearly double workforce to 8,000 by end-2026, FT reports www.reuters.com/business/ope...

-> because AI is so incredibly good at replacing skilled humans..... except when you want to push your business model beyond grifting and therapy chats

Meta backtracks on decision to end Horizon Worlds VR after fans speak up A day after announcing the Horizon Worlds would only be available on a mobile app starting in June, Meta said it changed its mind.

Meta backtracks on decision to end Horizon Worlds VR after fans speak up www.cnbc.com/2026/03/19/m...

"Meta previously announced on Tuesday that Horizon Worlds would be taken off the Quest Store"

-> I love a company that executes with a clear, bold vision lolz

still pretty hot off the griddle..... :)

Viral 'Quittr' Porn Addiction App Exposed the Masturbation Habits of Hundreds of Thousands of Users A couple of 20-year-old developers make $500,000 a month promising to help men to stop watching porn, but exposed their private porn watching habits.

Viral 'Quittr' Porn Addiction App Exposed the Masturbation Habits of Hundreds of Thousands of Users www.404media.co/viral-quittr...

-> If that doesn't help you quit I don't know what will.....ugh