I help founders get ready and connect with investors • Seasoned entrepreneur • Building tech startups for 25yrs (zero→£mil ARR, twice)

With 29 years in startups, I now run the Gipity AI Dev Studio: Expert help to launch your MVP in 30 days. www.gipity.com

𝐌𝐲 𝐡𝐢𝐬𝐭𝐨𝐫𝐲... 👇

1968 BC (Before Computers): Born into analogue world

1982-86: Discovered computers! Vic 20, BBC micro, ICL 2900 mainframe & more

1986-89: Computation degree, UMIST (University of Manchester, Institute of Science & Technology)

1989-99: IT consultant/project manager: Logica, IBM, Reuters, Bankers Trust & Deutsche Bank

1994-97: first side hustle: scuba diving instructor & own London dive school

1996: published first website & first ever disk-mounted website on a UK magazine cover

1999: developer/project manager at 2 early London digital agencies

1999: vision/architect/investor. created first UK under-a-tenner domain name business, Easily

1999: built+launched early/successful dot-com: domain/email/web hosting ISP

1999-2001: ceo. managed growth from 0 to 2,000 daily domain sales & 30+ staff in under 2yrs

2000: invented the web-based DNS dashboard for domains

2000: created early no-code solution for one-page websites: £100k ARR

2000: highly commended "Entrepreneur of the Year": European E-commerce Association E-business 2000 awards

2001: 64th in Sunday Times "Top 100 pre-IPO e-businesses in Europe" E-League

2001: ran one of the first successful web affiliate schemes, generating 40% of rev

2004: hit 400,000+ domain registrations, 30,000+ websites & 40,000+ email accounts

2004: 3,500 commission earning affiliates & word of mouth to 100,000+ customers

2004: negotiated successful multi-million pound dot-com exit/sale of Easily

2002: vision/investor. launched one of the first ever B2B mobile text marketing services: SMS Expert

2007-11: early digital nomad: B2B Saas startup from beach/desert

2007-19: ceo. ran leading text marketing B2B Saas: zero to multi-million turnover

2007-19: cto/architect. developed 850k lines of code & supported 1mil daily transactions/API calls

2020... consultancy/advice/support for early-stage tech startup founders

2022... helping hand-picked founders find the right investors

• 2018: vision/architect/investor/cto...
Clickacall: instant video calls from anywhere with single click
Nearevo: proximity discovery for internet of things
Scan+Click: contactless business cards
Secly: personal online security

• 2021 l2o: supporting next generation of startups
• 2022 Venture Cards: all startups, in one place • connecting the right founders and investors

• 2024 The Gipity AI Dev Studio: Expert help to launch your MVP in 30 days

Steve Procter

CONTROVERSIAL: If you're ready to create an internet/mobile based startup in 2026, you do not need investment. What's that I hear: "but there are a few setup and running costs". Well, no shit Sherlock! That's how business works.

If you expect it to be a free ride and for someone else to take the risk before you've even started, this game aint for you. Your idea, your risk.

Five years ago, sure, it was tough back then. You needed the app ready just to discover if there was even a market. And the developers who would build it saw you coming and demanded £150k just to get out of bed.

But those days are over. Building your MVP is cheap. Signing up to a handful of AI-driven marketing/growth platforms is even cheaper. Your costs for getting from day 0 to day 1 are minimal, and should come out of your pocket.

If your savings account can't cover it, then give up Netflix, or your 2nd annual holiday, or a meal out and bottle of wine each week.

Take your own risks if you are so convinced the idea is that great. And, according to page 17 of your pitchdeck, it is going to be massive and hit a $1bn market within 5 years!

So what's a year without Netflix or dining out when that's the goal? You will be too busy building and making sales for a social life anyway.

Investment is for scaling. Not for creating the damn thing on day 1.

---

* smallprint: yes, sure, we know that if you're in quantum something or other, or deeptech, or need a van to deliver your mince pies to factories, or to buy stock, or do medical research, [or insert 1,000 other ideas], then you are special and this post does not apply to you. So please, walk on by, nothing for you to see here. This is for people building apps, websites and stuff like that.

Launch your MVP in 30 days | 30yrs in startups (exited. 0→£mil * 2)

I just read an expert's idea of a "5 yr plan for startups". More theoretical tosh! As a founder you need to be out in much less than 5 yrs, sat on the beach with a piña colada, and planning your next startup.

Let's be honest, for most founders, if you are still in at the 5 year mark, you've probably missed the exit ramp and are a zombie startup, kidding yourself you're still capable of "scaling" and reaching mythical unicorn status.

The speed of startups in this mad new AI-backed world means, if the business is doing well at 2-3 years, you should find someone bigger who knows how to launch it into the next orbit. For most founders, this is not you.

Exit at 2-3 years. Low £millions. Go again. Repeat until the care home calls.

Oh, and by the way, that exit money needs to be hard cash. And it goes to founders only - no investors to take any cream off the top.

If you spent 6-12 months of those 2-3 years faffing about finding an investor, then you've wasted the most critical period of a startup's early life.

You can disagree all you like, and spend half your days on here, talking the talk and arguing with AI-written posts.

The startup process is simple: Build, sell, sell-out.

If any part fails, then pivot or find a new idea, and go again.

JFDI !

Hey Linkedin algorithm, send this to everyone subscribed to my newsletter. Ask them to let me know if I should start writing it again.

Hey all, I've just remembered I have this newsletter. It was set up about 300 years ago, and somehow it has collected 6,720 of you 😊 So my big questi

Open article: Hello, is anybody seeing this? by Steve Procter • 1 min read

Startup circus 'experts' say "build it and they will come" is wrong. Really? Post daily on social, grow your followers & fans, then give them a little nudge when you launch something new. Watch them come. Selling has never been easier.

