Ready to transform artificial intelligence from a buzzword into your personal revenue generator?

HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.

Inside you'll discover:

Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.

Get Your Guide

Sunday, June 8, 2026 Intelligence for Portfolio Executives closing the AI Wage Gap.

———————————————————————————

Something happened inside Anthropic's engineering org in May. Not a product launch. Not a press release. An internal report, then a blog post, then a number that got noticed and probably should have caused more of a reaction than it did.

Last month, 80% of the code merged into Anthropic's production systems was written by Claude.

Read that again carefully.

The engineers who built the most capable AI assistant in the world — who understand its failure modes and limitations more precisely than anyone alive, who could route around it if it wasn't actually faster and better — are now doing roughly a fifth of their own software development. Claude is doing the other four-fifths. And those same engineers are now shipping 8 times more code per day than they were two years ago.

This is not a case study. This is the lab.

Here's why this number matters more than any study, projection, or McKinsey scenario analysis published in the past two years.

When a consultancy tells you AI will transform knowledge work, they are modeling futures and selling engagements. When a CEO quotes an AI productivity stat at an all-hands, they are managing a narrative. When Anthropic publishes internal data showing that the humans who built Claude are now using it for 80% of their own production work — they are telling you what is actually happening inside the most safety-conscious, most technically credible AI organization in the world.

These are not people who overstate capability. The entire Anthropic brand is built on understatement and rigor. When they say 80%, they mean 80%.

The 8x output figure gets at something that most executive conversations about AI miss entirely. Eight times is not a productivity improvement. It's a category change. If you measured the output of every knowledge worker in your organization right now, and then asked what it would take to get 8x output from the same headcount, the answer you'd get back would involve years, capital, restructuring, new systems. Anthropic's engineers didn't restructure. They started using their own product seriously, and 8x happened.

There's a second piece of data from this week that connects directly to the first.

Anthropic's internal Mythos Preview model achieved a 52x speedup on AI training code optimization benchmarks — tasks that would take a skilled human 4–8 hours. Claude's success rate on complex, open-ended engineering problems has hit 76%, up 50 points in six months. The lab is calling this a "self-improvement threshold" — the point at which AI systems begin meaningfully contributing to the development of their own successors.

AI Fire called it "AInception." Rundown AI called it "Anthropic's self-improving AI warning." Anthropic itself called it a formal concern worth public disclosure.

I want to be precise about what this means and what it doesn't. Anthropic is not saying the machines are loose or that human oversight is over. They are saying something more specific and more immediately relevant: the gap between what AI can do and what most organizations are using it for has never been wider. The recursive loop is real and measurable inside their own walls, while the companies their clients work for are still running three-month pilots and asking for more evidence.

The 80% figure is the evidence.

Claude Opus 4.8 shipped this week alongside this data. Dynamic workflows running hundreds of parallel subagents in a single session. 2.5x faster processing in speed mode. One million token context. 128,000 token output. Mid-task instruction updates — you can redirect an agent mid-run without starting over.

Simultaneously, Anthropic filed confidentially for an IPO at a $965 billion valuation, surpassing OpenAI's private value. $47 billion annualized revenue run rate. Claude is now deployed across AWS, Google Cloud, and Azure. It has stopped being a product and is becoming infrastructure.

The automated traffic figure from this week is the final piece of context: for the first time in history, bot and agent traffic now accounts for 57.2% of all internet activity. More than half of what moves across the internet right now is not a human doing it. It is agents, scrapers, workflows, and AI systems operating on behalf of humans — or increasingly, operating without them.

What does all of this mean for the Portfolio Executive?

It means the 86/25 gap — 86% of employees supposedly having AI skills, 25% actually using AI at work — is no longer primarily a training problem or a change management problem. It's a leadership problem. Someone inside every organization has to look at the 80% figure and decide what the equivalent is in their function. What is the 80% equivalent in legal? In HR? In finance? In strategy? What would 8x output look like in the work your team does every week?

That question doesn't require a budget. It requires someone willing to ask it seriously and then do the uncomfortable work of mapping the answer.

The person asking that question — and shipping a working proof case against the answer — is the internal champion who becomes the CAIO. The IBM study confirmed it last month: 57% of CAIOs were promoted from inside. Not hired. Promoted.

The 80% figure just gave you the argument you needed for why this is urgent. Use it. Not as a thought piece to share on LinkedIn, but as the opening frame for the conversation you should be having with your executive team before the next board meeting.

And if you're the person who needs to make that case and doesn't yet have a working AI artifact to show when the conversation starts — the June cohort starts this week. That is the deliverable you'd leave with.

The Anthropic engineers didn't convince their leadership team with a presentation about AI potential. They shipped 8x, and the data spoke for itself.

That's still the move.

———————————————————————————

The 80% figure maps directly to three moments in Closing the AI Wage Gap:

Chapter 3 — The Multiplier is built around the same question the Anthropic data answers empirically: what does 8x output look like, and who is actually achieving it? The chapter profiles the 14% of mid-career executives who have restructured their work around AI leverage — not by using more tools, but by moving AI from assistant to primary producer. The Anthropic figure is the organizational-scale proof of the individual-scale pattern I documented across 2,300+ coaching engagements.

