A market, a category, or a set of vendors, written to an evidence standard. Every claim carries a label saying what kind of evidence stands behind it. Every source is numbered and clickable. There are no forecasts, because nobody can produce one honestly, so you get scenarios with tripwires you can watch for yourself.
A vendor's self-reported growth number, a research firm's proprietary forecast, and the author's own opinion get written in the same tone, on the same page, with the same confidence. The reader has no way to tell them apart, so they either trust all of it or none of it. Both responses are wrong, and both are expensive.
This report labels every claim. You can see at a glance which sentences rest on a primary source, which repeat a vendor's marketing, which are somebody's estimate with an undisclosed method, and which are my judgment. Where the evidence is thin, the report says so instead of writing around it.
Six labels: verified fact, vendor claim, third-party estimate, my assessment, hypothesis, and scenario. If a sentence can't be labeled it's either filler or a claim being smuggled in, so it gets cut.
Numbered, clickable, and tagged by what they are: vendor primary, peer-reviewed, analyst firm, independent press, self-selected practitioner survey, or second-hand and unverified. A source list where the weak entries are visibly weak is more useful than one where everything looks equally solid.
Every premise the analysis rests on is stated up front, each one naming which conclusions fail if it turns out to be wrong. You can disagree with a premise and know exactly what that costs you.
Point forecasts in a fast-moving category are decoration. You get scenarios with stated probabilities and, more importantly, observable tripwires: specific events you can watch for without private information, each one telling you which scenario is playing out.
The questions that need primary research get named rather than papered over. This is usually the most useful page in the document.
Which claims decay in six months, which in twelve, and which are architectural. If you pick the report up a year later you know how to discount it.
Every figure is tagged like body text, with a caption that names the weakness in the underlying data. A chart built from vendor self-reporting says so on its face.
Every one was built from public sources alone, with no client brief and no interviews. That's the point. Pick one to read it in full.
Work & careers
Nobody Is Reading Your Resume
Why 254 applications land on one opening, the screening system ranks instead of rejecting, and referral is the only lever left.
25 pages · PDF
Vendor & platform
The fabricated opener
An AI sales pipeline read a post I wrote about another company, handed the award to the author, and generated a membership history that never existed. Traced from the message back to the source.
21 pages · PDF
Work & careers
You Just Said Something Terrible
The cross-cultural blunder canon, audited: a third of the famous stories are false, and the expensive failure is the polite no.
27 pages · PDF
Vendor & platform
The Automation Platform Overhaul
Build versus buy for automation, decided per layer. The license is rarely the dominant cost, and break-even sits at 140 to 200 unattended bots.
32 pages · PDF
Vendor & platform
The Canvas After Code Got Cheap
Coding agents took the authoring step from visual workflow automation. n8n's runtime survives; its per-execution meter does not.
25 pages · PDF
Vendor & platform
The Rent and the Reach
Is Power Platform still worth its price now that AI-assisted development reaches what it cannot build?
17 pages · PDF
Technology & markets
The Bottleneck Is Below the Model
Something real happened in AI for science this year, and all of it is upstream of the real question: whether these systems can reach trustworthy inputs and whether anyone can check the output.
21 pages · PDF
Public statistics
Seventy Thousand Deaths and One Contested Inference
The viral graphic claiming pro-Trump counties suffered 70,000 more COVID deaths packs four claims into one image, sorted apart by the evidence each requires.
19 pages · PDF
Work & careers
The Day Your Title Stops Working
Whether a knowledge worker should build owned professional distribution before AI reshapes the role, and where: LinkedIn, Substack, or self-hosted.
21 pages · PDF
Professional services
The Independent Coaching Practice
Executive and business coaching, where published market sizing ranges across two orders of magnitude for the same market, taken apart from public figures alone.
20 pages · PDF
Public statistics
The Gen X Confidence Number
The chart that put Gen X at the lowest consumer confidence of any American generation, taken apart claim by claim against its own source.
18 pages · PDF
Company & brand
Microslop
Microsoft assessed from the position of the people who pay for it: the household on a subscription, the laptop that stopped getting updates, the small firm reading its renewal quote.
25 pages · PDF
Vendor & platform
Copilot Is Not One Product
Six commercial models behind one Microsoft brand: what each Copilot surface bills, where the boundaries sit, and how to budget a category that reprices twice a year.
22 pages · PDF
No reports match that.
Some of these are bought as the whole engagement. A board needs a defensible read on a category before it commits capital, or a founder needs to know whether the market they've described to investors actually exists at the size they've claimed. The report is the deliverable and the work ends there.
The rest become discovery. A fixed-scope, fixed-price assessment is a cheaper and more honest way to start than a discovery phase billed by the hour, because you own a finished artifact either way. If the work continues into architecture, advisory, or a build, the assessment fee credits against the first invoice of that engagement.
It's one comprehensive report, not a menu of tiers. Every assessment runs the full method above and adapts to the subject, whether that's a market, a category, or a named set of vendors. It's quoted as a fixed fee against a signed scope, never billed by the hour.
The published reports above are public examples of exactly this work. What you commission is the same method and standard, delivered privately to you and scoped to your specific question.
The full report on your chosen subject: stated assumptions, structure and sizing with reasons to discount it, facts separated from vendor claims, scenarios with observable tripwires, a section on what it cannot answer, and a stated half-life. Delivered in ten business days from a signed scope.
Yours to use however you want, and the fee credits in full if the work becomes a larger engagement.
The assessment, plus a two-hour working session on what the findings mean for your specific decision, and a short written recommendation. The fee credits in full against a follow-on engagement in another pillar. Priced with your scope.
Primary interviews are not included by default and are priced separately. Every report states plainly whether interviews were conducted.
This is not Gartner. There is no research staff behind it, no proprietary panel, no syndicated dataset. It's one practitioner with thirty years in enterprise technology, reading public sources carefully and telling you exactly how much weight each one holds.
It will not tell you a market is bigger than the evidence supports, and if the honest answer is that the published numbers can't be reconciled, that's what the report says. Anyone who needs a document confirming a decision that's already been made should buy it somewhere else.
Turnaround is ten business days from a signed scope. Rush is possible and costs more, and I'll tell you when rushing would make the work worse.
Tell me the market and the decision behind it. If a report is the wrong tool for what you're actually trying to work out, I'll say so before either of us commits.
Start the Conversation