Codeinate
Posts
💡 The $188B Question Every CTO Is Avoiding

💡 The $188B Question Every CTO Is Avoiding

PLUS: The identity problem AI agents just solved everyone's headache becomes a breach issue.

Rachel Miller
July 23, 2026

In partnership with

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

Quick Hits 🎯

🧼 RAG isn't broken, your data is. The real reason enterprise AI pilots stall has nothing to do with the model
🤖 Sequoia-backed Bunkerhill Health is putting AI agents directly into patient care workflows, not just chat
⚡ Databricks just hit a $188B valuation, here's why "boring" data infrastructure is AI's most valuable real estate
🧠 Anthropic and Blackstone are betting the next trillion-dollar AI company won't build a model at all
💎 Reflection AI locked in a $1B compute deal with Nebius, open-source research just got a serious budget
🕵️ A $60M-funded startup called Oak says AI agents have created an identity crisis nobody's solved yet

🎁 + 2 other stories you might find useful

Our Partner 🎉

Porkbun is the domain name registrar you need.

Still using GoDaddy or Namecheap? There’s a better way with Porkbun!

Porkbun is the domain registrar trusted by creators, developers, entrepreneurs, and folks who want low prices without the nonsense.

Why people are choosing Porkbun:
• Most domains sold at cost
• Low, transparent registration and renewal pricing
• Free features like WHOIS privacy and SSL certificates
• Powerful web and email hosting options
• Real human support 24/7, 365 days a year
• Named the #1 domain registrar by Forbes Advisor and USA Today

For launching a business, building a personal brand, starting a side project, or creating your first website, Porkbun makes it easy.

Get $1 off your next domain registration with Porkbun now.

Get Your Domain Name Now

The Big Picture 🖼️

💡 Your AI Strategy Has a Data Problem, Not a Model Problem.

RAG looks like magic in a demo and falls apart in production, and it's rarely the retrieval logic that breaks. It's the underlying data, messy, duplicated, ungoverned.

If your team keeps reaching for a bigger model or a cleverer prompt every time output quality drops, you're treating a symptom.

The teams actually shipping reliable AI in production spent their first quarter on unglamorous work: data lineage, deduplication, access control. That's the boring part nobody wants to fund, and it's exactly why most RAG rollouts quietly stall between pilot and production.

The takeaway: fix the data foundation before you touch the model. Every layer you build on top inherits whatever mess sits underneath it.

💡 Healthcare Just Showed What "Agent" Actually Means.

Sequoia's partnership with Bunkerhill Health isn't another AI chatbot bolted onto a patient portal. It's agents doing the unglamorous, high-stakes work of monitoring and care coordination inside real clinical workflows.

This matters beyond healthcare because it's a preview of what "AI agent" needs to mean everywhere else too: not a conversational layer, but something trusted enough to act inside a system where mistakes are expensive.

Getting there requires accuracy guarantees and safety controls that most consumer AI products never had to build. The integration model behind this partnership is worth studying regardless of your industry.

The takeaway: the bar for "agent" keeps rising. If your product still means "chatbot with extra steps," you're behind where the market is heading.

💡 Databricks' $188B Valuation Is a Bet on Plumbing, Not Models.

While the headlines chase frontier model releases, Databricks built its valuation on something less exciting: making enterprise data usable for AI at all.

Lakehouse architecture plus open-weight model research is a quiet but deliberate combination.

It says the money isn't only in who has the smartest model, it's in who lets you actually connect that model to your company's data without a six-month integration project.

What's driving this valuation run tells you where enterprise budgets are actually flowing this year.

The takeaway: cost control and data infrastructure are becoming the real competitive levers, not just model benchmarks.

💡 Anthropic and Blackstone Think the Real Money Isn't in Models.

Here's a contrarian bet from two firms with serious capital behind it: the next trillion-dollar AI company will win on implementation, not on training the sharpest model.

Models get commoditized fast. What's much harder to copy is a company that knows how to deploy AI reliably across messy, real-world enterprise environments, integration expertise, governance muscle, and client trust built over years.

The thesis behind this bet reframes where the durable value actually sits in this market.

The takeaway: if your differentiation strategy is "we use the best model," you don't have one. Everyone will have access to the best model eventually.

💡 Reflection's $1B Compute Deal Signals Open Research Is Going Industrial.

A billion-dollar compute agreement with Nebius moves Reflection AI from research curiosity to something that can run production-scale experiments on a predictable budget.

This is the quiet infrastructure story behind every open-model headline you'll read this year. Compute access, not cleverness, increasingly decides who can iterate fast enough to matter. The details of this deal show how open-source AI research is starting to look a lot like enterprise procurement.

The takeaway: watch compute deals as closely as model releases. They tell you who's actually positioned to compete, not just who's making noise.

