💡 The $100M Bet That Says Code Isn't the Future

In partnership with

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

Featured Tool 📊 

Your team ships AI code faster than anyone can review it.

And you're still accountable when it breaks in production. A Depot staff engineer lays out the guardrails that keep AI-assisted teams fast and safe: constraints that force reviewable code, reference tests agents can copy, and PRs small enough to approve.

Read the workflow →

Quick Hits 🎯

• 🧭 Anthropic just cut Claude Opus 5's price in half. Coding, agents, and enterprise workflows all got cheaper to run at scale

• 🏛️ Enterprise AI agent adoption is outrunning the governance built to control it

• 🔓 America's "open model" strategy might be a bigger gamble than it looks

• 💰 Reid Hoffman and Mark Pincus are raising $100M for a new lab betting on AI-run workflows, not more code

• 🖥️ IBM says the mainframe isn't dying. AI just moved the argument

• ⚡ A new chip partnership is betting the next fight is over milliseconds, not model size

🎁 + 4 other stories you might find useful

Our Partner 🎉 

Thinking about hiring globally? Start with an EOR.

The best person for your next role might not live near your office—or even in the same country.

More companies are realizing they don't need to open entities everywhere just to access global talent. Instead, they're using EOR to hire internationally faster, stay compliant, and avoid building local infrastructure before they're ready.

Oyster's EOR helps companies hire, pay, and support employees in 180+ countries while Oyster handles payroll, compliance, taxes, and local employment requirements.

See how it works

The Big Picture 🖼️

💡 The Governance Layer Just Became the Bottleneck.

Enterprises spent the last year deploying AI agents faster than any software wave before it. The policy, data, and runtime controls needed to supervise those agents didn't move at the same speed, and that gap is now showing up in budgets, vendor swaps, and rebuilt approval chains.

This matters because "move fast" only works until an agent touches a system nobody remembered to lock down. Teams that treated oversight as a phase-two problem are now doing phase-one work in production, under a deadline they didn't choose.

Where the gap is widest and who's already getting burned by it is a shorter list than most CTOs expect.

The takeaway: if your AI rollout plan doesn't have a governance line item, it's not actually a plan. It's a head start on a problem you'll be solving later, under worse conditions.

💡 Openness Just Became a Bet, Not a Value.

Open-weight models spread fast and build ecosystems quickly. They also raise the bar on safety and performance parity, because anyone can inspect what you shipped.

For a CTO, this isn't a philosophical choice anymore. It's a tradeoff between locking in developers early through openness and protecting a performance or compliance edge behind a closed system. Both paths are defensible, which is exactly why the decision is harder than it looks.

The full case for why this tradeoff is sharper than it's ever been is worth fifteen minutes before your next model-selection meeting.

The takeaway: whichever way you lean, interoperability and a real safety story are becoming the actual product, not the marketing around it.

💡 A New AI Lab Is Betting the Future Isn't Code. It's Workflow.

Reid Hoffman and Mark Pincus are in talks to raise $100M for Prentis, a lab built around automating the routine work inside a company, not writing more of it.

If that thesis is right, the competitive pressure isn't "does your team use AI coding tools" anymore it's whether your internal processes get automated before a leaner competitor's do. Incumbents who assumed AI meant "faster engineers" may be answering the wrong question.

What Prentis is actually building, and why two people who don't need the money are chasing it tells you where the automation money is starting to flow.

The takeaway: the next efficiency gain probably doesn't come from your codebase. It comes from the process nobody's touched since 2019.

💡 The Mainframe Isn't Dead. It's Getting a Co-Pilot.

After a rough quarter, IBM pushed back hard on the idea that AI is quietly killing off mainframe revenue. Their position: budget reallocation toward AI is temporary, not a replacement for systems that still run payroll, banking, and insurance at scale.

This is a useful reality check. The loudest AI narrative, rip and replace, rarely matches what's actually happening inside large, regulated organizations, where legacy systems are expensive to move and even more expensive to get wrong.

IBM's actual argument for why the old and new stacks are merging, not competing is worth reading before you write off legacy infrastructure spend.

The takeaway: hybrid wins. The teams building AI on top of what already works will out-execute the teams betting on a clean rebuild.

💡 Anthropic Just Moved the Price-Performance Line.

Claude Opus 5 landed at roughly half the price of the prior flagship, aimed squarely at coding, agents, and enterprise workflows.

Every competitor now has to answer a question they didn't have to answer last quarter: match the price, match the capability, or find a different differentiator entirely, safety tooling, ecosystem lock-in, latency. Price compression like this usually pulls adoption forward faster than any marketing push could.

The specifics on what Opus 5 actually changes for teams running AI at production scale are worth a look if you're budgeting AI spend for next quarter.

The takeaway: if your AI cost model was built six months ago, it's already out of date.

💡 The Next Fight Isn't Over Models. It's Over Milliseconds.

