Due-Diligence Hiring: Show Me Your Systems, Not Your CV

The résumé is a bandwidth workaround. When an employer can query the candidate's corpus instead of reading a summary of it, hiring stops being an interview and becomes due diligence.

I've spent the last while doing something oddly revealing: working through my own CVs and cover letters, sifting the exhaust of thirty years of work. I'm a bit of a data hoarder, so a lot of it survives at reasonable fidelity. And once you stare at a résumé long enough, you start asking what the thing is actually for.

Here's where I landed. A résumé stands in for a rough question: let's see what this person has done, how effective they'd be, what's in their head. It's a proxy for the quality of someone's judgment. And it's shallow at best. That's why we bolt an interview on top — which isn't much better. And it's why, in practice, the real test is the first three to six months on the job, when they can fire you anyway.

This connects to a rule I keep coming back to: don't accelerate a horse. If you've got a bad workflow in your company, the answer is not to make it go faster. The résumé is a horse. And most of what passes for hiring innovation is just a faster, shinier horse.

The résumé is a horse. Don't accelerate a horse — replace it.

The résumé is a bandwidth artifact

Why does the whole apparatus exist? One reason: human communication is slow. How do I explain how much I know in five minutes, or half an hour in an interview? I can't. Language and the thinking behind it move at a crawl relative to the sheer volume of what an experienced person carries. So civilisation built a cascade of lossy proxies, ordered by cost:

• The CV — minutes to read, two pages, catastrophic compression.
• The interview — hours, barely better, plus theatre.
• Probation — months, and the only stage that actually measures anything. It's the real interview wearing a legal disguise.

The entire hiring stack is an admission that the channel is too narrow. Everyone knows the first two stages are noise; we run them because the third is expensive and something has to shortlist. And the data agrees the top of that funnel is weak. In the canonical meta-analysis of selection methods, the unstructured interview — the everyday "let's have a chat" — scores about .38 for predicting job performance, well below general mental ability or a structured process at roughly .51.³ One assessment firm puts it flatly: unstructured interviews are "one of the worst predictors of job performance."⁴ We are shortlisting on the noisy proxies and only measuring during the costly one.

So what replaces it? Not a better résumé. The removal of the transmission step entirely.

"Show me your personal information systems" means the employer stops receiving a summary and starts getting access. Don't tell me what you'd do — let me interrogate your corpus with a question from our actual problem domain, and watch what comes back, with receipts.

There's a precise name for that. It isn't interviewing. It's due diligence.

Due diligence, not interviewing. You don't read the founder's CV — you audit the data room.

It's how you evaluate an acquisition. A buyer doesn't hire the founder off a two-pager; they open the data room and inspect everything before committing.⁹ A data room is simply the secure repository an acquirer interrogates during diligence — "the buyer will often need to get as much information as possible about the company being purchased before closing the deal."¹⁰ Swap "company" for "candidate" and you have the next hiring model.

The deeper shift hiding in this: the résumé assumed a candidate brings only the contents of their skull, rented by the hour. Someone with a compounding personal archive is bringing assets — years of compiled judgment that arrives on day one and keeps compounding on the employer's problems. Hiring that person is closer to acqui-hiring a one-person firm than filling a seat. Which is not a new idea in miniature — in an acqui-hire, "the company's product is of at most secondary interest"; the buyer is paying for demonstrated capability, evaluated through diligence.⁷ Cal Newport calls the accumulated version of this "career capital" — rare, valuable skill you build and then cash in.⁸ The corpus is career capital made inspectable.

The interview benchmarks the wrong unit

When I said this out loud to my AI thinking partner, it came out sharper than I meant it: I'm a cyborg now — a human with a cognitive exoskeleton. I've studied a lot of AI, documented it as IP frameworks, built the code, and wiki-indexed the whole thing so a model can self-reference it while we work on the next project. So why would you interview me without all that?

