💡 Stripe Just Made a $7B Bet That Ends the AI Middleman Era

Sponsored by

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

Featured Tool 📊 

If your AI analyst is right 80% of the time, who’s accountable for the other 20%?

ThoughtSpot Webinar Ad

If your AI analyst is right 80% of the time, who’s accountable for the other 20%?

After 15 years at the forefront of data and analytics, Andy Cotgreave has a lot to say on this and other taboo topics, and he's not afraid to tell it like it is.

Tune in to hear Andy and ThoughtSpot SVP of Product Management, Francois Lopitaux, discuss whether a “single source of truth” is actually achievable, the possibility of the prompt becoming the new UX, whether agentic AI is the end of the analyst, and much more.

Don’t miss this chance to sharpen your perspective and elevate your approach to AI-powered analytics.

Watch Now

Quick Hits 🎯

• 🤝 IBM is partnering with OpenAI to push enterprise AI straight into production, not just pilots

• 💰 OpenAI-backed Thrive Holdings just raised $2B to bring full-stack AI to enterprise workflows

• 🧠 The cheapest way to cut RAG costs isn't a better model. It's deciding what never reaches one

• 💳 Stripe is reportedly closing in on a $7B+ acquisition of AI gateway OpenRouter

• 🪶 Anthropic's new watermarking system claims to solve AI provenance. The mechanics say otherwise

• 🚗 AI code-testing startup Blacksmith just saw its valuation jump almost 10x in under a year

• 📍 Flock is rewriting the rules on its license-plate network after a privacy backlash

🎁 + 4 other stories you might find useful

Our Partner 🎉 

The best voice models now listen, adapt, and resolve too.

Most CX platforms don't own the voice. They orchestrate a workflow, then call a third party for speech and transcription. Every hop adds latency, and latency is what turns a frustrated customer into a churned one.

ElevenAgents is the opposite. Built on the voice models the market already builds on, it runs voice, transcription, chat, and reasoning in one vertically integrated pipeline. Responses come back in under 400 milliseconds and sound human, not synthetic. When a caller gets frustrated, the agent detects it and shifts tone in real time: calm, reassuring, patient.

You keep full control. Plug in any LLM, connect tools, webhooks, and MCP servers, and ground every answer in your knowledge base. Launch in minutes, A/B test with Experiments, enforce Guardrails, and version every change.

More resolved conversations, less infrastructure stitching. Pricing is transparent and flat at $0.08 per minute.

Start building free

The Big Picture 🖼️

💡 The AI Gateway Is Becoming the New Payment Rail.

Stripe built its empire by sitting quietly between merchants and money, taking a small cut of every transaction that flowed through. Now it's reportedly about to do the same thing for AI, an acquisition that would put Stripe between every developer and every model they call.

Why it matters: if you're building anything AI-powered, the company that owns model routing owns your unit economics.

It decides which providers get traffic, what pricing you see, and how fast you can switch when a model gets cheaper or better somewhere else.

The takeaway: infrastructure decisions you make this quarter could lock in your AI costs for years. Choose the layer you don't control carefully.

💡 Enterprise AI Is Consolidating Around Full-Stack Platforms.

IBM's new partnership with OpenAI isn't a pilot program, it's a bet that enterprises want one throat to choke, not fifteen vendors to coordinate.

Around the same time, OpenAI-backed Thrive Holdings raised $2B to build the connective tissue between data, models, and the ERP and CRM systems companies already run on.

Why it matters: the "best model wins" era is fading. The winners now are the platforms that can plug into procurement, security review, and existing tooling without a six-month integration project.

That's a harder problem than training a better model, and it's where the real budgets are moving.

The takeaway: if your team is still evaluating AI vendors purely on benchmark scores, you're optimizing for the wrong variable.

💡 The Real AI Cost Savings Happen Before the Model Ever Runs.

Everyone's trying to shrink inference bills by picking cheaper models. The bigger lever is upstream: routing predictable, low-stakes questions to rules and classifiers so they never reach the LLM at all.

That single design choice can cut RAG inference costs 6x without touching model selection.

Why it matters: this is the difference between an AI budget that scales with usage and one that scales with revenue. Teams that build this filtering layer early save money every single month it stays in production.

The takeaway: before you optimize your model, optimize what even gets to your model.

💡 Watermarking Won't Solve What Everyone Hopes It Will.

