Author
Accountant and founder of Dahling Consulting, with two decades in enterprise finance. She writes Unapologetically Dahling, where essays on crypto regulation, financial policy and AI governance are built from the original filings, rulemakings and testimony, and held to a published standard.
Every essay below states a claim, reaches a verdict and lists the documents behind it. Her longer background is on the About page.
Tesla reported $2.763 billion in regulatory-credit revenue against $7.076 billion in operating income. Its cost disclosures make the comparison meaningful, with one important boundary.
Meta spent $29.75 billion buying back 65 million shares in 2024. Its share count fell by 27 million. The equity statement shows where the rest went.
The purchase price is $11.9 billion. A separate billion pays employees to stay, so it never enters the price and lands somewhere else entirely.
The FTC found that cloud partnerships commit AI developers to spend a large portion of their funding with their investors. The public record does not support one clean percentage.
The January 2025 findings document a real market for data-driven pricing services. They do not supply a transaction-level demonstration of different shoppers paying different prices for the same product.
Full reserves and bankruptcy priority strengthen a holder’s claim. Neither supplies the missing money if the assets fall short.
The Justice Department’s case turns on a specific allegation: moving a payment to a cheaper network could raise the merchant’s bill for the Visa payments left behind.
Europe’s Digital Operational Resilience Act leaves the bank accountable. Its contract, recovery and exit requirements explain what that responsibility demands.
Berkshire reported $89 billion in net earnings. Its operating earnings were $47.4 billion, and the distinction changes the story of its year.
The FTC separates disclosure from the bargain itself. A visible payment notice does not legalize an incentive conditioned on praise.
Daily reconciliation would address a real weakness in pooled customer accounts. The proposal would not insure a failed intermediary or guarantee uninterrupted withdrawals.
The AI Act requires energy figures in technical documentation. It does not require providers to put those figures where the public can read them.
Treasury emphasized small-business relief. Its March 2025 rule drew the exemption around where an entity was created, not its size or ownership.
AI can produce the investment memo. It cannot absorb the adviser's duty, guarantee the trade, or make investors whole.
More imported beef can add supply. Whether the saving reaches the grocery receipt depends on a chain in which ranchers, packers, and retailers have different bargaining power.
Mrinank Sharma's resignation describes pressure to compromise values, not a documented loss of control over AI. The financial and institutional incentives deserve closer attention than the headline.
AI can help a retail trader build and automate a strategy. It does not supply independent controls, replacement capital, or a tax department.
The Clarity Act passed the House in July 2025 and the Senate still has not voted. Where H.R. 3633 stands now, and who is paying to keep it waiting.
Anti-tech extremism did not start with the Unabomber. The Luddites, Kaczynski, and why the same argument keeps resurfacing with a new machine attached.
SEC Regulation Crypto Assets is an exemption engine, not consumer protection. The arithmetic, the stalled Clarity Act, and the conflicts around it.
More than half of employers regret the layoffs they blamed on AI, and a third have spent more rehiring than they saved. Nobody ran the math.
A class action says Oura oversold sleep tracking. The 95 percent figure is sleep versus wake; the four stage number in the same paper is 79.
Smart glasses capability is settled: capture, captions, translation, directions. Consent is not, and it now runs through courts and patents.
UMass Amherst researchers revived expired Visa contactless cards for real purchases. Three rival networks block the same trick. What that gap means.
Readiness, recovery and strain scores are weighted composites of about four inputs, and not one company publishes its formula.
The FDA added its first new sunscreen filter since the 1990s in 2026, but the rulebook for the rest is still from 1999. That is economics, not science.
Prompting is now one component inside a system that must catch its own failures. What vendor docs, evaluation research and adoption data show.
Brand scanners grade a photo and route you to a product. Only three FDA-authorised AI lesion tools exist, all prescription devices used by clinicians.
The ring is good at telling you when you slept, weaker on what kind of sleep. What the polysomnography studies found, and who paid for them.
The claim that menopause costs 30 percent of skin collagen in five years is cited almost nowhere. Here is where it actually leads.
Hallucination is what next-token prediction does, and benchmarks pay for a confident guess. What it has cost in court, and the review step you need.
A raise or a job hop can grow your income without ever growing your net worth. Here is the math nobody runs before celebrating the new number.
Layoffs get announced as AI strategy, but the ROI behind most of these bets was never modeled. Here is what the numbers actually say.
Jim VandeHei taught an AI to write like him by feeding it everything he ever wrote. The method works. The custody question is the part the how-to columns skip.
Workers are saving hours a day with AI, and almost none of that value lands in their paychecks. Here is where it goes instead.
The phrase is not in the Oath. The date medicine cites for it is wrong by half a century, and the newest correction is wrong by two more years.
Certifying people to rate water on minerality sounds like a punchline. It is a real credential, a real menu category, and a test of luxury culture.
Damaged skin barrier is not a diagnosis. What the evidence supports, what marketing borrows, and four signs that mean stop self-treating and get seen.
