A DTC brand briefed Viktor inside Slack: one landing page per Meta ad group, mapped to a different headline variant. He wrote the code, deployed each page to their subdomain, posted the URLs back in #marketing, and now monitors performance across the set.

Their content team uses him to draft email flows, generate creative variants, and audit Klaviyo segments every Friday. Their growth lead uses him to catch spend anomalies before the day starts.

20,000+ teams now have the same setup: one AI employee across every marketing tool. A teammate who ships work in Slack and Microsoft Teams. SOC 2 certified.

"Not only have we caught up on several months of work, we are automating manual tasks and expanding to things previously not possible at scale." Jesse, Director, Torque King 4x4.

Get Started for Free

Sunday, June 28, 2026  Intelligence for Portfolio Executives closing the AI Wage Gap.

This week AlphaSignal ran a deep dive on a voice tool called Wispr Flow. The number that traveled was the speed: people talk at roughly 220 words a minute and type at about 45, so voice input runs four times faster than the keyboard. It's already in use at OpenAI, Vercel, Replit, and Warp. Tidy productivity story.

But the line that mattered wasn't about words per minute. It was Andrej Karpathy's, quoted in the coverage: expressing your goals to the agent is now the bottleneck.

Not the model. You.

Sit with that, because it quietly closes an argument we've been having for two years. The entire enterprise conversation about AI has been a referendum on the model. Is it good enough? Is it accurate enough? Can we trust it with real work? Every pilot, every steering committee, every "let's wait for the next version" was a bet that the constraint lived inside the machine. The constraint was the model's capability, and the responsible thing was to wait for it to improve.

It improved. Quietly, across every function, the frontier models started clearing the bar. Last week I wrote about the 400,000-session study where lawyers and managers kept pace with engineers — the moat moved from coding skill to domain judgment. This week's signal is the next link in that chain, and it's more uncomfortable. Because if the model is good enough, and your domain judgment is the moat, then the thing standing between you and the output is no longer the technology. It's the bandwidth and precision with which you can move what's in your head into the machine.

That's the bottleneck. And it's sitting in your chair.

———

Wispr is the literal version of the problem — your fingers can't keep up with your thoughts. But typing speed is the small version. The real bottleneck isn't how fast you type. It's that the machine doesn't know anything about you when you start.

Every time you open a blank chat, you re-explain. Who you are. What the company does. What you're working on this quarter. Your standards, your voice, the three constraints that make this decision different from the textbook version. You spend the first several minutes of every interaction rebuilding context the machine had no way to remember — and then you do it again two hours later, from scratch. That re-explanation tax, paid ten times a day, is the actual gap between you and the people getting 8x. It is not a model problem. You and the Anthropic engineers have access to the same models. The difference is that they've collapsed the distance between intent and execution, and most executives haven't even measured it.

Picture the tax in a real week. A CHRO opens a chat to draft a sensitive termination memo and spends four minutes laying out the state, the tenure, the policy year, the tone the CEO prefers. An hour later, a board-update request — and she re-types who's on the board, what they care about, what last quarter's narrative was. After lunch, a vendor evaluation, and again she rebuilds the criteria from memory. None of that is the work. It's the throat-clearing before the work, and she pays it every single time because the machine started the day knowing nothing about her. Multiply that across a function, across a quarter, and you're looking at hundreds of hours spent telling a brilliant system things it should have already known.

And here's where it meets the wage gap. The premium the market is paying isn't for access to AI — everyone has access. It's for leverage, and leverage is exactly what the re-explanation tax destroys. Two executives with identical judgment and identical tools produce wildly different output if one of them has front-loaded their context and the other re-enters it by hand all day. The gap between them isn't talent or tenure. It's setup. That's the uncomfortable, liberating truth underneath Karpathy's line: the highest-leverage hour of your month right now is probably not spent on strategy. It's spent teaching the machine, once, the things you currently re-explain a hundred times.

———

Here's the part that should change how you think about this. The bottleneck being you is the best news in the entire AI story.

A model limitation is something you wait on. You file a ticket with the future and hope the next release fixes it. But a context-and-configuration problem is something you can solve this weekend, with no budget, no engineering team, and no permission. The constraint moved from a thing you can't control to a thing you entirely can. That's not a downgrade. That's the unlock.

What does solving it actually look like? It's not a personality trait and it's not technical aptitude. It's configuration. A project structure that holds your standing context, so you stop re-introducing yourself ten times a day. A set of custom instructions that encode how you think, what "done" looks like, and the things you never want to see again — so every output lands closer to final on the first pass. A small library of prompts for the four or five high-value moves that recur every week — board prep, a vendor evaluation, a hard email, a first draft — built once and run on demand. And the connections to the tools your real work already touches, configured and tested rather than left as a someday. Configure once. Compound forever.