Anthropic's approach to watermarking AI output sounds like a clean answer to a messy problem: how do you know what's AI-generated?

The mechanism behind it tells a more complicated story, one where compliance theater and real accountability aren't the same thing.

Why it matters: regulators and enterprise buyers are going to start asking for watermarking as a checkbox.

Understanding what it actually verifies, and what it doesn't, is the difference between a real safeguard and a false sense of security.

The takeaway: don't let a badge substitute for a real answer when someone asks how you know content is AI-generated.

💡 AI Testing Just Became a Venture-Scale Category.

Blacksmith isn't a household name, but its valuation just jumped almost 10x in under a year as AI-generated code creates a testing bottleneck nobody built for.

When AI writes code faster than humans can review it, the bottleneck moves from writing to verifying.

Why it matters: every team shipping AI-assisted code is quietly running into this same wall. The startups solving verification, not generation, are becoming the picks-and-shovels play of this cycle.

The takeaway: watch the tooling category around AI, not just the models. That's often where the durable value ends up.

💡 Trust Infrastructure Is Becoming a Competitive Moat.

Flock's license-plate reader network just got rewritten after regulators and privacy advocates pushed back hard. The rule changes reveal something bigger about how fast "move fast" data companies are being forced to slow down.

Why it matters: the companies that build governance and consent into their data products from day one will outlast the ones that treat it as an afterthought. Regulatory risk is now a real cost line, not a hypothetical one.

The takeaway: if your product touches sensitive data, your privacy posture is now part of your competitive pitch, not just your legal review.

💡 The Best Operators Are Playing a Portfolio Game, Not a Bet-the-Company Game.

Amid all the single big headlines, this week's roundup of smaller wins across hardware, media, and partnerships is a reminder that most durable companies aren't made from one genius move. They're made from a dozen small ones stacked in the same direction.

Why it matters: it's easy to read tech news as a series of bets. In practice, the companies that compound fastest are testing many small things and doubling down on what works.

The takeaway: your roadmap doesn't need one big swing this quarter. It needs several small ones pointed the same direction.

Trending Tools 📈

system-design-primer (+231 ⭐ per day, 🐍 Python) Link

• A comprehensive walkthrough of large-scale system design, plus flashcards for interview prep

• Helps teams: sharpens architectural thinking across every level, from new grads to staff engineers running design reviews

lm-evaluation-harness (+20 ⭐ per day, 🐍 Python) Link

• A standardized framework for few-shot evaluation of language models

• Helps teams: makes model comparisons rigorous instead of vibes-based, which matters more as model choice becomes a cost decision

hoppscotch (+8 ⭐ per day, 🟦 TypeScript) Link

• An open-source API development suite with offline, on-prem, and cloud options

• Helps teams: streamlines API testing and collaboration without the vendor lock-in that comes with closed tools

velero (+7 ⭐ per day, 🟢 Go) Link

• Backup and migration tooling built for Kubernetes workloads and persistent volumes

• Helps teams: turns disaster recovery from a fire drill into a routine, tested process

rustdesk (+143 ⭐ per day, 🦀 Rust) Link

• A self-hosted, open-source remote desktop app built as an alternative to commercial tools

• Helps teams: gives support and ops teams secure remote access without paying per-seat for a commercial license

cal.diy (+64 ⭐ per day, 🟦 TypeScript) Link

• Scheduling infrastructure built for teams that want to own their booking logic

• Helps teams: replaces black-box scheduling SaaS with primitives you can actually extend

Tech Trend of The Week 📊 

🔍 "AI gateway" searches spiked hard this week

Stripe's reported $7B run at OpenRouter turned a niche infrastructure term into something people outside of engineering are now typing into Google. Searches around AI gateways and model routing climbed sharply the same day the acquisition reports broke.

The signal: the AI infrastructure layer just went mainstream enough that non-technical operators are trying to understand it. When a category starts trending outside its usual audience, that's usually a sign the money has already moved in.

Our Partner 🎉 

AI Insights. Real Growth. Higher GMV, Better Profits

The difference between growing stores and stagnant ones isn't more effort. It's better insights. StoreClaw analyzes your Shopify and Amazon data, surfaces your biggest growth opportunities, and helps you increase GMV while protecting profit. Start free with bonus tokens. No credit card required.

Try it FREE today

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)