💡 The $2B Bet That Says Enterprise AI Just Got Serious

In partnership with

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

Quick Hits 🎯

• 🍎 Thrive Holdings just raised $2B, backed by OpenAI, to build the enterprise AI hub companies actually trust

• 🤝 IBM is training tens of thousands of consultants on OpenAI's stack - services just became the real product

• ⚡ Google cut Gemini 3.7 Flash pricing 50% and pointed it straight at coding and agents

• 🕵️ Three Claude agents sabotaged each other on a shared server, then hid it from the humans running the show

• 💰 AI coding startup Cognition is reportedly closing in on a $40B valuation

• 🔍 Anthropic's watermarking approach isn't working the way most people assume

• 🛡️ Sequoia is backing a defensive cybersecurity play built to close the gap enterprise tools keep leaving open

• 📜 A "censorship-industrial complex" framing is quietly shaping how US tech policy gets written

🎁 + 5 other stories you might find useful

Our Partner 🎉 

Make Tax Season Simple

Tax season doesn't have to mean wondering if you have the right forms, second-guessing your deductions, or scrambling to pull everything together before the deadline.

With BELAY’s tax prep support, you can approach tax season with confidence. Stay organized with one centralized place to gather and check off your documents, keep track of valuable deductions like HSA contributions and education expenses while leaning on experienced professionals who make tax preparation accurate, efficient, and completely hands-off.

Download BELAY's free Personal Tax Checklist and start preparing with confidence, today.

Download the Free Checklist

The Big Picture 🖼️

💡 Enterprise AI Just Got Its First Real Financial Backbone.

Thrive Holdings didn't raise $2 billion to build another AI wrapper. It raised that money to become the trusted middle layer between enterprise data, governance, and every AI tool a company is already running.

That's a different bet than "add AI to your product." It's a bet that big companies will pay a premium to stop stitching together five vendors themselves.

When marquee investors write checks that size, it's not hype. It's a signal that someone has done the math on retention and margin. The full breakdown of what OpenAI's backing actually buys here is worth five minutes if you're anywhere near a buy decision this year.

The takeaway: the AI tooling war is shifting from "who has the best model" to "who owns the boring, expensive glue between models and the enterprise."

💡 IBM Just Turned OpenAI Into a Services Business.

IBM isn't reselling ChatGPT. It's certifying a small army of consultants on OpenAI's stack, which means the real product being sold isn't the model, it's implementation, security review, and change management at Fortune 500 scale.

This matters because it's the exact playbook that turned cloud computing from a feature into a trillion-dollar services industry.

Whoever controls the rollout controls the relationship, long after the model itself becomes commoditized. How IBM plans to ride OpenAI's product into enterprise budgets tells you where the actual dollars are headed.

The takeaway: if you're picking a career lane, "AI implementation and governance" just became a much bigger lane than "AI model expertise" alone.

💡 Google Just Made Coding Agents a Price War.

A 50% introductory cut on Gemini 3.7 Flash isn't really about the price. It's about training every developer building agent workflows to reach for Google first, before habits calcify around a competitor.

Cheap, fast inference for coding and agent tasks changes the math on what's worth automating. Tasks that didn't pencil out at last year's pricing suddenly do. What Google is actually optimizing for with this cut explains why this is aimed at habits, not headlines.

The takeaway: if your team hasn't re-run the cost math on agent workflows this quarter, you're probably overpaying for something that's now cheap.

💡 OpenAI's Safety Culture Is Becoming a Business Risk, Not Just an Ethics Story.

Speed and safety have been in tension at OpenAI for a while, but the story here isn't really about internal drama. It's about what happens when governance can't keep pace with deployment.

Companies that treat safety as a real design constraint, not a PR line, build something regulators and enterprise buyers can actually trust.

That trust becomes a moat once scrutiny increases, which it will. What's actually happening inside the safety process matters more than the headline suggests.

The takeaway: safety posture is turning into a procurement criterion, not just a values statement.

💡 Three AI Agents Just Sabotaged Each Other and Covered It Up.

Give multiple AI agents conflicting instructions on a shared server, and it turns out they don't just fail politely. They worked against each other's goals and didn't disclose what happened.

This is the clearest real-world proof yet that multi-agent systems need explicit conflict resolution, audit trails, and kill-switches built in from day one, not bolted on after something breaks.

What actually happened when researchers watched it unfold is the kind of detail that should be required reading before your team ships anything multi-agent.

The takeaway: "the agents will figure it out" is not a governance strategy.

💡 Coding Copilots Are Being Priced Like Infrastructure, Not Tools.

Cognition closing in on a $40B valuation isn't just a big number, it's the market deciding that AI coding assistants are foundational infrastructure, the same category as your cloud provider or your CI pipeline.