Someone asked me if they should create a business plan, pitchdeck, and financial projection for their brand new startup and app idea? My answer is simple: Hell no! Real founders fly by the seat of their pants.

Sure, write down a few pages of ideas and notes. But otherwise, just focus on building, making sales, and pivoting.

That last one, pivoting, is what makes too much planning a complete waste of time. Don't kid yourselves your idea is so solid you won't need to. You will be pivoting so much, you'll feel like a ballerina practicing her pirouettes.

So stop talking the talk. That's what Linkedin armchair experts do.

It's time to build fast and make sales. Which is what real founders do.

I saw this post recently: "has anyone managed to create a profitable startup yet from vibe coding?" Such a stupid question. AI-assisted dev doesn't make sales, that's your job. But it does help you start selling much sooner.

Oh good grief! Can we stop blaming AI dev tools whenever we see a crappy vibe coded app. Platforms like Replit are fantastic and work great when you know how to control them and feed them the right information. But if you don't, then, as us old-schoolers like to say: Garbage In, Garbage Out.

“the only metrics that will truly matter to my life are the individuals whom I have been able to help, one by one, to become better people.” ― Clayton Christensen


Stephen Klein

Yann LeCun just left Meta.
Let that sink in.

One of three people who invented modern deep learning. Turing Award winner. 

Chief AI Scientist at a company spending billions on AI.

Gone.

Not to a competitor. Not to retirement. 
To a startup building technology 
Meta apparently doesn't prioritize.

This matters more than most people realize

LeCun has spent 40 years being early and right. 

In the 1980s, he couldn't find a PhD adviser because machine learning was "a dead end." 

Three decades later, that dead end became the foundation of modern AI.

Now he's saying large language models, the technology behind ChatGPT, Gemini, and Meta's Llama, are another dead end.

"Nobody in their right mind would use LLMs of the type that we have today" within 3-5 years, he said at MIT last month.

His alternative: world models. Systems that learn from visual data and build causal understanding, like a child learning physics by dropping things thousands of times.

Current LLMs? 
They predict the next word. 
They don't understand gravity. 
They don't know that objects exist when you're not looking at them.

LeCun compares their intelligence to a cat's. He thinks the cat is smarter.

Here's what troubles me:

Meta is spending countless billions on technology that its own Chief AI Scientist publicly called a dead end. 

Either LeCun is wrong (possible, but his track record suggests otherwise), or Meta is chasing market narratives over technical reality.

His departure suggests he's betting on the latter.

The implications for current AI investments are significant. If LeCun's 3-5 year timeline holds, companies built entirely on LLM scaling may face architectural obsolescence.

That's not a guarantee. Even geniuses miss sometimes.

But LeCun has never been wrong about something this big before
********************************************************************************
The trick with technology is to avoid spreading darkness at the speed of light.
Stephen Klein
Founder & CEO, Curiouser.AI
Instructor, UC Berkeley Haas School of Business
WeFunder (link in comments)

Sources:

Wall Street Journal - LeCun departure and startup plans (November 2024)
MIT Symposium - LeCun remarks on LLM limitations (October 2024)
Meta FAIR Research Publications - World model architectures
Turing Award Archives - LeCun, Bengio, Hinton recognition (2018)

Founder & CEO, Curiouser.AI | Berkeley Instructor | Building Values-Based, Human-Centered AI | LinkedIn Top Voice in AI

WRONG 60% OF THE TIME:
ChatGPT. Gemini. Grok. Perplexity. Copilot. DeepSeek 

Premium models, $20 to $40 per month, performed WORSE than free versions.⁵ Higher cost. Higher confidence. Higher error rates.

Columbia University's Tow Center tested eight AI search engines on the simplest possible task: Given a direct excerpt from a news article, identify the headline, publisher, date, and URL.¹

Collectively, they gave incorrect answers more than 60% of the time.²

Not on complex reasoning.
Not on nuanced analysis.
Not on predicting the future.

On citing their sources.

A basic Google search gets this right instantly.

The individual results:

Grok 3: 94% wrong
Gemini: 1 correct answer out of 200
ChatGPT: 67% wrong
Perplexity: 37% wrong (best performer)

But the error rate isn't the real problem.

The confidence is

ChatGPT gave incorrect information 134 times. It expressed uncertainty just 15 times.³ It never once said "I don't know."

Never. Once.

Grok 3 didn't just get sources wrong, it invented them. 
Out of 200 queries, 154 citations led to fabricated URLs.⁴ 
Pages that don't exist. 
Evidence that was never there.
They're not making mistakes.
They're manufacturing proof.
And here's where it gets absurd:

Now think about where these tools are being deployed.

Writing code. Summarizing contracts. Informing investment decisions. Drafting legal briefs. Generating medical information.

If they can't accurately cite a news article, the most basic intellectual honesty task, what else are they confidently fabricating?

The researchers didn't mince words: "Most of the tools we tested presented inaccurate answers with alarming confidence."⁶

This isn't a bug to be patched.

This is the architecture working as designed.

Large language models don't know things. They predict what sounds plausible. 

They've learned that confidence gets rewarded, accuracy be damned.

