💡 The $500M Deal That Quietly Merged Security and IT

Sponsored by

☕ Morning! 💡 Your Weekly 5-Minutes of Caffeine and Tech Clarity

Quick Hits 🎯

• 🧠 Chipmakers are redesigning memory and storage around real-time inference, not training runs

• 🔧 The "AI transformation" that sticks is 20% new tools, 80% process redesign

• 🟥 Microsoft 365's outage stretched into a multi-day reliability story (and a governance lesson)

• 🛰 Ukraine is selling battlefield drone data, and it's quietly reshaping the AI data market

• 🧭 OpenAI's Greg Brockman on Astra, and why alignment is now a product decision

• 📉 New research: startup ARR is less durable than the SaaS playbook assumed

• 🛡 Palo Alto Networks paid $500M for Console to fuse security with IT automation

• 🚕 Travis Kalanick's Atoms may be pivoting toward robotaxis

🎁 + 4 other stories you might find useful

Our Partner 🎉 

285 Free Prompts to Save Time

Your inbox isn't the problem. Neither are your meetings. It's the hours you spend on work AI could do in minutes.

285 copy-paste prompts: built for email, meetings, admin, and delegation. Each one has a clear role, a specific task, and a defined output.

No setup. No prompt engineering. Just copy, paste, and get your time back. Free when you subscribe to Mindstream.

Get 285 free prompts

The Big Picture 🖼️

💡 Inference Is Becoming a Memory Problem, Not a Compute Problem.

Everyone's still counting GPUs, but the bottleneck for real-time AI is shifting to how fast you can move and hold state. Latency-optimized memory tiers and high-bandwidth interconnects are turning into the actual backbone of production inference.

The counterintuitive part: the fix for a memory wall isn't always more memory.

The takeaway: if your AI roadmap only budgets for compute, you've mispriced it.

💡 Tools Don't Transform Companies. Process Does.

AI can augment specialists and speed up workflows, but it can't supply the discipline that decides how choices get made and how outputs get audited. The teams pulling ahead pair the model with a repeatable process around it.

That gap is exactly what Benedict Evans keeps pointing at.

The takeaway: governance isn't the brake on AI adoption - it's the thing that makes the adoption pay off.

💡 Your Uptime Isn't Yours Anymore.

Microsoft 365's multi-day disruption is a reminder that when you buy SaaS, you're buying someone else's resilience along with the features. SLAs improving doesn't remove the dependency.

The lesson lands harder when you see how this outage actually unfolded.

The takeaway: map your critical dependencies now, or discover them during the next incident.

💡 Data Provenance Just Became a Product Feature.

The sale of battlefield drone data shows how fast dual-use inputs are flowing into commercial AI. Unvetted data slipping into a model is no longer a theoretical risk, it's a shipping one.

Ukraine is opening millions of annotated combat frames to over 100 companies, a world-first data pipeline running straight from the front line into commercial models.

The takeaway: "where did this training data come from" is becoming a question your customers ask, not just your lawyers.

💡 Alignment Is Turning Into a Sales Argument.

OpenAI's push around Astra reframes safety as something you sell, not something that slows you down. In regulated and high-stakes buying, provable alignment is starting to close deals.

Watch the framing: Brockman pitched Astra as the company's most aligned model yet, alignment as a headline product virtue, not a footnote.

The takeaway: trust is moving from the compliance column to the revenue column.

💡 Recurring Revenue Stopped Being Predictable.

New research shows startup ARR is less secure than the SaaS model assumed, buyers churn faster and commit more cautiously as AI reshuffles their stacks. Incumbents with deep integrations get the benefit of the doubt; newcomers have to prove ROI fast.

The breakdown of why ARR got fragile is the one to forward to your board.

The takeaway: land-and-expand only works if the "land" proves value in weeks, not quarters.

💡 Security and IT Automation Are Merging Into One Budget Line.

Palo Alto Networks didn't spend $500M on Console to add a feature. it bought a way to unify AI-driven IT service automation with security. Consolidation at this tier means your policies can't live in separate silos anymore.

