Donald Knuth is 88 years old. he's considered the father of algorithm analysis. wrote the most important computer science textbook series in history.
earlier this month he published a paper titled "claude's cycles." opens with the words "shock! shock!"
what happened: claude opus 4.6 solved an open graph theory problem that knuth had been working on for weeks.
a problem in a field knuth helped invent. solved by an AI. and the man himself called it "a dramatic advance in automatic deduction and creative problem solving."
when the greatest computer scientist alive publishes a paper because AI beat him at his own game, we're in a different era.
this story got almost no mainstream coverage. it should have been front page everywhere.
the best time to start building with AI was 6 months ago.
the second best time is this weekend.
pick one task. build one system. see what happens.
if you need ideas, we drop fresh workflows every tuesday and thursday:
https://t.co/9ED5tBloo1
have a good weekend. go build something.
this week in AI is INSANE.
the numbers:
$1T. nvidia's demand forecast through 2027 (doubled from last year)
$19B. anthropic's annualized revenue
$57B. projected AI ad spend for 2026
75%. GPT-5.4's score on human productivity benchmark (humans: 72.4%)
24,000. fake accounts chinese labs made to reverse-engineer claude
15,000. potential meta layoffs to fund AI infrastructure
$2.2T. norway's sovereign fund now using claude for ESG screening
$2.5B. claude code annual revenue (launched 14 months ago)
every number is a signal. the question is what you're building with that information.
block (the company behind cash app and square) just cut 40% of its team.
that's roughly 4,000 people.
their CEO didn't sugarcoat it. said those roles were made unnecessary by AI tools.
not "restructuring." not "strategic realignment." AI made those jobs obsolete. his words.
same week oracle announced plans to cut up to 30,000 employees to redirect $8-10 billion toward AI infrastructure.
two companies. 34,000 jobs. one reason.
the pattern is getting harder to ignore. the companies building AI are hiring. the companies using AI are firing the people AI replaced.
which side of that equation are you on?
the real AI moat isn't the tool you use.
it's the system you build around it.
tools change every 6 months. the model that's best today won't be best next year.
but a system that connects your voice memos to your task manager to your calendar to your email? that transfers to any model.
invest in architecture, not brand loyalty.
build systems that are model-agnostic. the plumbing matters more than the water.
friday prompt that takes 4 minutes:
"here's what i accomplished this week: [paste your done list]. tell me:
what patterns do you see? am i spending time on the right things?
what should i do MORE of next week?
what should i STOP doing?
give me a brutally honest productivity score 1-10"
the score stings sometimes. that's the point.
most productivity systems track WHAT you did. this one tells you WHETHER you should keep doing it.
run it every friday. watch your focus sharpen week over week.
5 things i learned this week:
1. chatgpt ads are struggling (0.91% CTR vs 6% benchmark). attention doesn't equal ad revenue.
2. claude code channels turned "sit at desk" coding into "message from phone" coding. game changer.
3. norway's $2.2T sovereign fund uses claude for ESG screening. if they're using AI, your company should be.
4. anthropic walked away from pentagon contracts over ethics. respect that or disagree. but it's historic.
5. the AI writing study proved what i suspected: AI makes everyone sound the same. voice is now the moat.
what did you learn this week?
heads up: claude's doubled usage limits end tonight at 11:59 PM pacific.
if you haven't used the extra capacity this week, today and this weekend are your last shot.
one thing to try: take your most annoying weekly task. describe it to claude. ask for a system.
worst case: you waste 15 minutes.
best case: you never do that task manually again.
openai just picked AWS as its exclusive cloud provider for its enterprise platform.
not azure. not google cloud. amazon.
jensen huang said this will drive "enormous consumption" of cloud computing.
here's the bigger picture: the company that makes the most popular AI model chose the cloud that already runs 31% of the internet. that's not a partnership. that's infrastructure consolidation.
the AI stack is settling. the winners are locking in. if you're building workflows on top of these tools, you're building on a foundation that isn't going anywhere.
the question isn't which AI company wins. it's whether you're building on any of them.