None of that is exotic. All of it is skipped, because it isn't urgent and nobody assigns it to you. The executive who does it operates at a different clock speed than the one still typing context into a blank box — and the gap widens every week, because the configured operator's setup keeps paying off while the other keeps paying the tax.

So three moves this week. First, measure the tax: for two days, count how many times you set the same context from scratch. The number will annoy you, and the annoyance is the point. Second, build the standing-context layer once, even a rough version, for the one slice of your work that eats the most time. Third, shift your default from typing at the machine to directing it — start treating setup as the work, not the thing you'll get to after the work.

———

I've spent the last two years building exactly this layer for executives and for my own businesses, and it's most of what I do all day. So starting today, I'm opening the part of this newsletter where I show the actual builds.

It's called The Build Vault, and it's the premium tier of The Leverage Brief. Every week I take one real system from the Portfolio Leverage Co. stack and break it down to the studs — the project structure, the actual Claude instructions, the prompt library, the field schemas, the connections — copy-ready, plus the walkthrough. Free readers get the map. Members get the build. The first one is live below, and it's the obvious place to start: the exact Claude configuration that turns a $20 chatbot into something that runs like a chief of staff.

The model stopped being the thing between you and 8x. You are. The Build Vault is where you stop being it.

This week's signal maps onto three chapters of Closing the AI Wage Gap:

Chapter 3 — The Multiplier is the chapter this week's data validates most directly. Its argument is that 8x output never comes from a better model — it comes from a domain expert who has configured AI to hold their context and standards, then moved it from assistant to primary producer. Karpathy's "expressing goals is the bottleneck" is the one-sentence version of the whole chapter: the multiplier is the configuration layer, and almost nobody builds it.

Chapter 9 — The Internal Champion is about who gets the CAIO title, and why it's the person who ships a working setup rather than the one who waits for the perfect model. The 57%-promoted-from-inside statistic lives here. This week sharpens it: when the bottleneck is configuration rather than capability, the champion is simply whoever did the unglamorous setup work first.

Chapter 15 — The Recursive Advantage is the compounding chapter, and "configure once, compound forever" is its thesis in four words. The operator who builds the standing-context layer pulls away from the one who re-explains himself daily — not linearly, but on a curve. Over twelve months that gap stops being a productivity difference and becomes a category difference.

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 joined Slack as a coworker. Anthropic shipped a native integration that drops Claude inside Slack as an enterprise teammate, alongside Managed Agents and the new Claude Tag — capped by its Openday 2026 event. The read: Anthropic isn't trying to beat ChatGPT, it's trying to become Salesforce. Claude is becoming the infrastructure your organization standardizes on, which means the time to own that standard internally is before procurement does it without you.

OpenAI built its own inference chip in nine months. Codenamed Jalapeño and co-designed with Broadcom, it already powers GPT-5.5 Instant and is a direct shot at Nvidia dependence. Compute is being verticalized by the labs themselves. For anyone whose roadmap assumes stable model pricing and availability, that assumption is now load-bearing on a supply chain that's actively being rebuilt — plan for movement.

GPT-5.6 got a government leash. The Rundown reported tightening federal guardrails on frontier models this week. The CAIO just quietly became the Chief Compliance Officer too. If you're the person responsible for AI in your org, model your regulatory exposure now — governance has moved from a slide to a board agenda item in about a quarter.

Alibaba allegedly scraped 28 million Claude queries. Three separate stories from one actor this week — mass scraping, alleged model cloning, and a grey-market reselling Claude API access. Model IP is becoming a geopolitical problem, not just a competitive one. The operator takeaway is narrower and immediate: the AI vendors in your supply chain carry IP and security risk you should be diligencing, not assuming away.

Wispr Flow put voice-to-text into the IDE — 4x faster than typing. Voice input with an AI cleanup layer, already adopted at OpenAI, Vercel, Replit, and Warp, $12/month. The input layer is the next productivity unlock, and the keyboard is the thing being unlocked away. If your team produces a lot of text, the interface — not the model — is where the next hour gets saved. wispr.flow

The $0 agent stack got closer. n8n shipped a built-in AI Agent builder with memory, tools, and guardrails, while OpenCode went fully open source. The open agent toolchain is solidifying: orchestration, execution, and observability are now assemblable for free. The build-vs-buy math just shifted — a chunk of what a $50K automation agency sells is increasingly something a configured operator wires up in a weekend. n8n.io

Six items from a week that had sixty. Full daily version coming — see below.