Chapter 9 — The Internal Champion covers the 57% stat in detail: why the path to the CAIO title runs through internal proof cases, not external credentials, and what it takes to build one deliberately. This week's data gives the chapter its sharpest possible opening argument.

Chapter 15 — The Recursive Advantage deals with compounding — what happens to the output gap between operators who use AI seriously and those who don't, over six months, twelve months, two years. The Anthropic 52x figure on optimization benchmarks is the clearest published data point for what the tail of that compounding curve looks like.

Manuscript in Tier-1 agent querying — Levine, Halpern, Sagalyn pending. If you know an editor or agent at the intersection of work, AI, and organizational economics, reply to this email.

———————————————————————————

The best six from a week that had sixty. Full daily version coming — see below.

Claude Opus 4.8 ships with dynamic workflows. Hundreds of parallel subagents in a single session, mid-task instruction updates, 2.5x speed mode, 1M token context, 128K output. The practical impact: workflows that used to require human handoffs between steps can now run end-to-end. If you're still running Claude on individual prompts, you're using it like a calculator.

57.2% of internet traffic is now automated. TAAFT confirmed this week that bot and agent traffic crossed the majority threshold for the first time — more than half of all internet activity is now generated by non-humans. SEO, content monetization, API pricing, and the value of verified human attention are all affected. If your business model depends on traffic volume, the definition of "traffic" just changed.

John Peslar built 17 drop-in Claude Opus 4.8 SDR skills. Research, prospect, write outreach, book meetings — all running in parallel, no human SDR in the loop. 137 prospects researched, 41 replies, 9 meetings in 4 days. This is the most documented, most concrete example of AI replacing a revenue-generating human role I've seen published. The playbook is free at johnpeslar.com. Worth reading even if sales automation isn't your function — because the pattern generalizes.

Meta Business Agents went global. WhatsApp, Instagram, and Messenger now support AI agents that close sales, book appointments, qualify leads, and recommend products — across languages, at scale. 1 million+ businesses already live. If you have clients or portfolio companies running customer acquisition through messaging, this is live infrastructure, not a roadmap item. meta.ai/business

Life-Harness: 88.5% agent performance boost without touching the model. New open-source framework that improves AI agent performance across 116 model combinations purely by fixing the runtime wrapper — the translation layer between model and environment. No retraining required. The implication: the biggest gains in the near term are coming from better infrastructure around models, not bigger models. GitHub: Life-Harness

First AI-designed vaccine enters human clinical trials. TAAFT confirmed this week that a vaccine designed autonomously by AI has cleared the preclinical phase and is now in human trials. This is not a product launch or a demo. It is a clinical trial milestone in the most regulated, most scrutinized domain in science. Physical AI has left the screen.

———————————————————————————

Each week: one real tool I built, the stack behind it, what it took, what changed. This is a free preview — the full breakdown series, covering every tool in the Portfolio Leverage Co. stack with copy-paste architecture and adaptation guides, is coming as a paid add-on.

The problem had a specific shape. I was working with a Fortune 500 CHRO whose HR team was fielding the same eight to twelve policy questions roughly two hundred times a week. Parental leave calculations. PTO carry-over edge cases. State-specific termination procedures. Benefits eligibility after a role change. The questions were slightly different each time — different employees, different circumstances, different policy years — but the underlying structure was the same. Senior HRBPs were spending around forty hours a week on recurrence.

That's not a staffing problem. That's an architecture problem.

The solution I built is called AIHRPilot. Here's the stack: Flask for the backend, scikit-learn with TF-IDF vectorization for ticket classification, Claude API for handling the actual policy interpretation, and HTMX for a lightweight UI that HR could use without training. No vector database. No fine-tuning. No infrastructure team needed. About forty dollars a month to run.

I built the first version in a weekend. Three weeks later it was in production.

The classification layer — TF-IDF — handles the routing problem: which of the eight to twelve recurring question types is this? Once it's routed, Claude handles the interpretation. That's the critical handoff. The reason this works where a rule-based system wouldn't is that HR policy questions are almost never clean. They involve edge cases, exceptions, recent amendments, employee-specific context. A decision tree collapses on contact with reality. Claude doesn't, because it can hold the ambiguity and reason through it the way a senior HRBP would — just faster and at any hour.

The result: forty hours a week of HRBP time went to six. The six hours that remained were the genuinely complex cases that benefited from human judgment. The other thirty-four hours freed up for actual people work.

Three things made the difference that I didn't expect going in. First, the system prompt design — not the model choice, not the stack, the prompt. Getting Claude to consistently cite the relevant policy section and flag when it wasn't certain took more iteration than the code did. Second, the handoff logic: defining precisely when the system should auto-answer versus route to a human for review was a judgment call that required real HR expertise, not engineering. Third, cost. At forty dollars a month, the ROI conversation takes about thirty seconds.