There's a name for what that interview is doing: benchmarking the wrong unit of analysis. My deployed capability is the system — me, the frameworks, the wiki, the loop between them. The interview amputates the system and tests the residual, like assessing a racing driver by having him run laps on foot.

And history says this is lag, not principle. Nobody tests accountants on mental long division anymore. Once the spreadsheet became the job, the test migrated to judgment about the spreadsheet. VisiCalc — the first spreadsheet, the killer app that sold the personal computer² — didn't kill accountancy. By 2022 there were 1.4 million accountants and auditors, more than ever; they had simply "outsourced the arithmetic to the machine."¹ The manual, tedious, low-judgment layer got automated; the judgment layer got elevated. The person who insisted on testing arithmetic by hand was measuring the part the machine now owned.

Here's the honest cut, because the complaint is sharper than "you're testing me without my tools." There is a legitimate question about the naked human — it's just not the one they're asking. Strip the exoskeleton away and what remains is precisely the part that can't be delegated: taste, the North Star, knowing when to stand pat, the scar tissue, the judgment that directs the machine. So the real indictment isn't "you tested me without my tools." It's: you cut off the judgment and interviewed the memorisation — and the memorisation is the part the machine now owns. They've got the amputation exactly backwards.

The model is not the memory. Well — the candidate is not the résumé either. The interview samples the head, and the head is deliberately not where I keep it.

The prototype already ran

This isn't a prediction with no evidence. It had its first quiet run this year, on me.

I was working through a job application that asked for machine learning — which, who's doing machine learning anymore? My honest gut said I didn't have much. But the AI CV writer didn't interview me; it interviewed my archive. And it came back with: we can say twenty years, counting back from Chompster — my chess engine. I'd never filed that under "machine learning." I'd filed it under "hobby, chess, 2003" and closed the drawer. A different question reopened the drawer and found it was also evidence. The archive revalued my own history along an axis my self-narrative never used — it recompiled the past under a fresh lens and handed me the exhibit, not a verdict.

Now, the honest-friend note — because the line is worth drawing precisely. Chompster is unambiguously twenty years of AI: negamax, alpha-beta, quiescence, transposition tables, hand-tuned evaluation — the classical, symbolic branch of the field. What it mostly isn't is machine learning, strictly; nothing in it learned from data. So "twenty years of AI, including hand-built game-tree search engines competing internationally" is bulletproof. "Twenty years of machine learning" is the stretch.

Here's the thing though: you don't need the stretch, because the corpus has the real receipt sitting right there. The crypto project — sixteen generations of TensorFlow/Keras hybrid models: Conv1D towers, BiLSTM and GRU branches, multi-head attention, Huber loss, scikit-learn pipelines, Weights & Biases tracking, TensorRT inference. That's machine learning by anyone's definition, hands-on and recent. The honest sentence — AI since 2003, deep learning and model training since the TensorFlow work, applied LLM systems now — is a stronger story than the stretch, because every clause survives the follow-up question. The stretch has exactly one failure mode: a smart interviewer asking "tell me about your ML work in 2008," and the drawer being empty.

Provenance is the anti-fraud layer

"But candidates will just fake it." This is where the corpus is structurally different from a CV — and it's the part I find genuinely new.

A faked CV takes an afternoon. What can't be faked in an afternoon is thirty years of cross-referenced, timestamped, mutually-corroborating exhaust. In my own archive, a mate's email agrees with the chess-server log agrees with the tournament entry agrees with an account password from 2003. Each source independently props up the others. That web of corroboration is essentially unforgeable. Provenance is the anti-fraud layer the résumé never had.

The software world already knows this instinctively. Open-source contributions, one developer-hiring resource notes, "leave behind a permanent and public trail of how someone writes code, solves problems, and responds to feedback" — unlike résumés, "which often rely on self-reported achievements."⁵ A designer put the anti-fraud point even more precisely: "AI can fake a portfolio in an afternoon. It cannot fake twelve months of public thinking."⁶ A corpus is twelve months — or thirty years — of thinking, with the timestamps intact.

