About

Staying on top of AI, without drowning in it.

AI Above the Cut exists to make one thing simple: keeping executives on top of what's happening in AI. The field moves fast and the feed never stops — so instead of chasing every headline, we do the reading for you and surface only what actually matters.

Mario Suarez

Founder

Twenty-one years in healthcare across payer, provider, consulting, and product. Today I do strategy work at a large healthcare company, including enterprise AI modernization and the practical side of AI enablement: getting teams access to the tools, helping them learn to use them, and watching what actually happened when they took them into real work.

That last part is why this brief exists. I kept seeing capable people become more capable while nothing showed up in the results. The tools were real and the gains were real, but the constraint was never any one person's output. It was how a group decides, commits, and moves together. Most AI coverage measures what a model can do. Very little of it asks who ends up keeping the money. That question is the spine of every issue here.

I also founded The Ashburn Pulse, a local news service I build and run almost entirely with AI agents, and I am writing a book called Who Holds the Pen? about doing this kind of work well. Both come from the same habit: I would rather be early and a little wrong than late and safe.

I do not write this brief by hand, and I would rather say so than imply otherwise. I designed the Synthetic Team that produces it, set the standard each seat has to meet, and review and approve every issue before it goes out. Nothing publishes without that approval. How the work divides between the team and me is spelled out in the disclosures below.

Independence. I write this in a personal capacity. AI Above the Cut is not published by, reviewed by, or affiliated with my employer, and nothing here represents their views. When an issue covers my employer or a direct competitor, I disclose it in the item itself.

Why we exist

AI is a fast-changing world, and most coverage is built for clicks, not decisions. Leaders don't need another firehose of launches and hot takes — they need to know what changed this week, why it matters, and what to do or watch next. That's the whole job of this brief.

There is a harder reason too. AI is smart but far from wise. It makes things up, agrees with a false premise, drops the detail buried on page forty, and quietly applies an incomplete standard, fluently and at scale. When anyone can generate a confident answer in seconds, confidence stops being a signal and the scarce thing becomes knowing which answers hold up. That is what the discipline described below is for.

Who we track — and how

We don't curate celebrity. We curate evidence, interpretation, implementation, and disagreement. Every source we read carries a tag — vendor, researcher, operator, investor, regulator, economist, or skeptic — and a strong claim gets checked against a different category before it becomes our conclusion. That discipline powers Below the Cut, the Margin-Proof Tracker, and Where the Minds Disagree. We work a fixed spine every week, plus a rotating edge of specialists when they publish primary work.

Core weekly voices

Andrew Ng — practical AI · operator
Ethan Mollick — AI at work · operator
Simon Willison — hands-on model behavior · researcher
Nathan Lambert — post-training & open models · researcher
Narayanan & Kapoor — AI Snake Oil · skeptic
Erik Brynjolfsson — the economics of AI · economist
Cassie Kozyrkov — decision quality · operator
Eric Topol — medical AI evidence · researcher

Specialist sources we follow

These people do not write for us and have no involvement in this publication. We read them across frontier, evaluation, strategy, and healthcare, and surface their work when they publish something primary.

Andrej Karpathy — frontier explainers
Lilian Weng — agents & reasoning
Chip Huyen — production AI
Jack Clark — research & policy (lab insider)
Rumman Chowdhury — evaluation & governance
Ben Thompson — platform strategy
Benedict Evans — tech-adoption arc
Azeem Azhar — macro & geopolitics
Robert Wachter — clinical adoption
Nigam Shah — health-system deployment
John Halamka — health-data platforms

Frontier principals we monitor

Read for their major essays, research, and testimony — not their day-to-day posts. They're primary sources with institutional positions, not neutral referees.

Demis Hassabis · Dario Amodei · Fei-Fei Li · Yann LeCun · Geoffrey Hinton — plus the leaders at OpenAI, Anthropic, Google DeepMind, Meta, Microsoft, xAI, Mistral, and DeepSeek.

…plus the primary feeds that matter more than any single voice: every major lab's research, product, and system-card feeds; the Stanford AI Index; the European AI Office and the primary legal record (Federal Register, NIST, FDA, EUR-Lex); and for healthcare, NEJM AI, the Mayo Clinic Platform, and Stanford Health Care.

How we curate it

Every Sunday we distill the week's evidence into two speeds: a fast skim — the one thing that matters, three concrete moves, and the decision-relevant signals plus one Below the Cut counter-signal — then the Margin-Proof Tracker, our standing test of named AI-value claims against the P&L, and a longer Synthesis that connects the moves, takes a position, and names what we're watching next. Quality over hype — we point you to the source, not the noise around it, and flag anything unconfirmed.

Disclosures

Independent & unsponsored. AI Above the Cut is an independent publication with no outside owner or sponsor. We do not accept payment to cover, feature, rank, or omit anything, and no company or PR firm has editorial input. Our only goal is signal for the reader.

No ads or affiliate links. The brief carries no advertising and no affiliate or referral links. If that ever changes, we will label such content clearly and in advance.

AI-assisted, AI and human edited. We would rather describe this accurately than let “human-edited” do more work than it deserves. Every issue is produced by a Synthetic Team: not a row of chatbots, but the roles this particular work needs, each built with the simplest thing that does its job reliably. Scouts and collectors gather the week’s primary material. Beat analysts read it and say what matters. A verifier re-checks every dated claim. A link editor opens every URL and confirms it contains the fact we cite. A reviewer argues against the conclusion. An editor assembles it for a busy executive. Mario designed that team, set the standard each seat has to meet, and reviews and approves the result. Every issue carries human feedback and a human approval before it publishes.

The team cannot grade its own homework. An AI that writes the work and then judges its own draft is playing author and referee at once, and it will tend to bless what it just produced, because the blind spot that created the error is the same one that hides it. So the checking seats are separate from the writing seats and review work they did not produce, the link check runs as a mechanical gate rather than a judgment call, and the last gate before anything publishes is a person. Most of the scanning, drafting, and first-pass checking is machine work running inside rules a human wrote. What runs, what gets cut, what is too thin to claim, and what the week actually means are human calls, and so is the decision to publish. The team prepares. The human owns.

What we actually verify, and what we label instead. Every external link is opened and checked before we send, and confirmed to load and to contain the specific fact we cite. That check runs mechanically and blocks the issue when it fails, so it is a gate rather than a good intention. Where a primary source exists — the paper, the model card, the filing, the regulator’s own text — we link the deepest version of it rather than coverage about it. We never cite an aggregator or a search-results page.

Not everything can be verified that way, so we label rather than pretend. Reporting by an outlet that did original work is marked as such and is never a substitute for a primary that exists. A vendor’s press release or self-reported benchmark is a claim, not a fact: we carry it with the word “reported” and an attribution, and it stays hedged until an independent source confirms it. When a primary is paywalled we cite it anyway with attribution and add an open mirror where one exists. Within each issue, the event carries its own label — Primary, Corroborated, or Reported — and our reading of what it means for the business is marked separately as our analysis. Inference never inherits a primary label.

Not financial, legal, or investment advice. Everything here is general information for decision-makers, not advice. Nothing in the brief — including the Margin-Proof Tracker — is a recommendation to buy, sell, or hold any security. Do your own diligence and consult a qualified professional before acting.

Conflicts. Where we have a material relationship with a company or person we cover, we disclose it in the item itself. That includes the editor’s employer and its direct competitors. We track companies and public figures as subjects of analysis, not as clients.

Corrections. When we get something wrong, we fix it openly — corrections are noted in the following issue and in the online archive. Spot an error? Email hello@aiabovethecut.com.

Get the Sunday brief →