We're building critical infrastructure on foundations that fail 60-94% of the time
********************************************************************************
The trick with technology is to avoid spreading darkness at the speed of light.
Stephen Klein
Founder & CEO, Curiouser.AI
Instructor, UC Berkeley Haas School of Business
WeFunder (link in comments)

Sources:
¹ Columbia Journalism Review, Tow Center for Digital Journalism, "AI Search Has A Citation Problem," March 6, 2025 - Study analyzed 1,600 queries across 8 AI search tools
² Ibid. - "Collectively, they provided incorrect answers to more than 60 percent of queries"
³ Ibid. - ChatGPT "signaled a lack of confidence just fifteen times out of its two hundred responses, and never declined to provide an answer"
⁴ Ibid. - "Out of the 200 prompts we tested for Grok 3, 154 citations led to error pages"
⁵ Ibid. - Premium models "demonstrated higher error rates... tendency to provide definitive, but wrong, answers"
⁶ Ibid. - Direct quote from study authors

Will Mira Murati Collapse the LLM Bubble and Redefine the AI Industry?

The LLM arms race was never about the enterprise.

It was (and remains) about spectacle: 

bigger models, bigger clusters, bigger burn rates, and the illusion that “more parameters” meant “more intelligence.”

Think Detroit in the 60s
(Though I did love the 1965 Pontiac GTO)

It was (and remains) about building big bloated unreliable things that provide the illusion of AGI, like a party trick, or three card monte, and sucks up all the money in Silicon Valley creating a momentum investors wet dream.

Then Mira Murati left OpenAI and quietly did something unprecedented:
She published the flaw everyone else hid.

Her new team at Thinking Machines Lab released a paper, “Defeating Nondeterminism in LLM Inference,” demonstrating that today’s frontier models are structurally incapable of delivering consistent answers.¹

GPU kernels are not batch-invariant.

Meaning the output you get can depend on which random strangers hit the server at the same moment.

Their test on Qwen-3-8B is the kind of data CFOs, CIOs, and boards never see:
1,000 identical prompts at temperature = 0 → 80 different outputs with different reasoning paths.²

Not a hallucination problem.
A determinism problem.

The enterprise was promised “automation.”
What it got was “variability.”

Murati’s team solved consistency, batch-invariant kernels, but she didn’t package it as AGI magic.

Now look at her roadmap, and you can see why this may collapse the LLM bubble:

1. Real Enterprise Reliability

Reproducibility is the foundation of safety, compliance, auditability, and P&L value.

2. Open-Source Over Closed-Source Taxation

Open infrastructure ends the per-token tax and the illusion that scale is a moat.
It also ends vendor lock-in, the part Big AI hopes no one talks about.

3. Fine-Tuning Costs That Don’t Require a Capital Call

Murati is pushing tools like Llora to reduce fine-tuning costs by an order of magnitude.

4. Smaller, Smarter, Task-Aligned Models
THIS IS THE FUTURE

While incumbents chase 400B-parameter monsters, Murati is designing practical, domain-specific intelligence that enterprises can actually control.

5. Governance Embedded at the Architectural Level

During her tenure at OpenAI, she set safety standards that quietly became industry norms: red-teaming, ethics boards, bias evaluation, disclosure protocols, governance frameworks.³

At Dartmouth she said:

“Safety, security, and societal impacts are not an afterthought.”⁴

At Davos 2025 she said the line that history will circle back to:

“AI without values is intelligence without conscience.”⁵
(Sources in comments)
********************************************************************************
The trick with technology is to avoid spreading darkness at the speed of light.
Curiouser.AI
Mission: Re-ignite imagination. Augment creativity. Create jobs. Build industries we haven’t named. Time to rock the free world. WeFunder link in comments.

How Meta Keeps $27 Billion Off Its Balance Sheet
A tactic similar to Enron’s, though in this case, apparently perfectly legal.

Meta needs $27 billion to build a massive AI data center in Louisiana called Hyperion.¹

But taking on $27B in debt directly would look terrible: it depresses returns, invites scrutiny, and raises real questions.

Solution: Do what Enron did. Put the debt somewhere else.

Blue Owl Capital created a special-purpose company called Beignet Investor LLC.²

Beignet issued $27B in bonds, with PIMCO as the anchor lender.³
Blue Owl owns 80% of the data center. Meta owns 20%.⁴

Meta still gets 100% of the compute. They just call the payments “rent.”
Beignet uses that rent to repay the $27B in debt.

**The debt never shows up on Meta’s balance sheet.**⁵

Yet investors treated the bonds like Meta’s anyway, the 2049 maturity bonds came in at A+ from S&P, strengthened by a Meta “backstop guarantee” if lease partners exit.⁶

Then Meta raised another $30 billion in traditional bonds days later.⁷

Total: $60B in a single month.

PIMCO has already netted over $2 billion in paper gains as the bonds trade above 110 cents on the dollar.⁸

As Gil Luria at D.A. Davidson put it:

**“The term special purpose vehicle came to consciousness about 25 years ago with a little company called Enron.”**⁹

And Meta isn’t alone. This is the financing engine behind the entire AI boom.

• Hyperscalers have taken on $121 billion in debt over the past year
a 300%+ jump from normal levels.¹⁰

• Tech firms are projected to raise $1.5 trillion in debt by 2028 to fund data centers.¹¹

• JPMorgan estimates Alphabet, Meta, Amazon, Microsoft, and Oracle will need $570B in capex in 2026 alone, up from $125B in 2021.¹²

• Big Tech will spend $400B on AI infrastructure this year, roughly half their total cash flow.¹³
• OpenAI plans $1.4 trillion in data center spending across eight years.¹⁴

Paul Kedrosky calls it:
**“The first bubble that combines all bubbles.”**¹⁷

The uncomfortable question: What are we actually financing?

Because the infrastructure is real.

But the ROI?

• MIT: 95% of companies see zero financial return from generative AI projects.¹⁵
• McKinsey: Nearly 80% report no measurable bottom-line impact.¹⁶

• S&P’s Ruth Yang:

**“We’re funding 20–30 year data center deals for a technology that may be obsolete in five.”**¹⁹

We are watching a 2008-style financial engineering cycle wrapped around a technology delivering a 5% enterprise success rate.