The wiki is the portfolio format for thinking

None of this is unprecedented as a pattern. Portfolio professions already hire this way. Designers show work. Academics have citation graphs. Open-source developers get hired off their public repositories without a single whiteboard question. Evidence of the actual work has always been the strongest signal available — it sits at the very top of the validity tables, above the interview.³

What never had a portfolio format was judgment — architecture, strategy, the "it depends" professions. GitHub made code legible. The wiki is the portfolio format for thinking. I've spent eight months accidentally building the first one, and the diligence trail is what makes it credible.

GitHub made code legible. The wiki is the portfolio format for thinking.

So what does a hiring due-diligence session actually look like? The move is to stop asking permission for the exoskeleton and make it the exhibit. Ten minutes in the systems-design round: "Here's a question about my own twenty-year history I can't answer from memory. Watch the system answer it, with receipts, live." You run queries from the employer's real problem domain against the corpus; you check provenance where it matters. It simultaneously proves the retrieval architecture, the judgment that built it, and the working method they'd be hiring. (This is the market between organisations; the mirror-image shift inside one — when "what does the wiki say?" replaces "ask the veteran" — is its own story.)

Two honest edges

Good predictions deserve stress-testing, so here are the two edges where this one strains.

Edge one — the IP boundary. My exhaust is mine: founder, consultant, data hoarder. But most people's best evidence is locked inside former employers' tenancies — their real work lives on someone else's servers under someone else's NDA. The future I'm describing quietly requires people to start owning their professional exhaust, which is itself a shift, and an argument I'm early to. It makes "own your career capital" a literal, technical instruction, not a motivational poster.

Edge two — the widening gap. When the interview becomes "show me your systems," the candidates without one aren't slightly disadvantaged. They're illegible. New grads are fine: they start their corpus at twenty-two, and forty years of compounding from graduation is a staggering thought. The stranded generation is the mid-career professional with twenty years of experience and no receipts — all of it real, none of it queryable. Their experience exists in the one format the new diligence can't read: their own head, behind exactly the narrow channel the résumé was invented to compress.

What you can't delegate

Let me close on the objection a smart skeptic actually raises: "You're too reliant on AI." The answer is unusually strong if you say it plainly. The system is made of my thoughts. I'm not outsourcing my thinking — I built a system where AI brings my own prior thinking back at the right moment. Every node in it is something I thought, argued, built, or shipped. The AI is the librarian; I wrote the books. Reliance looks like helplessness without the tool. I'm demonstrably better with it and still formidable without it. That's not dependence — that's what training with an exoskeleton does to the muscles.

Which tells you what the remaining interview is for. Not memory — the machine owns that now. It's for the non-delegable core: taste, the North Star, the scar tissue, the judgment that directs the machine. Every "it depends" question already probes it, and it's completely testable without a phone. The trick is dosage. Make it a nudge, not a verdict. Volunteered unprompted at length, the system is a party trick; delivered as a one-sentence answer to a question they asked, it's a credential. Let their curiosity do the pulling. (I've written elsewhere about why the small signal, offered once and then dropped, beats the big one pushed.)

And when a senior person who genuinely gets AI leans in, something bigger than a good interview beat happens. The interview inverts. You stop being a candidate describing capabilities and become a live exhibit of the thing their clients would pay for. The meeting quietly turns into a consulting demo, and the person across the table starts recalculating what you are. The confident senior hires you because of it, already imagining you in front of clients. The insecure one realises you're a peer wearing a candidate costume, and flinches. You can't control which — but the flinch is core-sample data about the team, and that role was never a good home for the machine you've built anyway.

Stop being a candidate describing capabilities. Become a live exhibit of the thing their clients would pay for.

The résumé had one job: compress a life down to something you could transmit through a thirty-minute channel. That job is ending. When the employer can query the corpus instead of reading the summary, hiring becomes what it should always have been — diligence on a system, not an interview with an amputee. Show me your systems.