The $500M price tag on Console tells you where enterprise spend is pooling.

The takeaway: fragmented automation pipelines are becoming a liability, not a rounding error.

💡 The AI Stack Now Runs From the Street to Orbit.

The frontier isn't clever software anymore, it's an end-to-end, governed stack that can operate anywhere. This week that ranged from Kalanick's quiet robotaxi move to Europe's first fully commercial orbital launch.

The takeaway: time-to-value now depends as much on regulatory footing and data integration as on the model itself.

Trending Tools 📈

donnemartin/system-design-primer
(+31 ⭐ per day, 🐍 Python) Link

• Open study guide for designing large-scale systems, with Anki flashcards

• Helps teams: fast shared vocabulary for system-design reviews and interview prep

codecrafters-io/build-your-own-x
(+31 ⭐ per day, 📝 Markdown) Link

• Guides for rebuilding real technologies from scratch

• Helps teams: turns "how does X actually work" into hands-on understanding for newer engineers

avelino/awesome-go
(+14 ⭐ per day, 🟢 Go) Link

• Curated index of Go frameworks, libraries, and tooling

• Helps teams: cuts research time when a Go project needs a vetted library fast

openai/whisper
(+13 ⭐ per day, 🐍 Python) Link

• Robust speech-to-text model that performs well out of the box

• Helps teams: drop-in transcription for voice features, meeting notes, and accessibility pipelines

rclone/rclone
(+4 ⭐ per day, 🟢 Go) Link

• Rsync for cloud storage - syncs and moves data across 70+ providers

• Helps teams: backbone for cross-cloud backups and migration jobs

slint-ui/slint
(+4 ⭐ per day, 🦀 Rust) Link

• Declarative UI toolkit for native apps in Rust, C++, JS, and Python

• Helps teams: one UI layer across languages for desktop and embedded builds

bitwarden/server
(+2 ⭐ per day, 🟪 C#) Link

• Backend for the open-source Bitwarden password manager

• Helps teams: reference for secure auth, secrets handling, and self-hosting

oven-sh/bun
(+2 ⭐ per day, ⚡ Zig) Link

• All-in-one JS runtime, bundler, test runner, and package manager

• Helps teams: consolidates the JS toolchain and trims build and test times

Tech Trend of The Week 📊 

🔍 "GPT-6," OpenAI's new flagship, saw a 5000% spike in searches vs. the previous week.

OpenAI shipped GPT-6 Astra on September 3, and searches for the model number surged the moment it dropped.

The number that tells the story is the price: $10 / $50 per million tokens, 2.5x the previous flagship, with a 1M+ token context window and gated cybersecurity features. It's being pitched less as a better chatbot and more as a computer operator that browses, codes, and checks its own work.

The signal: frontier labs are done competing on chat quality and are now charging a premium for models that do things on your behalf. When the best model costs 2.5x more and OpenAI still gates its strongest features, the agentic era just got an official price floor.

Our Partner 🎉 

Find the perfect marketing agency, for free.

Stop writing RFPs. Tell us your budget and needs, and our experts send back 3 or 4 agencies worth pitching. The no cost way to source the best agency, without lifting a fingers

Get Matched Free

Hit Reply And Tell Me 💬 

What's the most impressive tech you've seen recently that actually works?

I read every single reply

Go where your customers actually are 📍

Most of the internet copies Reddit.

Blogs, SEO pages, AI models, product reviews. They all pull from the same well.

Reddit is the 3rd biggest site on earth. And it trains the bots your users ask for help.

Odd Angles Media runs Reddit campaigns for over 45 brands each month.

Luckily, we’ve convinced them to give away all of their strategies in a (free) ebook ⬇️⬇️

Grab The Free Ebook (Free) →

Reach the People Who Sign the Checks 💰️ 

40,000+ CTOs and engineering leaders. 96% US-based. 80%+ corporate emails.

They evaluate vendors. They shortlist solutions. They buy.

Advertise with us on Paved → (discount applied)