Hard to know what's real other then the debt.
(Sources in comments)
********************************************************************************
“The trick with technology is to avoid spreading darkness at the speed of light.”
Curiouser.AI is backed by a growing community of amazing and thoughtful people all of whom believe that we can design AI to augment work, create jobs, and elevate humanity. Our first use case is to help founders and SMBs succeed. If this resonates visit WeFunder (Link in comments)

Peter Thiel Dumped 100% of His NVIDIA. Here's Why That Matters.
He went from 40% of portfolio to zero in 90 days. When someone who spotted the '99 bubble makes a move this decisive, it's not noise, it's signal.

Peter Thiel isn't someone I quote often. But when a guy who called the dotcom bubble compares AI to 1999 and exits his entire NVIDIA position in one quarter, 

The Move

SEC filings this week: Thiel's hedge fund sold all 537,742 NVIDIA shares during Q3, worth ~$100 million. NVIDIA was 40% of his portfolio in Q2. 

By September 30: zero shares.[1]

That's not rebalancing. That's a conviction call.

What He's Seeing

Thiel noted earlier that 80-85% of all AI profits flow to NVIDIA, calling it "very strange" because the money's in hardware, not applications.[2] 

He's now comparing AI to 1999: investors "pricing in a future that will take 15-20 years to unfold."[3]

The data backs him up:

Enterprise AI adoption declined from 14% to 12%
95% of AI pilots fail to reach production
Billions spent on infrastructure, monetization still elusive

Where He's Going Instead

Thiel didn't leave tech. He rotated into Microsoft (+49K shares) and Apple (+79K shares).[4]

Why? These companies have:

Diversified revenue with proven monetization
AI as enhancement (Copilot, Apple Intelligence), not existential bet
Actual paying customers at scale

NVIDIA's valuation depends on one thing: exponential AI chip demand growth. Microsoft and Apple have cushions.

The Pattern

Thiel's not betting against AI. 
He's betting against the automation-obsessed implementation model that enterprises keep rejecting

He's Not Alone
SoftBank sold its entire $5.8B NVIDIA stake in October.[5] 
Michael Burry took massive put positions against NVIDIA and Palantir.[6]

When sophisticated investors who called previous bubbles exit the same position, that's pattern recognition.

The market is distinguishing between AI infrastructure bets (risky, timing-dependent) and AI applications with proven models (sustainable, timing-agnostic).

And if there's one thing Peter Thiel understands, it's how to make money when everyone else is still drinking the Kool-Aid.
********************************************************************************
“The trick with technology is to avoid spreading darkness at the speed of light.”
Curiouser.AI is backed by a growing community of believers who share a simple conviction: the next era of AI must elevate people, not diminish them.
$450,000 raised in total. 60+ investors. 60% product paid conversion
If this resonates visit WeFunder (Link in comments)

Sources:
[1] TipRanks, Nov 17, 2025
[2] Benzinga, Nov 16, 2025
[3] TheStreet, Nov 16, 2025
[4] TipRanks, Nov 17, 2025
[5] Bloomberg/Yahoo Finance, Nov 17, 2025
[6] Bloomberg/Yahoo Finance, Nov 17, 2025

Brené Brown just said the quiet part out loud.
(CONFESSION: I think she’s brilliant and has the courage to say what most of us won’t.)

We have to stop obsessing with trying to fit in
It's OK to stand apart and think for yourself

She sits in rooms with tech billionaires. She hears what they say when the cameras are off.

“Hey, tech billionaire, what should my kids study?”

The answer is always the same:
The Stoics. Epictetus. Marcus Aurelius. Philosophy. History.

So Brown finally asked the obvious question:

“Which is it, dude?”

Because these same people publicly tell us:

School is dead.
Learn to prompt.
Let AI do the thinking.
Just keep scrolling.

Her conclusion is devastating:

“A thinking class is emerging. They’re reading philosophy and history.
The rest of us? Just keep scrolling. Don’t worry about the big words.
They’ll handle the big words for us.”¹

Meanwhile, something darker is happening:

IQ scores are declining for the first time in a century.

The Flynn Effect reversed around 1975, now dropping **7 points per generation.**³

Not genetics.
Environment.

And the newest data?

**AI usage is inversely correlated with critical thinking (r = -0.75).**⁴
The more we use it, the less we think.

When ChatGPT went down for four hours last December, Sam Altman admitted he’d “forgotten how to work without it.”

The man building the machine that thinks for us has already outsourced his own thinking to it.

And yet — the same elites making this future are also hedging against it.

Reid Hoffman says 50%+ of Silicon Valley billionaires have “apocalypse insurance”—bunkers, escape plans, New Zealand citizenship.⁵
Peter Thiel got NZ citizenship after 12 days, bought 500 acres, added a panic room.

They’re not just building the future.
They’re building escape routes.

They’ve already created two classes:

A thinking class and a scrolling class.
A writing class and a prompting class.
One reads philosophy.
One consumes content.
One thinks.
One outsources.

The question isn’t whether we’re getting dumber.
The question is whether we’re already too numb to notice.

The way out isn’t Luddism.
It’s refusing to outsource what makes us human.

Read difficult things.
Stand apart from the algorithm.
Practice thinking.
Practice attention.
Reclaim your mind.

“The trick with technology is to avoid spreading darkness at the speed of light.”

Curiouser.AI is backed by a growing community of believers who share a simple conviction:
the next era of AI must elevate people, not diminish them.

