🍭 OpenAI Designed Their Own Chip 🤯

Good morning. Somewhere in San Francisco, a server just became self-aware enough to request PTO.

Let’s dig in.

🍭 What’s Cookin’:

• Be Boring, Get Rich

• OpenAI x Broadcom

• 5 AI Tools to Steal Back Your Workday

• Everything Else: Google’s $15B India hub, Walmart in ChatGPT, NVIDIA’s mini AI rig, and more

_ToolBox™
⚒️ 5 AI Tools to Steal Back Your Workday

⏱️ DoneThat
→ Privacy-first AI time tracker and work coach that auto-logs your work, sets realistic goals, and supports teams.

🍓 Strawberry
→ A self-driving browser with AI companions that navigate sites and complete multi-step workflows for you.

📧 Mailmodo AI
→ Run end-to-end email campaigns from a single prompt—strategy, copy, design, automations, and reports.

🧩 Jotform Salesforce Agent
→ An AI chatbot inside Salesforce that turns chat into action—create/update records, qualify leads, and manage cases.

⚡ n8n AI Workflow Builder
→ Describe what you want, get a draft automation or AI agent you can refine; rolling out on n8n Cloud.

_{OpenAI <> Broadcom}
🎬 AI That Designs It’s Own Chips: The Flywheel

Moore’s Law on Steroids

The Bite: OpenAI and Broadcom just signed a long-term deal to co-develop custom AI chips.

But the twist is who’s doing the designing.

OpenAI is using its own AI models to help engineer the next generation of accelerators, optimizing layouts, power, and timing.

The chips will then train smarter versions of those same models, completing what insiders are calling an “AI hardware flywheel.”

Snacks:

• AI helps design chips. OpenAI’s models are used in early-stage design — floor planning, thermal layout, and block optimization.

• Broadcom builds at scale. The partnership covers up to 10 gigawatts of AI racks rolling out between 2026–2029.

• Full-stack ambition. This is about owning the stack — from training models to controlling the silicon they run on.

• Faster R&D loop. AI accelerates chip design → better chips train better AI → repeat.

• Energy remains the ceiling. Even with smarter chips, every watt still matters.

Why it bites: This deal is more about time and speed than hardware.

If OpenAI can cut chip R&D cycles from years to months, it rewrites the speed limit of AI progress.

Every improvement feeds back into the next one, tightening the loop between intelligence and infrastructure.

Still, physics hasn’t gone away.

Power, heat, and cost will decide how far this flywheel spins before it hits resistance.

But for now, OpenAI’s not waiting for Moore’s Law.
It’s rebuilding it in-house.

Steal This Prompt
📚 Use the 80/20 Principle to Learn Faster Than Ever

Find the critical 20% of any topic so you understand 80% of it—fast, focused, and practical.

Optional chaos-free flexes: learn Python over a weekend; brief your boss on quantum by lunch.

1. Click this link (Prompt)

2. Paste prompt into your favorite AI

3. Replace the #’s with your topic, time box, and output format

4. Watch it cook — study notes appear while your coffee’s still hot

   
   🍭 Try the prompt on Snack Prompt 🍭

_{Everything Else}
🧠 You Need to Know

🛰️ Google to Invest $15B in India AI Hub

→ Google will build a 1-GW data center and AI hub in Visakhapatnam through 2030, with plans to scale and add subsea connectivity.

🛒 Walmart Shopping Comes to ChatGPT

→ A new partnership lets customers link Walmart accounts and buy items directly in ChatGPT, with a rollout expected later this fall.

💄 Google Meet Adds AI-Powered Makeup

→ Meet now offers 12 virtual makeup options under “Portrait touch-up,” rolling out on web and mobile; it’s off by default.

📆 Gemini Can Schedule Google Calendar Meetings

→ Gmail’s “Help me schedule” suggests times from your availability for one-on-one meetings and inserts them into your email.

🖥️ NVIDIA Sells $3,999 DGX Spark AI PC

→ The 2.6-lb mini system includes a GB10 SoC, 128GB RAM, 4TB storage, and NVIDIA’s full AI stack; DGX Station with GB300 is coming next.

Cisco AI Readiness Index
🤑 Be Boring, Get Rich: The Unsexy Truth Behind AI Winners

The Bite: Cisco just dropped its 2025 AI Readiness Index, and the results are humbling:
Most companies talk about AI, but very few are actually ready for it.

Out of thousands of organizations studied, only 13%, called “Pacesetters”,
are turning AI into real business results.

These Pacesetters aren’t the loud ones posting flashy demos.

They’re the ones doing the slow, steady work:
Cleaning their data, building scalable infrastructure, assigning ownership, training people, and tracking results.

Cisco calls it the “difference between chasing AI and capturing value.”

Snacks:

• The 13% Club: Pacesetters have stayed around 13–14% for three years — small group, big results.

• Big Plans, Small Pipes: 83% of companies plan to launch AI agents, but only 34% have infrastructure that can support them.

• Money Over Hype: Pacesetters are three times more likely to measure their AI impact (95% vs 32%).

• Proof Beats Promise: 77% of Pacesetters finalized use cases; most others are still experimenting.

• Profits Follow Discipline: They report stronger revenue, productivity, and innovation across the board.

• New Risk Alert: Cisco introduced the term AI Infrastructure Debt — the cost of ignoring upgrades and underfunding systems that AI depends on.

Why it Bites: The companies seeing the best returns treat AI as part of how they operate, not a side project.

They plan ahead, build solid systems, and track what actually works.

For builders: slow down before you ship the next “AI agent.”

Make sure your data, infrastructure, and teams can handle it.

The companies that invest in the boring stuff today will be the ones making real money tomorrow.

Boring wins.