Starting today, the Build Breakdown has two halves. Free readers get the map — the architecture, the order, the logic. Build Vault members get the build itself: the actual files, the instruction skeleton, the copy-ready prompt library, and the walkthrough video. This is the first issue of the premium tier. Welcome to The Build Vault.

Most executives I work with are paying for frontier AI and getting a smarter search box. They open Claude, ask a question, get a decent answer, close the tab. That's maybe 10% of the thing.

The gap between default Claude and a properly configured Claude is the gap between a tool and a chief of staff — something that knows your business, your voice, your standing context, and your standards, and applies all of it the second you ask. Closing that gap isn't about being technical. It's about knowing what to build, in what order. That's most of what I do all day, and it's the entire reason this newsletter now has a paid tier.

Here's the free version: the map. Four pieces, and the order matters more than any single piece.

1. A project structure that holds your context. Default Claude forgets you between chats. The fix is a small set of projects, each loaded with the standing context for one part of your work — your company, your role, your portfolio, your writing. You stop re-explaining yourself ten times a day.

2. Custom instructions that encode your standards. This is where you tell Claude how you think, what you never want, and what good looks like. Done well, every output lands closer to final on the first pass.

3. A prompt library for your recurring moves. The same handful of high-value tasks come up every week — board prep, vendor evaluation, a sharp reply, a first draft. Build each once, then run it.

4. The connections that matter for you. Calendar, mail, drive — whatever your real workflow touches, configured and tested, not left as a someday.

That setup replaced what used to be six to eight hours a week of context-setting and rework. That's the whole game: configure once, compound forever. It's worth a Saturday.

The free version of this issue gives you the map. The build itself — the actual project files, the custom-instruction skeleton you paste and fill, the copy-ready prompt library, the refreshed connections and 2026 add-on stack, and the screen-share walkthrough — is below, for Build Vault members.

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

The Leverage Signal is the five-minute weekday read for the same audience: the two or three highest-signal tool or agent releases from the prior 24 hours, each with a one-sentence "what this means for your work" framing; one open-source repo worth knowing; one deployment pattern from the operator community; one macro signal — model, funding, regulatory, talent — in your working context before 9 AM.

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 and L&D leaders whose clients are navigating AI transformation.

I want to know 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 pay 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 its 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 in 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 across law, STEM, finance, linguistics, coding, and safety. No prior AI training experience required. Typically responds within 48 hours. Strong fit for specialized domain experts. → Apply at meridial.ai

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

The July cohort starts in under three weeks. Three seats remain.

This week's signal is the cohort's argument in one line: the bottleneck is no longer the model, it's your ability to direct it — and that's a buildable skill, not a trait. The cohort is the room where you build it against your real work, with a working artifact and before-and-after data to show for it.

Fifteen seats, three left. 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; and a positioning narrative that names your value at the convergence of talent and technology, because that's where the CAIO titles are going.

You don't close your gap by waiting for a better model. You close it by configuring the one you already have and shipping proof that you did.

→ Apply: portlev.com/cohort

For the next two days, keep a tally. Every time you open a fresh AI chat and start by explaining who you are, what your company does, or what you're working on — make a mark. Don't change anything yet. Just count.

On day three, look at the number. That's your re-explanation tax, and it's the clearest measure of the bottleneck this week's lead is about. Now take the single context block you typed most often — your role, your company, your current priorities — and paste it once into a Claude Project's instructions. That's it. That's the first brick of the configuration layer.

You just converted a tax you were paying daily into a setup you built once. Do it for one slice of your work this week. Then watch how much closer the first draft lands.

• The Build Vault (premium) — one real build, broken down, every week + full archive + member sprint pricing. Join
• Portfolio Executive Cohort (July 2026) — 3 seats left. Apply
• The Leverage Signal (daily briefing) — SIGN UP HERE
• The Claude Setup sprint — $4,000 ($2,800 for Build Vault members). Reply "Claude Setup."
• Custom AI Build — from $5K. Scoping calls open now (BIG DISCOUNT FOR THE BUILD VAULT ANNUAL SUBSCRIBERS)
• Fractional CHRO / CLO — $15K/mo. Two Q3 slots open.
• Pre-order Closing the AI Wage Gap — portlev.com/preorder

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

TODAY’S BUILD BELOW (SUBSCRIBE TO ACCESS FULL BUILD BREAKDOWN IN THE BUILD VAULT)