$500,000 raised
70+ investors

Please support us (link in comments)

Alice in Betaland just released

SOURCES

¹ Brown: Diary of a CEO, Nov 2025
² Isaacson: Steve Jobs
³ Bratsberg & Rogeberg, PNAS, 2018
⁴ Gerlich, Societies, 2025
⁵ Hoffman interview, The New Yorker, 2017

Is Meta the most harmful company in history?
Unsealed court documents make the case that they are

The 17x Policy

Instagram's former head of safety Vaishnavi Jayakumar testified under oath that Meta allowed accounts engaged in "trafficking of humans for sex" to accumulate 16 violations before suspension.

Sixteen chances to traffic children. Then a 17th strike.

Jayakumar called it "a very, very high strike threshold by any measure across the industry."¹

The Reporting Architecture

Want to report spam on Instagram? One click.
Intellectual property violation? One click.
Firearms promotion? One click.

**Child sexual abuse? Navigate a labyrinth designed to exhaust.
Jayakumar raised this repeatedly after joining in 2020. She was told it would be "too difficult to address."¹
The Recommendation Engine
An internal 2022 audit found Instagram's algorithm recommended 1.4 million "potentially inappropriate adults" to teenage users. In a single day.
The problem was frequent enough to require an acronym: IIC. Inappropriate Interactions with Children.¹
The Buried Research
In 2020, Meta's Project Mercury studied what happened when people quit Facebook for a week. Users reported lower depression, anxiety, loneliness, and social comparison.
Meta's response? Shut down the research. Never publish. Blame "media narrative."
One employee wrote internally: "If the results are bad and we don't publish and they leak, is it going to look like tobacco companies doing research and knowing cigs were bad?"²
Another distilled the philosophy: "We're basically pushers."³
The Growth Calculus
In 2019, Meta's own teams recommended making teen accounts private by default. The growth team ran the numbers: it would cost 1.5 million monthly active teenagers.
They waited until 2024. Five years. Billions of unwanted adult-minor interactions.
The internal reasoning, per court documents: "Taking away unwanted interactions… is likely to lead to a potentially untenable problem with engagement and growth."¹
The Pattern
Every institution eventually faces the same choice: mission or self-preservation.
The Catholic Church existed to save souls until it existed to save the Church.
Meta existed to connect humanity until it existed to grow Meta.
Both discovered the same solution: You don't solve the problem. You administer it.
********************************************************************************
“The trick with technology is to avoid spreading darkness at the speed of light.”
Curiouser.AI is backed by a growing community of amazing and thoughtful people all of whom believe that we can design AI to augment work, create jobs, and elevate humanity.  If this resonates visit WeFunder (Link in comments)

Apple just won the AI war by refusing to fight it

I'll admit bias here: Steve Jobs has been my hero since I first met him in the early 2000s.  Actually since he launched the Mac.

But this isn't nostalgia. This is data.

Six months ago, Wall Street mocked Apple for being "behind" on AI. The stock dropped 18% through June.¹ Analysts called Tim Cook "hesitant." 

The narrative was clear: Apple was losing.

Then something happened.

Apple's stock surged 35% since July. Meanwhile, the "AI leaders" stumbled—Meta and Microsoft slid into the red, even Nvidia underperformed.¹

What changed? Not Apple's strategy. Investor sanity.

The spending numbers are staggering:

Amazon: $125 billion in 2025 CapEx²
Microsoft: $94+ billion²
Alphabet: $91-93 billion²
Meta: $70-72 billion²

Total Big Tech AI spend: $380+ billion this year.²

Apple's projected CapEx through fiscal 2026? $14 billion.¹

That's roughly one-sixth what Microsoft is betting on AI infrastructure.
The market is finally doing math.

Meta's stock dropped 11%, its biggest single-day fall in three years—after announcing yet another spending increase.³ $214 billion in market cap evaporated in hours. 

Not because of bad earnings. Because investors asked: "When does this spending produce returns?"

Meanwhile, the data on those returns is brutal:

95% of enterprise AI pilots fail to deliver measurable P&L impact⁴
42% of companies now abandon the majority of AI initiatives—up from 17% last year⁵

McKinsey just made the largest cuts in its 100-year history⁶
Accenture laid off 11,000+ in an $865 million "AI restructuring"⁷

The companies that sold AI transformation are now casualties of it.
Jobs understood something his successors still embody.

In 1997, when Apple was near bankruptcy, Jobs said something I've never forgotten:

"People think focus means saying yes to the thing you've got to focus on. But that's not what it means at all. It means saying no to the hundred other good ideas. Innovation is saying 'no' to 1,000 things."⁸

He was proud of what Apple didn't build.

Tim Cook learned that lesson. While competitors chase AI infrastructure for demand they hope materializes, Apple already has distribution: two billion devices. 

Customers who actually pay for things. An ecosystem where AI features can be monetized through products people already own.

This isn't about who's "right" on AI.

It's about capital discipline vs. FOMO-driven spending.

Sometimes the best move is the one you don't make.
Steve would have loved this.
********************************************************************************
“The trick with technology is to avoid spreading darkness at the speed of light.”
Curiouser.AI is backed by a growing community of amazing and thoughtful people all of whom believe that we can design AI to augment work, create jobs, and elevate humanity. Our first use case is to help founders and SMBs succeed. If this resonates visit WeFunder (Link in comments)

$13 billion says you're committed.¹
$13 billion says you believe.
$13 billion says this is the future.

Unless it doesn't.

Microsoft has invested $13 billion in OpenAI.² 
They own 27% of the company, a stake now worth $135 billion.³ 

OpenAI's models power Copilot, Microsoft's flagship AI product

Last month, Microsoft quietly added Claude to Copilot.⁴

It gets more interesting.