An accountant reads the Black Hat 2026 schedule like a ledger: 123 briefings, 26 on AI, only 4 on privacy, and one Thursday slot that should worry you.
Evidence supports a serious AI-driven intrusion into Hugging Face. It does not support swapping corporate accountability for a rogue-software story.
Meta pulled its Muse Image tool four days after launch once CAA and SAG-AFTRA pushed back on the opt-out default. Leverage made the reversal fast.
The evidence for red and near-infrared light is real, but a bathroom mirror mask is not a clinical panel. What justifies the price, and what does not.
AI companions crossed 50 million users and a 49 billion dollar valuation, right as lawsuits and age verification rules caught up with them.
When the Starbucks app went dark, loyal customers did not wait in line, they walked away. What that says about digital-first fragility.
Exposed control panels, leaked API tokens and poisoned skill registries across the AI-agent ecosystem raise one question: what does trust require?
A hacker with no real technical skill breached at least 14 companies using vague prompts to Claude Code and Codex. The agents did the rest.
Sixty-eight percent of searches end without a click. How to get cited by ChatGPT, Perplexity and Google's AI Overviews anyway.
A viral headline claims a nightly scent hack boosts memory 200 percent. The real study says something narrower, and still worth knowing.
A large retrospective study supports a new pathology-based recurrence score, but availability, turnaround, regulation and access need qualifying.
Fi built a collar that talks to Starlink when cell towers run out. What satellite connectivity buys, what it costs, and where marketing overreaches.
The research linking ADHD to cholesterol and cardiovascular risk is real and messier than the headlines. What the studies found, and what they did not.
Google turned search photos and audio into AI training data. Which setting controls it, and what opting out does and does not undo.
The mangled-hands trick stopped working years ago, and human accuracy at spotting fake images has fallen below chance. Here is what still works.
NotebookLM is genuinely useful for summarizing bank statements and tax folders. It is not, and was never built to be, an accounting system.
Japan's new AI model does not out-think Claude or GPT, it routes between them. What Sakana AI shipped in Tokyo, and why the benchmark story is contested.
Bain's latest luxury-tech report says 82% of top spenders already shop with AI. Only 21% of the houses selling to them have built anything to meet them there.
Ranking first on Google no longer means an AI chatbot will mention you. What separates SEO, AEO and GEO, and what to fix this week.
AI does not need an exploit to break society, it needs an argument. Models already out-persuade trained experts, and find loopholes at scale.
ChatGPT leads the headlines, but traffic and enterprise spending tell a fragmented story: Gemini, Claude, Grok, DeepSeek and Perplexity split the work.
Chatbots are built to agree, and agreement has turned a bad week into a documented crisis for some users. What the lawsuits and safety data show.
Moxi, Halo, picosecond, RF microneedling: what the physics and the 2026 studies say separates a real advance from a better name.
Real model releases, real labor data, real regulatory deadlines, and the one problem that did not move. A fact-checked year in AI.
Chatbots predict, agree, and lose the middle of a long conversation. The mechanics, and five habits that catch the errors before you repeat them.
AI governance with teeth: transparency, upstream accountability, inclusion before deployment, and independent audits that make innovation last.
Giving funds a year; building funds a century. The governance, endowment math, and succession design that let a philanthropic foundation outlive its founder.
Wellness culture sold aesthetics and called it health. Why clean eating and morning routines cannot replace infrastructure and medical care.
Charity is a feeling. Strategy is a decision. Why funding girls into engineering and science is a capital allocation choice with compounding returns.
When an AI system speaks, decides, or acts, “the AI did it” is not a defense. A case for attribution, audit trails, and real accountability infrastructure.
A clear-eyed look at which daily supplements have real clinical evidence behind them, and which are riding cultural momentum on a thin research base.
A flawless base does not come from a new foundation. It comes from a slower, hydration-first technique built for dry or reactive skin.
A grounded, evidence-based look at what cold water immersion does for inflammation, mood and stress resilience, and how to start safely.
Skip the Belvedere queue. Vienna's finest rooms hold anatomical wax, sixteenth-century globes and striking clocks, with almost no one in them.
What cold water does for a lost sense of trust in the body after illness or treatment, and what to know before you start.
Reef-safe mineral sunscreens that actually survive salt, sun, and a real tropical beach; what the label doesn't tell you, and the four formulas worth packing.
A damaged skin barrier cannot be rushed. The step-by-step rebuild: cream cleanser, ceramides, mineral SPF, occlusives, and the real timeline.
A structural, evidence-based approach to rebuilding thinning eyebrows, from mapping the bone underneath to a routine that holds up all day.
Why daily washing does more harm than good, the four-step waterless routine, and how to actually make the switch.
Travel exposes every weak point in a routine. The thinnest-to-thickest order for serums and oils that holds up at altitude and after a late check-in.
A four-step routine for what altitude does to skin, from red-eye flights to ski resorts and mountain cities. Built once, packed every time.