A new partnership is pushing hardware built specifically to make inference faster and cheaper, not another general-purpose chip, a purpose-built one.

Model quality gets most of the headlines, but latency and cost-per-inference decide who can actually ship real-time AI features at a price that scales. Whoever solves that quietly wins the deployments nobody writes press releases about.

The bet behind building inference-specific hardware instead of chasing bigger models explains why some of the smartest capital in the room isn't chasing another foundation model.

The takeaway: access to fast, cheap inference is becoming as strategic as the model itself. Don't sleep on the plumbing.

💡 Fast Integration Has a Dark Side.

The push to stitch together best-in-class AI components, one lab's model, another's tooling, a third's data layer, is accelerating. So is the risk that alignment gets lost somewhere in the seams.

Speed and safety pull against each other here in a very literal way. The faster you plug modules together, the harder it becomes to guarantee the combined system behaves the way any single piece was designed to.

What actually happened when the pieces didn't line up the way everyone assumed is a preview of a failure mode most teams haven't planned for yet.

The takeaway: modular AI is fast to build and hard to fully trust. Budget time for the seams, not just the parts.

💡 "Open" and "Closed" AI Are Fighting a Cold War Now.

Open ecosystems, capital-backed labs, and state-supported efforts are all pulling in different directions at once, and the gap between them is starting to look permanent rather than temporary.

For anyone building a long-term AI strategy, this means picking a lane matters more than it used to. Geopolitics, supply chains, and data provenance are no longer background noise, they're inputs to the actual roadmap.

The full read on how this standoff is reshaping who gets access to what is worth the longer sit-down read this week.

The takeaway: strategy now has to account for who controls the compute and the data, not just who has the best model.

💡 Layoffs Are Becoming AI's Loyalty Test.

Monday.com laid off hundreds of employees this week to refocus the company around AI-driven product lines.

This is the part of the AI story that doesn't show up in keynote demos: reallocating headcount toward AI only works if the product actually lands. Hiring your way to an AI pivot without shipping something customers want is just an expensive way to find out you guessed wrong.

What Monday.com is betting the layoffs will buy them is a case study worth watching over the next two quarters.

The takeaway: AI headcount reshuffling is now a real signal of confidence or desperation. Watch what ships next, not what got announced.

Trending Tools 📈

sindresorhus/awesome (+409 ⭐ this week, no language) Link

• A curated index of curated lists, spanning nearly every technical and career topic

• Helps teams: gives you a vetted starting point when exploring a new stack, instead of trusting a random blog post

1N3/Sn1per (+121 ⭐ this week, 🔷 Shell) Link

• Automated pentesting and attack surface management with 600+ exploits and 90+ integrations

• Helps teams: gives security teams one console to find and document gaps instead of stitching together a dozen scanners

lukasz-madon/awesome-remote-job (+73 ⭐ this week) Link

• A curated list of remote-friendly companies and job boards

• Helps teams: hiring managers building distributed teams get a vetted sourcing list instead of starting from zero

darkroomengineering/lenis (+72 ⭐ this week, 🔷 TypeScript) Link

• A smooth-scroll library that makes web pages feel less stiff, without a heavy framework

• Helps teams: frontend teams add a polish layer to marketing sites and product UI in an afternoon

supabase/supabase (+71 ⭐ this week, 🔷 TypeScript) Link

• An open-source Postgres backend with auth, storage, and real-time built in

• Helps teams: ship a production backend without wiring together five separate services

paperless-ngx/paperless-ngx (+54 ⭐ this week, 🐍 Python) Link

• A self-hosted document system that scans, OCRs, and indexes paperwork automatically

• Helps teams: ops and legal teams cut the time spent digging through PDFs for one clause

simplex-chat/simplex-chat (+45 ⭐ this week, 🟢 Haskell) Link

• A messaging network built with no user identifiers at all, not even a phone number

• Helps teams: security-minded teams get a private channel for sensitive coordination with no account trail

tailscale/tailscale (+36 ⭐ this week, 🔷 Go) Link

• A mesh VPN built on WireGuard that sets up in minutes, not a weekend

• Helps teams: remote and hybrid teams get secure access to internal tools without a hardware VPN box

Tech Trend of The Week 📊 

🔍 "Claude Opus 5" spiked this week

The price cut on Anthropic's flagship model didn't just make headlines, it sent engineering leads straight to Google to figure out what changed and whether a migration is worth the effort.

The signal: when a lab halves the price on its top model, the search spike isn't curiosity. It's teams already running the numbers on a switch.

Our Partner 🎉 

Stop switching apps. Your browser can do it all.

Every tab you open, every copy-paste into ChatGPT, every lost train of thought — that's your browser failing you. Norton Neo fixes it. Built-in AI works directly inside your session. Hover to preview. Search everything from one bar. VPN and ad blocking included, free.

Try Norton Neo 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)