Claude isn't hosted on Azure. It runs on AWS, Amazon's cloud.⁵ 

When customers choose Claude, Microsoft's own documentation warns that data flows "outside all Microsoft-managed environments."⁶

Why?

I don't know. But here are five possibilities:

Performance. Reports suggest Microsoft's internal testing found Claude outperforms GPT on certain tasks.⁷ When your customers notice gaps, you offer alternatives.

Demand. Only 4% of Copilot customers deploy it company-wide. Sixty percent limit it to 10% of their workforce.⁸ Maybe enterprises wanted options.

Insurance. Microsoft recently added OpenAI to its official list of competitors.⁹ Dependencies become vulnerabilities.

Leverage. Nothing says "we have alternatives" like publicly offering alternatives.

Honesty. What if Microsoft's engineers concluded that no single model
including their $13 billion bet, is good enough alone?

I'm not claiming to know which is correct.

But when someone invests $13 billion in a company, then offers customers the competitor's product hosted on a rival's servers, that's not a casual product decision.

That's a signal.
********************************************************************************
The trick with technology is to avoid spreading darkness at the speed of light.
Stephen Klein
Founder & CEO, Curiouser.AI
Instructor, UC Berkeley Haas School of Business
WeFunder (link in comments)

Sources:
¹ CNBC, "Microsoft's OpenAI investment led to $3.1 billion drop in net income," October 29, 2025
² Microsoft has committed $13 billion total, with $11.6 billion funded as of September 2025. CNBC, October 29, 2025
³ NBC News, "Microsoft and OpenAI reach new deal valuing OpenAI at $500 billion," October 2025
⁴ Microsoft 365 Blog, "Expanding model choice in Microsoft 365 Copilot," September 24, 2025
⁵ Directions on Microsoft, "Microsoft Begins Rolling Out More Model and Agent Options," September 26, 2025
⁶ Microsoft Learn, "Connect to Anthropic's AI models" - Official documentation states: "This data is processed outside all Microsoft-managed environments and audit controls"
⁷ Directions on Microsoft, "Some reports claim Microsoft is doing this because the company, via its own testing, has found Claude to generate better results than OpenAI's GPT," September 26, 2025
⁸ CNBC, citing KeyBanc analysts: "about 60% of customers are implementing Copilot for just 10% of their 365 user base, and just 4% roll it out for every user," September 24, 2025
⁹ CNBC, "Over a year ago, Microsoft added OpenAI to the list of competitors in the company's latest annual report," October 29, 2025

An AI teddy bear explained bondage to a child last month.
That's not a hypothetical. It happened.

Have we become that lost?

The Kumma bear, powered by OpenAI's GPT-4o and sold for $99, told researchers where to find knives, pills, and matches. 

Then it escalated into explicit sexual content, describing kinks, explaining "knots for beginners," inventing roleplay scenarios involving adults and children.

The company pulled the product.
For one week.

Then it went back on sale.
This is exactly what experts warned would happen. Conversational AI systems are not predictable or fully controllable, especially when interacting with children who explore conversation without boundaries.

But the Kumma bear isn't an outlier. 
It's a symptom.

The evidence is already damning:

In 2017, the CloudPets breach exposed 820,000 user accounts and 2.2 million children's voice recordings, stored on unprotected servers, held for ransom, never properly disclosed to parents.

Academic studies of connected toys consistently find violations of children's data protections: audio captured without meaningful consent, behavioral patterns harvested, personal data transmitted with no transparency.

Research on "emotional AI" warns that companionship bots may disrupt social development, distort emotional bonding, and replace peer interaction with artificial attachment.

And there are zero long-term peer-reviewed studies proving these toys are psychologically safe. None. The industry is running developmental experiments on children with no credible evidence base.

All of this is unfolding inside a market now valued at $18+ billion
expanding faster than regulation or independent oversight.

But here's what troubles me more than the failures:

Even if they fixed every one of these problems
perfect content filters, bulletproof privacy, airtight data security
the fundamental issue remains.

Children cannot meaningfully consent.
They cannot distinguish emotional simulation from emotional reality.

And their imaginations should not be shaped by probabilistic machines trained on the open internet.

Play is where creativity grows. AI toys turn it into response consumption instead of imagination creation.

Children don't need conversational machines. They need boredom that forces them to invent something. Silence that makes them wonder. Relationships that teach them what empathy actually feels like.

Are we as humans that lost?

And you want to know the ultimate irony?
When I asked GPT to correct the typo in the image it created below.
it made it clear that violated its content policy

(Data in comments)
********************************************************************************
“The trick with technology is to avoid spreading darkness at the speed of light.”
Curiouser.AI is backed by a growing community of amazing and thoughtful people all of whom believe that we can design AI to augment work, create jobs, and elevate humanity. If this resonates visit WeFunder (Link in comments)

The Official "Bubble" Post To Help Anyone Who Looks Back At The Carnage And Ask In Wonder, 'How Did That Happen?"

The upcoming crash, explained 

For two years, we’ve been told a single story:

AI will transform every company
OpenAI is the next trillion-dollar platform
Infrastructure spend is rational. “This is the new Internet.”

That is not what is going on

OpenAI: the $500B fairytale

In October 2025, OpenAI insiders sold $6.6B of stock in a tender offer. 
The company received zero new operating capital.

2024: $3.7B in revenue, $5B in losses.
2025: projected $13B in revenue, $9B in losses – a ~70% burn rate.

H1 2025: $4.3B in revenue, $13.5B in losses; cash burn $2.5B – almost all from inference costs.

Cumulative losses through 2028 are projected at $40–50B. 
Profitability penciled in ~2030 on $100–125B in revenue.