The full build walkthrough — including the six-phase architecture, the four core Claude prompts, and an adaptation table for candidate screening, RFP scoring, board prep, deal flow, and compliance Q&A — is at portlev.com/build/how-i-built-aihrpilot.

Next week: DueDrill — the AI due diligence platform I built for my own VC work. Sixteen DD categories. Two hundred and fourteen data fields. Goldman-style PDF memos, auto-generated. Built because I was tired of grinding through deal packets at eleven PM.

The full Build Breakdown series — every tool in the Portfolio Leverage Co. stack, the exact architecture behind each one, and how to adapt it for your org — is coming as a paid add-on to The Leverage Brief.

———————————————————————————

The research that feeds this newsletter — AlphaSignal, Rundown AI, TAAFT, AI Fire, Superhuman Code, and four LinkedIn operators I track daily — generates 50–60 items a week. The Leverage Brief carries six.

I'm building The Leverage Signal: a five-minute weekday morning read, calibrated for the same audience as this newsletter.

Who it's for: Portfolio Executives, CHROs, CTOs, CAIOs, CLOs, and fractional executives in AI-exposed roles. Founders and operators running AI-leveraged businesses. Executive coaches, L&D leaders, and organizational designers whose clients are navigating AI transformation. People who have moved past "I've tried some tools" and want daily intelligence calibrated to actual work, not product hype.

What it covers: The two or three highest-signal AI tool or agent releases from the prior 24 hours, with a one-sentence "what this means for your work" framing. One open-source repo worth knowing. One workflow or deployment pattern from the operator community. One macro signal — model, funding, regulatory, talent — that belongs in your working context before 9 AM.

With enough subscribers, The Leverage Signal gets auto-drafted from the same Notion research catalog that feeds this newsletter and sent daily. I want to know if the readership is there before I build the pipeline.

→ SUBSCRIBE TO THE (M-F) LEVERAGE SIGNAL — Ten seconds to opt in.

———————————————————————————

Paid AI training and evaluation work for senior operators. Flexible, remote, NDA-bound.

The expert-data market is tracking toward $100B/year by 2027. Frontier labs are paying domain experts for post-training evaluations, RLHF, and agent environment design. Senior operators and licensed professionals in HR, finance, legal, medicine, and engineering consistently land in the $50–$150/hr band.

Mercor — Premium rates ($75–$150+/hr for qualified domain experts). Strictest screening, best fit for CHROs, attorneys, physicians, and senior engineers. Valued at $10B after October 2025 Series C. Note: impacted by a March 2026 supply-chain attack — review their post-breach disclosures before onboarding. → Apply via referral link

micro1 — Faster onboarding via the Zara AI interview; multiple attempts allowed. Crossed $100M ARR December 2025. Expanding into robotics pre-training and agent simulation. $20–$150/hr depending on domain. → Apply via referral link

Meridial (by Invisible Technologies) — Expert contractor work in law, STEM, finance, linguistics, coding, and safety. No prior AI training experience required. Typically hear back within 48 hours. Strong fit for specialized domain experts. → Apply at meridial.ai

Apply to all three. A $50–$150/hr income node that builds on expertise you already have.

———————————————————————————

The July cohort starts in 3-4 weeks. The timing is deliberate.

The 80% figure is the board-level argument. The cohort is where you build what you bring to that board meeting.

Fifteen seats. Twelve weeks. Three things you leave with: a redesigned operating week built around compounding output rather than calendar entropy; a custom AI workflow or tool you actually ship in your real work — not a prototype, a working artifact with before-and-after data; and a positioning narrative that names your value at the convergence of talent and technology, because that's where the CAIO titles are going.

The Anthropic engineers didn't need a cohort. They had access to everything. You don't have to work at Anthropic to get to 8x. You have to restructure how your week runs and ship one thing that proves you did.

→ Apply: portlev.com/cohort

———————————————————————————

Find the 80% equivalent in your own work.

Pick one category of output your team produces regularly — reports, analyses, proposals, responses, documents, whatever is most recurring and most time-consuming. Estimate what percentage of that output an AI agent, running with the right context and a well-designed workflow, could produce as a first draft requiring only your verification.

Write that percentage down. Then write what the headcount or time equivalent of that percentage represents in your org.

That number is your AI Wage Gap — the gap between what you're currently doing and what the Anthropic engineers are doing. It has a dollar value and a competitive implication.

Now: what would it take to close 20% of that gap by the end of June?

That's the question. The answer is the project.

———————————————————————————

Reply to this email. I read every one.

Yuri Kruman Founder, Portfolio Leverage Co. · 3x CHRO · AI Trainer: OpenAI, Meta, Microsoft PortLev.com · LinkedIn · AIWageGap.com

———————————————————————————

"80% of Anthropic's production code last month was written by Claude. Their engineers are shipping 8x what they shipped two years ago. The technology isn't coming. It's already doing most of the work at the lab that built it."

— The Leverage Brief, June 2026

———————————————————————————

© 2026 Portfolio Leverage Company. You're receiving this because you subscribed to The Leverage Brief. Unsubscribe in one click.