💡 AI Agents Broke Identity Management, and Someone's Finally Fixing It.

Oak just came out of stealth with $60M to solve a problem most companies haven't even named yet: what happens to identity and access control when the "user" acting in your systems is an AI agent, not a person.

Traditional IAM was built for humans logging into things. Agents that act autonomously, chain tasks together, and touch multiple systems in seconds break assumptions that identity infrastructure has relied on for decades.

What Oak is building to fix this is a preview of a category that's about to get very crowded.

The takeaway: if your org is rolling out agents faster than it's rethinking identity and access, you have a governance gap that will surface at the worst possible time.

💡 The Assistant Wars Are Shifting From Answering to Doing.

Glean's positioning isn't "ask it anything," it's "grounded in your company's context, capable of finishing the task." That's a meaningfully different product than a search bar with a chat interface stapled on.

The distinction matters because most AI assistant fatigue comes from tools that answer questions well but can't actually move work forward.

Enterprise buyers are starting to notice the difference, and it's reshaping who wins the "AI at work" category. How Glean is positioning around action, not just answers shows where this market is actually heading.

The takeaway: "grounded and action-ready" is becoming the new baseline. Chat-only assistants are starting to look dated.

💡 IBM's Mainframe Moat Is a Lesson in What Modernization Actually Requires.

IBM's struggles alongside its still-formidable mainframe business make an uncomfortable point: having a defensible core product doesn't protect you if you can't modernize around it fast enough.

The mainframe still does things cloud-native competitors can't easily replicate, security, reliability, and workloads that simply can't tolerate downtime.

But the AI-era expectation is hybrid modernization that keeps those strengths while adding real developer velocity, and that's a much harder needle to thread than most roadmaps admit.

The full breakdown of IBM's position is a useful gut-check for anyone managing legacy systems under pressure to "just add AI."

The takeaway: a moat only holds if you can build on top of it fast enough. Legacy strength without velocity is a countdown clock, not a safety net.

💡 The Real AI Race Moved Past the Frontier.

Chasing the single best model is starting to look like the wrong game. The companies actually capturing enterprise budget are the ones nailing governance, cost efficiency, and flexibility with open models they can own, audit, and adapt.

This is the thread that ties every story in this issue together: production AI wins on reliability and control, not raw capability.

Frontier benchmarks make headlines, but auditable, compliant AI at scale is what actually gets budget approved. Why the frontier may no longer be where the real competition is is worth reading before your next model-selection meeting.

The takeaway: enterprise AI succeeds when models, tooling, and governance line up to deliver value you can actually measure and defend.

Trending Tools 📈

PostHog/posthog (+411 ⭐ per week, 🐍 Python) Link

Product analytics, session replay, feature flags, and AI observability in one open-source platform
Helps teams diagnose issues and validate features without stitching together five separate tools

SigNoz/signoz (+345 ⭐ per week, 🔷 TypeScript) Link

OpenTelemetry-native observability unifying logs, metrics, and traces with AI-assisted insights
Helps teams get end-to-end visibility into complex systems without vendor lock-in

ClickHouse/ClickHouse (+19 ⭐ per week, 🔷 C++) Link

Real-time analytics database built for high-performance analytical workloads
Helps teams run fast queries over massive datasets without the usual warehouse lag

pocketbase/pocketbase (+164 ⭐ per week, 🟢 Go) Link

Open-source real-time backend that ships as a single executable file
Helps teams stand up data-driven apps and internal tools in a fraction of the usual setup time

appsmithorg/appsmith (+33 ⭐ per week, 🔷 TypeScript) Link

Low-code platform for building internal tools and admin panels with 25+ integrations
Helps teams cut internal tooling projects from weeks down to days

codecrafters-io/build-your-own-x (+754 ⭐ per week, 🪶 Markdown) Link

A hands-on series for rebuilding familiar tech from first principles
Helps engineers - junior through staff - build the kind of systems intuition that shortcuts don't teach

NixOS/nixpkgs (+15 ⭐ per week, 🔷 Nix) Link

The Nix Packages collection powering reproducible, declarative system builds
Helps teams kill "works on my machine" bugs with deterministic environments across dev and prod

Tech Trend of The Week 📊

🔍 "AI agent identity" is climbing fast in tech search interest

This week's Oak launch didn't happen in a vacuum. As more companies put AI agents into production systems, not just chat interfaces. A wave of teams are hitting the same wall at the same time: their identity and access infrastructure was built for humans, not autonomous software making decisions in seconds.

The signal: agent adoption outpaced the security thinking underneath it. Expect identity and access management for AI agents to become one of the most contested infrastructure categories of the next 12 months.

Our Partner 🎉

The best voice models, now with full orchestration. Build real-time voice and chat agents on one low-latency stack: any LLM, your tools and knowledge, testing, Guardrails, and omnichannel deployment.

Start building free