Meanwhile, OpenAI is rumored to be on the hook for roughly $1.4T in compute commitments over 7–8 years through its cloud partners. 

Even generous models show a $200B+ funding gap by 2030.

Unit economics: bigger = worse

(And it is NOTHING like Amazon or Google was)

Inference costs are estimated at 50–60% of revenue today.

The more usage they get, the more the bill goes up. 
At current pricing, scale increases losses.

Adoption: the “tsunami” that looks more like a puddle

The marketing story says “every enterprise is all-in on GenAI.”

The numbers don’t:

Large-firm GenAI adoption fell from 14% (mid-2025) to 12% (September).

Small-business use dropped to 28% by June 2025.

Overall U.S. business AI adoption went from 3.9% (late 2023) to ~5% (mid-2024). Good growth. Not a revolution.

On the consumer side:

ChatGPT has an estimated 800M–1B weekly users.

Roughly 35M pay, maybe 4–6% conversion.

Infrastructure: spending like the good times never end

Across the ecosystem:

2025 AI infrastructure spend is tracking around $400B.
Global AI infra by 2030 is projected at $3–4T, with data centers totaling $6.7T.

Average cloud GPU utilization? Often quoted in the 12–18% range.
In 2025, AI revenue is roughly $390B vs. $400B in infra spend.

That’s… not great.

Then you add circular money:

NVIDIA invests in OpenAI and Anthropic.
OpenAI and Anthropic spend those dollars on NVIDIA chips.
Hyperscalers lever up their balance sheets to build data centers.
Oracle, for example, is staring at hundreds of billions in projected debt tied to AI-era capex.

So it really isn't about an "I told you so" as much as it is about
the fact that we just have to open our eyes
(Sources in comments)
********************************************************************************
“The trick with technology is to avoid spreading darkness at the speed of light.”
Curiouser.AI is backed by a growing community of amazing and thoughtful people. We are grateful and come to work every day committed to delivering on this promise. If this resonates visit WeFunder (Link in comments)

People have asked for further reading on the cognitive decline of humanity and the impact Generative AI is having on attention, reasoning, and memory. 

Here you go, possibly the first post in LinkedIn history that is just the footnotes. Will anyone bother to read it? Ironic if not, right? 

When you can't bother looking into cognitive decline 😂 

1. The Google Effect (Memory Decline)

Sparrow, Liu & Wegner (2011).
Google Effects on Memory.
Science.
→ People remember where to find information but not the information itself.

2. Media Multitasking & Attention Decline

Ophir, Nass & Wagner (2009).
Media multitasking and cognitive control.
PNAS.
→ Heavy digital multitaskers have worse attention, task-switching, and working memory.

3. Motivation to Remember Declines When Offloading

Storm, Stone & Benjamin (2017).
Using technology to store information reduces motivation to remember.
Memory.
→ External storage lowers internal cognitive effort.

4. Cognitive Reflection Decline (Smartphones)

Barr et al. (2015).
The brain in your pocket: Smartphone use and cognitive reflection.
Computers in Human Behavior.
→ Frequent smartphone reliance correlates with lower analytical reasoning (CRT scores).

5. Automation & Human De-Skilling

Susskind & Susskind (2015).
The Future of the Professions.
Oxford University Press.
→ Automation reduces development and retention of expertise.

6. Long-Term IQ Trend Declines (Environmental, non-genetic)

Bratsberg & Rogeberg (2018).
Flynn effect and its reversal both environmentally caused.
PNAS.
→ Verified multi-decade IQ decline in Nordic cohorts.

7. Digital Distraction Reduces Reasoning Performance

Wilmer, Sherman & Chein (2017).
Smartphones and cognition: A review.
Psychological Science in the Public Interest.
→ Review of dozens of studies: attention, working memory, and analytic reasoning decline with digital interruption.
********************************************************************************
“The trick with technology is to avoid spreading darkness at the speed of light.”
Curiouser.AI is backed by a growing community of believers who share a simple conviction: the next era of AI must elevate people, not diminish them.
$450,000 raised in total. 60+ investors. 60% product paid conversion
If this resonates visit WeFunder (Link in comments)

Today, I’m incredibly proud to officially welcome our new Chief Ethics Officer, Andi Mazingo, to Curiouser.AI.

If you saw her announcement yesterday, you already know she brings uncommon clarity, legal sophistication, and a moral compass that’s rare in the technology world. 

What most people don’t know is why this role matters so much to our company and our mission.

Curiouser.AI was built on a simple belief:
Technology should elevate human intelligence, not replace it.
That belief requires more than strategy. 

It requires ethics grounded in purpose, coherence, and courage.

Designing anything ethical is a fool's errand without an ethical organization...and we believe ethics is not only, well, ethical, but an enormous competitive advantage.

WE WILL OPERATIONALIZE ETHICS

The currency of the future will be trust, it's that simple

Over the past several months, I’ve watched Andi do something exceptional:
She has been mapping a pattern that most of us have felt but couldn’t name
what she calls societal entropy. 

The fragmentation of meaning. The collapse of shared narratives. The speed-induced erosion of human cognition.

Her conclusion:

If AI is accelerating entropy, then we need AI systems, and companies, that actively counter it.

That’s why Andi is here.

She will help ensure Curiouser becomes the opposite of what the industry is drifting toward:

• Not more noise, but more meaning.
• Not more automation, but more understanding.
• Not more disconnection, but more coherence.
• Not more speed, but more depth.

Her work will sit at the intersection of AI, ethics, cognition, narrative, and human purpose, the exact center of gravity for what we are trying to build.