6 months ago i didn't know what n8n was.
3 months ago i'd never built an API integration.
today i have an automated pipeline that turns voice recordings into structured data across 5 different apps.
i'm not a developer. i'm not technical. i'm a navy vet who runs businesses.
the tools are that accessible now. you just have to start.
every expert was once a beginner who decided not to quit on day 3.
everyone talks about zapier. almost nobody talks about n8n.
the difference:
zapier charges per task. n8n cloud starts at $24/month for unlimited.
at high volume, zapier gets expensive fast. n8n stays flat.
plus n8n lets you self-host for free if you want full control.
i run my entire meeting-to-notion automation through n8n. costs me pennies.
it's more powerful than zapier, cheaper at scale, and the learning curve is about a weekend.
if you're paying $50+/month for zapier and running more than a few workflows, do the math. you might be surprised.
google just rebuilt workspace with AI baked into everything.
docs: "help me create" generates first drafts from your existing files and emails.
sheets: "fill with gemini" populates spreadsheets with real-time web data.
slides: generates full decks from your existing documents.
maps: gemini-powered search answers complex location questions.
gemini in sheets scored 70.48% on the SpreadsheetBench dataset. that's a new state-of-the-art for AI spreadsheet automation.
AI is moving from a destination (separate app you visit) to a feature (inside every tool you already use).
your coworkers will be using AI whether your company has a strategy for it or not.
norway's $2.2 trillion sovereign wealth fund is using claude to screen for ESG risks.
that's the largest investment fund on the planet using AI to detect forced labor, corruption, and supply chain issues.
this is what AI adoption looks like at scale. not chatbots. not content generation.
risk analysis across a $2.2 trillion portfolio.
the enterprise AI use cases that matter most aren't the ones that make headlines. they're the ones that quietly replace entire departments of analysts.
stuck on a decision? try this:
paste this into claude:
"i'm deciding between [option A] and [option B]. here's the context: [your situation]. build me a decision matrix comparing both across these dimensions: cost, time, risk, upside potential, and what i'd regret most in 6 months. then give me your honest recommendation and why."
i use this for everything. hiring decisions. tool choices. whether to take on a project.
it doesn't make the decision for you. it organizes your thinking so the right answer becomes obvious.
most "hard decisions" are just poorly organized information.
claude is the best research assistant i've ever had.
here's how i use it:
"i'm writing about [topic]. find the strongest arguments for and against. include specific numbers, studies, or examples. organize them so i can see both sides clearly. then tell me which angle has the most data behind it."
one prompt. 30 seconds to type. 5 minutes to review.
replaces an hour of googling, tab-hopping, and trying to remember where you saw that one stat.
the quality of your output is only as good as the quality of your research. let AI handle the gathering. you handle the thinking.
first ever thursday issue of stacked drops in 2 hours.
i'm calling it THE BUILD. you walk in. you build something. you walk out.
two claude workflows today:
one that auto-triages your email into notion tasks. 30 min setup.
one that turns a 5 min voice memo into 5 pieces of content. 45 min setup.
no code. no terminal. grandma-level walkthroughs.
visa is building infrastructure for AI agents to make payments.
not people using AI to shop. AI agents shopping on their own. initiating transactions, handling authentication, making purchasing decisions within rules you set.
they're testing it right now for automated procurement and recurring purchases.
think about that for a second. the financial system is being rebuilt to process transactions where no human is involved at any step.
we went from "AI writes my emails" to "AI buys things with my money" in about 18 months.
the companies building for this future are the ones that will own the next decade.
the richest people in the world all have one thing in common:
they build systems that work without them.
AI just made that accessible to anyone with a laptop and an internet connection.
you don't need employees. you don't need investors. you don't need a CS degree.
you need one workflow that saves you 2 hours a week. then another. then another.
stack enough of them and you've built something that used to require a team of 10.
that's not hype. that's math.