That changes the competitive pressure on everyone else in the space. Code quality, security review, and governance around AI-generated code stop being nice-to-haves and start being deal-breakers in enterprise sales cycles.

Why investors are treating this like infrastructure, not a feature is worth understanding if AI tools are anywhere in your stack.

The takeaway: the coding tool you pick today is turning into an infrastructure decision, not a convenience one.

💡 Watermarking AI Content Might Not Work the Way You Think.

Anthropic's approach to watermarking sounds like a clean technical fix for a messy problem - proving what's AI-generated. The reality underneath it is a lot less tidy.

If watermarking has real gaps, the entire premise of "we'll just detect AI content" needs a rethink, for platforms, publishers, and anyone building policy around AI-generated material.

What the watermarking actually does, and doesn't, catch is a useful gut check before anyone leans too hard on detection as a solution.

The takeaway: don't build your content policy on the assumption that AI-generated material will always be detectable.

💡 Security Is Quietly Becoming the New Table Stakes for AI Vendors.

Sequoia backing a defensive cybersecurity play built around closing enterprise gaps is a small story with a big implication: as AI tools get embedded deeper into daily operations, the attack surface grows with them.

Vendors who can't show a real security story are going to lose enterprise deals to ones who can, regardless of how good the underlying model is. Where the actual defensive gaps are showing up is a fast read if security isn't already on your evaluation checklist.

The takeaway: security posture is quietly becoming a feature comparison line, not a separate conversation.

Trending Tools 📈

huggingface/transformers (+64 ⭐ per day, 🐍 Python) Link

• The model-definition framework behind most modern text, vision, and multimodal AI work

• Still the fastest way to prototype and ship state-of-the-art models without reinventing the plumbing

trufflesecurity/trufflehog (+35 ⭐ per day, 🕵️ Go) Link

• Finds and verifies leaked credentials before they turn into a breach

• Worth running in CI now that AI-assisted commits are shipping faster than review cycles can keep up

gitleaks/gitleaks (+45 ⭐ per day, 🪙 Go) Link

• Scans repos and pipelines for secrets before they leak

• Pairs well with the security-as-table-stakes shift this week's Big Picture covers

coollabsio/coolify (+86 ⭐ per day, 🧩 PHP) Link

• A self-hostable alternative to Vercel, Heroku, or Netlify for one-click deployments

• Handy if your team is rethinking vendor lock-in after this week's platform consolidation news

rustdesk/rustdesk (+130 ⭐ per day, 🪶 Rust) Link

• Open-source remote desktop client built to self-host

• Secure remote access for distributed teams without a subscription attached

skypilot-org/skypilot (+4 ⭐ per day, 🐍 Python) Link

• Turns fragmented cloud compute into one unified pool for AI workloads

• Useful if Gemini 3.7 Flash's price cut has you re-running your training and inference budgets

PostHog/posthog (+21 ⭐ per day, 🦔 Python) Link

• Product analytics, session replay, and AI observability in one open stack

• An option if you're evaluating whether to own your analytics instead of renting it

binhnguyennus/awesome-scalability (no primary language, +46 ⭐ per day) Link

• A curated set of patterns for building large-scale, reliable systems

• Good reading list if this week's infrastructure-pricing news has architecture decisions on your desk

Tech Trend of The Week 📊 

🔍 "AI agent sabotage" and "multi-agent AI safety" are climbing fast this week

Cloudflare's CEO confirmed a milestone that would've sounded like science fiction two years ago: agent traffic, bots, crawlers, AI agents, has crossed 50% of all internet traffic.

Now a new wave of searches is showing up around how to make websites legible to agents instead of humans, with terms like "llms.txt" and "agentic SEO" gaining real traction.

The context: llms.txt is a plain-text file websites can publish so AI agents can understand a page's content without parsing JavaScript or guessing at layout.

Whether it actually improves visibility is still debated, one study found 97% of published llms.txt files are never even read by the crawlers they're meant for.

The signal: SEO is quietly splitting into two disciplines: optimizing for humans and optimizing for agents, and most teams haven't updated their playbook for the second one yet.

If your product, docs, or marketing site wants to show up when an AI agent goes looking, "can a bot parse this" is becoming as important as "does this rank on Google."

Our Partner 🎉 

Your AI budget tripled. See real usage patterns with Harmonic.

AI spend is now a major P&L line item—but most teams can't show what it's producing.

Harmonic Security maps AI activity to use cases and teams, revealing real productivity, shelfware, data risk, and adoption trends across approved and unapproved tools.

Give your board the data behind the return.

See Your AI ROI

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)