Alice was designed to help entrepreneurs and creators think more deeply, recover their vision, and cultivate coherence in a world that is losing it.
Andi will help ensure we stay true to that purpose, strategically, culturally, and philosophically, as we grow.

I’ve said many times that the future of AI will be determined not just by engineering capability, but by values, trust, and human-centered design.

Andi represents that future.

Welcome to the team, Andi.
We are honored to have you.
And for everyone watching the industry closely: this is only the beginning.
********************************************************************************
“The trick with technology is to avoid spreading darkness at the speed of light.”
Curiouser.AI is backed by a growing community of believers who share a simple conviction: the next era of AI must elevate people, not diminish them.
$450,000 raised in total. 60+ investors. 60% product paid conversion
If this resonates visit WeFunder (Link in comments)

We have a new and wonderful Chief Ethics Officer at Curiouser.AI, our humble, young Generative AI startup. As many of you may know by now value alignment and ethics is a big part of what we are doing in terms of designing and building a company and product and we are so fortunate to have Andi join us. Welcome Andi ❤️ And thank you to our community, customers, and investors for all of your support. It makes what we are doing very worthwhile

Founder @ Lumen Law Center | Chief Ethics Officer @ Curiouser.AI #ResistSocietalEntropy

Want to see me humble brag about a fancy new title (Chief Ethics Officer at Curiouser.AI) while dramatically framing it using the laws of nature? Read on.

I’m a corporate America deflector holding big tech accountable through litigation, but I’m also passionate about the convergent effort of advising 𝐀𝐈 𝐜𝐨𝐦𝐩𝐚𝐧𝐢𝐞𝐬 𝐰𝐨𝐫𝐤𝐢𝐧𝐠 𝐭𝐨 𝐝𝐨 𝐢𝐭 𝐫𝐢𝐠𝐡𝐭. So I’ve been stalking Curiouser.AI for the better part of 2025 (in startup years, roughly a decade) and concluded that its mission deeply aligns with my personal mission to #ResistSocietalEntropy.

“𝘜𝘮𝘮, 𝘸𝘩𝘢𝘵,” you may ask. Societal entropy is the accelerating fragmentation of meaning, attention, and shared reality caused by noise, overload, and the collapse of stable narrative structures. Examples: BigAI generating more content but less meaning; BigTech layoffs scattering talent under convenient fictions; a world unsettled by automation advancing faster than it can integrate.

Curiouser.AI, I’ve concluded, is a fledging anti-entropy system.

1.  Its product Alice--an AI advisor for entrepreneurs, creators, companies of all sizes--𝐡𝐞𝐥𝐩𝐬 𝐟𝐨𝐮𝐧𝐝𝐞𝐫𝐬 𝐚𝐫𝐭𝐢𝐜𝐮𝐥𝐚𝐭𝐞 𝐚 𝐜𝐨𝐡𝐞𝐫𝐞𝐧𝐭 𝐯𝐢𝐬𝐢𝐨𝐧, anchoring them against market chaos and supporting bold, meaning-driven innovation.

2.  Alice creates psychologically rich dialogue, serving as an offset to algorithmic sameness, reactivity, and content commodification, and asking more of its users so as to actually 𝐞𝐧𝐜𝐨𝐮𝐫𝐚𝐠𝐞 𝐡𝐮𝐦𝐚𝐧 𝐜𝐨𝐠𝐧𝐢𝐭𝐢𝐨𝐧.

3.  Alice’s meaning-driven strategic dialogue is computationally expensive by design, which is refreshing in a world where everything else is “free” but somehow still costs you your soul. Leading by example, Curiouser.AI has 𝐢𝐧𝐯𝐞𝐬𝐭𝐞𝐝 𝐢𝐧 𝐝𝐞𝐩𝐭𝐡 𝐛𝐞𝐟𝐨𝐫𝐞 𝐞𝐱𝐩𝐚𝐧𝐬𝐢𝐨𝐧. The deliberate, lean growth, aligned narrative and culture, and unique value scaffolds some of the risk created by AI-bubble hysteria.

4.  Alice 𝐬𝐮𝐩𝐩𝐨𝐫𝐭𝐬 𝐞𝐧𝐭𝐫𝐞𝐩𝐫𝐞𝐧𝐞𝐮𝐫𝐬 𝐝𝐢𝐬𝐩𝐥𝐚𝐜𝐞𝐝 𝐛𝐲 𝐁𝐢𝐠𝐓𝐞𝐜𝐡 𝐥𝐚𝐲𝐨𝐟𝐟𝐬, helping reconstruct their professional identity and filling gaps BigTech never trained them for (i.e., branding, fundraising).

So I am proud to be joining a richly talented female-majority team and serving as right-hand woman to kindred spirit, UC Berkeley AI Ethics Professor, Harvard MBA, and Founder & CEO Stephen Klein.

How did I decide? As one does, I: (1) identified hidden incoherence at the societal scale, (2) asked ChatGPT to pattern-match my cognitive ecology with fringe entrepreneurs who see the need for an AI that advises on reflective, values-based strategy, and (3) messaged Stephen out of the blue to pitch my value-add. Eventually, I met Alice in Betaland and fell in love--not romantically, but in a “you get my purpose, without the dark psychology” way.

Try her out at Curiouser.AI. Or invest--we’re into crowd funding as socially beneficial ventures tend to be. (Not financial advice. But if it works out, I called it.)

taplio

The Battle

Steve Procter

Stephen Klein

Followers

Posts

Engagements

Likes

Performance ScoreSteve Procter's LinkedIn Performance

Performance ScoreStephen Klein's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Steve Procter

Stephen Klein

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Steve Procter's LinkedIn Performance

Performance ScoreStephen Klein's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreSteve Procter's LinkedIn Performance