Welcome to your weekly Brew & AI
Each week, I’ll break down AI, the headlines, and the social buzz — no jargon, no hype, just what’s worth your attention.
P.S. - To ensure you receive this newsletter, please move it to the primary tab and set it for all future communications. Or just add this email as a known contact. I wish AI could solve this for me.
Grab your coffee - let’s dive in. 👇
☕ AI in the news
This week’s biggest AI power plays, leaks, and infrastructure moves, and why they actually matter beyond the headlines
Meta's Superintelligence Team Just Showed Its Hand
Meta quietly dropped Muse Spark, the first model out of its newly formed superintelligence unit. It's less a product launch and more a signal: Meta is no longer just playing catch-up.
With a dedicated team focused on AGI-level ambitions, this is the clearest sign yet that the race has moved beyond labs and into organizational structure.
The next 18 months just got more interesting. And with Zuckerberg personally steering the AI bet, don't expect Meta to stay quiet for long.
ChatGPT Is Becoming an Ad Business. That Should Surprise No One.
OpenAI is projecting $2.5 billion in ad revenue this year, scaling to $100 billion by 2030.
The product that started as an AI assistant is quietly becoming one of the largest ad platforms ever built. If those numbers hold, ChatGPT won't just be a ChatGPT product, it'll be infrastructure for how brands reach people in an AI-first world.
The bigger question: how does ad-driven incentive change what the model actually shows you?
Anthropic Just Passed OpenAI in Revenue. Enterprise Won It for Them.
Anthropic hit a $30B annualized revenue run rate, overtaking OpenAI's $24B.
The driver? Enterprise.
While OpenAI chased consumer scale and ad dollars, Anthropic went deep on B2B deals, and it's paying off. The quiet player in the AI race just became the one to watch on the business side.
It also reframes the whole narrative: in AI, safety positioning and commercial success are not mutually exclusive.
☕ Trending on social
The posts everyone in AI is screenshotting, subtweeting, and arguing about this week; distilled for your sanity.
😄 Someone built a literal digital whip - yes, a whip that cracks every time Claude pauses mid-response, as a joke tool to "motivate" the AI to work faster, and 14 million people found it deeply relatable.
It went massively viral because, apparently, we've all been sitting there silently judging our AI for taking too long, and this person just decided to do something about it.
It's absurd, it's petty, and it's a perfect snapshot of where we are right now: so dependent on these tools that we're now inventing ways to boss them around.
🔥 Anthropic just dropped a new model called Mythos under Project Glasswing, and almost nobody can use it yet.
Access is currently limited to FAANG companies and cybersecurity firms, which has the internet simultaneously impressed and deeply jealous, especially after early testers started hinting that nothing else comes close.
The exclusivity is doing exactly what exclusivity always does: making everyone who doesn't have access want it even more, while sparking a very loud debate about whether keeping frontier AI locked behind a velvet rope is safety-conscious or just good marketing.
🤔 Andrej Karpathy, one of the most respected names in AI, formerly of OpenAI and Tesla - dropped a detailed thread on how to use AI to build your own personal knowledge base, essentially a searchable second brain made from your notes, documents, and ideas.
It blew up because Karpathy has a rare gift for making complex technical ideas feel approachable, and people are hungry for practical ways to actually use AI in their daily lives rather than just marvel at it.
☕ AI tool of the week
A lesser‑known gem, handpicked for its uniquely powerful edge.
Clicky is an AI-powered cursor that sits right on your screen and answers questions about whatever you're looking at - no switching tabs or copying text into a chatbot required. Just ask it out loud and it'll explain what's in front of you, whether that's a confusing dashboard, an unfamiliar app, or a document full of jargon. It's a surprisingly natural way to get help without breaking your flow.
Perfect for anyone learning a new tool or software, Clicky removes the frustration of not knowing where to click or what something means by letting you simply ask your screen.
☕ AI workflow of the week
Plug‑and‑play workflows you can steal in under an hour to make AI actually save you time and money, not just impress you.
Turn any repeatable workflow into a reusable AI skill your whole team can run consistently
Who this is for: Product teams, operators, founders, marketers, and internal tool builders
Problem: Teams keep solving the same tasks by hand, which wastes time and produces inconsistent results. Without a reusable structure, every person reinvents the wheel - and the output quality depends entirely on who's doing it that day.
How to run it (30–45 minutes)
Step 1 — Gather your workflow's raw material
Pick one high-frequency task your team does repeatedly — something with clear inputs (what goes in) and clear outputs (what comes out). Write down the steps you already follow, the edge cases that trip people up, and what 'good' looks like when it's done right. One page of rough notes is enough to start.
Step 2 — Prompt Claude to draft your SKILL.md
A SKILL.md is a plain-text instruction file that tells the AI exactly how to run your workflow — every time, the same way. Ask Claude to turn your notes into a structured skill file with a purpose statement, a 'when to use this' section, numbered step-by-step behavior, and a short acceptance checklist so you know when the output is good enough.
Step 3 — Refine by stress-testing on one hard real example
Don't test on an easy case — find the messiest, most ambiguous real example of this task you've seen in the last month. Paste it in alongside your SKILL.md and run the workflow. Where the output breaks down or feels off, go back and tighten the instructions.
Step 4 — Package it and put it to work
Save your final SKILL.md as a file you can upload at the start of any Claude conversation — or paste it into a shared team doc so colleagues can copy and use it instantly. Run it on two or three more real cases, note where it still slips, and update the file after each failure. ☕ Weekly Reads
Hand‑picked long‑reads that go deeper than the feed - strategy, markets, and how AI is quietly rewiring real industries.
☕ Resources
If you missed my writing, here’s everything I’ve done - bite‑sized, practical, and very AI‑obsessed. Think of this as the “in case you blinked” shelf for my recent posts, threads, and experiments.
✍ Blogs - In-depth articles and insights to expand your AI knowledge
🛠 Tools - Discover powerful AI tools to enhance your workflow
💭 Tips - Practical tips and tricks to make your AI journey smoother
💛 P.S.
That’s it for this week’s brew.
I’d love to hear what you think - what you liked, what could be better, or what you’d love to see next.
Just hit reply - I read every message over my morning coffee ☕.
Brew & AI
Making AI simple, one sip at a time
1