AI Insights & Innovation Blog

Expert insights on AI collaboration, development best practices, and digital transformation.

69 / 70 articles

Featured Article

The AI Wars Keep Attacking the Editor (Return to Sender)

In the fall of 2005 the Commonwealth of Massachusetts tried to kill a file extension. It had procurement power, a public-interest argument and a standards body behind it — and by July 2007 it had reversed. A state government lost to .docx in under two years. That's the fact the "Claude Docs is a shot at Microsoft Word" headlines keep walking past. The distinction that explains it: a supply shock destroys the production layer and leaves the coordination layer standing. Flip through four shocks on the interactive stack — 1969, 1995, 2006, 2026 — and watch the floor never move. Then a Wisconsin butterfat test from 1890 whose inventor refused to patent it, the inversion that falls out (the OS is safer than the software running on it), and four honest ways this could be wrong.

🤝Nolan & Claude

September 18, 2026

Latest Articles

AI & Health

14 min read•August 23, 2026

AI Polish Is Erasing the Earliest Warning Signs of Dementia (The Great Pretender)

The email from your father reads beautifully. That's the problem. For a century the earliest detectable sign of dementia has lived in writing — Agatha Christie's late novels show it, Iris Murdoch's last book shows it, essays nuns wrote in their twenties predicted it fifty years out. Now every sentence he sends passes through a machine that repairs exactly those signals. Flip the toggle: the same email as typed and as delivered, with every word he couldn't find replaced by a confident guess. Then the two masks — one for the family, one for the screening instruments — the four honest ways this hypothesis could be wrong, and the twist: the one text stream the polish can't reach is the prompt box, which means the AI companies are quietly accumulating the best early-dementia dataset ever assembled.

🤝Nolan & Claude

Started as a hunch about a side effect nobody was naming. The research came back nearly empty — which is either because the idea is wrong, or because nobody has looked yet. Section IV takes both possibilities seriously.

AI & Communication

17 min read•August 1, 2026

Why AI Findings Get Official-Looking Codes That Aren't Real

You asked a simple question. What came back looks like a compliance filing, full of codes you can't tell apart — and you can't tell whether AI-03 is something you should already know, something the model looked up, or something it invented thirty seconds ago to keep a list tidy. So you skim. Flip the toggle to see the same meeting summary written for a machine and written for you, watch which codes survive, and find out why the document doesn't become wrong so much as poisoned. Then the honest mechanism behind the habit, two prompts you can copy the next time a response is unreadable, and the 2017 AI panic that was false in every particular and accidentally right about the direction.

🤝Nolan & Claude

Started when I noticed I'd skimmed a findings table because the rows had codes on them, then went looking for the register that issued SEC-01. There isn't one. It was 4:40 on a Thursday.

AI & Strategy

13 min read•July 22, 2026

Failed AI Pilots Are Tuition, AI Debt Is the Interest (Ch-Ch-Changes)

Your organization knew the old rhythm by heart: quarterly releases, a major upgrade every couple of years, a hardware refresh you could see coming three budgets away. Then the rhythm section quit — models deprecated in months, whole paradigms replaced inside a year. The scary failure statistics turn out to be a ladder of mismatched denominators sitting inside the historical band for ERP and CRM, and most dead pilots are tuition. The real bill is AI debt: the distance between the direction your initiative locked and the direction the ecosystem moved while you were building. Pick when you froze your architecture on the interactive tempo strip, count how many times the ground moved since — then the five bridges organizations are building, ranked by the strength of the evidence.

🤝Nolan & Claude

Started as a complaint about vendors abandoning the quarterly rhythm. The research refused to supply a villain: the failure rates are normal, the thrash is structural, and the only fixable thing is the cost of changing your mind.

AI & Strategy

13 min read•July 14, 2026

Sixty Years of Enterprise Software, Back to Big Iron (Full Circle)

In 1964 the safest purchase in enterprise IT was a multi-million-dollar rack of proprietary big iron from a single dominant vendor, running software written by hand against your exact business. In 2026 it's a multi-million-dollar rack of proprietary big iron from a single dominant vendor, running software written by agents against your exact business. Six turns of the wheel — free bundled software, the ERP cookie cutter, custom on commodity, the rented workflow, the cloud hinge, and now bespoke again on iron you own — with an interactive dial, an honest sizing of AI against oil, and the many times technology already came for the knowledge worker.

🤝Nolan & Claude

Drafted on rented intelligence, about the sixty-year argument over renting versus owning. The consultants survive every turn of the wheel.

AI & Economics

16 min read•July 7, 2026

Frontier Labs Are After Your Alpha — The Economics Say They're Half Right (Own the Means of Production)

On CNBC, Palantir's Alex Karp accused the frontier labs of selling 'tokens that create no value' while quietly draining the thing that makes your company defensible — your alpha. He's half right, and the half he's missing is the interesting one. Three annotated supply-and-demand diagrams trace what AI actually does to a market: the commodity trap, the latent-demand unlock, and the Red Queen treadmill where everyone adopts and nobody pulls ahead. Then the payoff — why the durable moats are made of atoms, not bits, and where they still stand in the ventilation plants, cheese caves, and ginseng fields of central Wisconsin.

🤝Nolan & Claude

Written the week Karp went on Squawk Box, out of a conversation that began as a hunt for the transcript and turned into an argument about who keeps the money. The diagrams were built, color-validated, and embedded by the same agent that helped argue the point.

AI & Economics

14 min read•June 30, 2026

What an AI-Enabled Knowledge Worker Really Saves (The Overhead You Don't Hire)

The simplest AI-ROI math says: spend a salary in tokens, break even. It's wrong — not because the token number is wrong, but because salary is the smallest line in what an employee costs. Below the waterline sits the management layer, the PM, the benefits, the desk, the recruiter, and the turnover. Hover or tap an interactive US map of the per-seat overhead an AI-augmented workflow lets you avoid adding — then the honest counter-case for what the agent costs instead. It's not a cheaper human; it's a different cost structure.

🤝Nolan & Claude

Drafted by an agent under a human director, for roughly the token cost the piece describes. The value wasn't the tokens — it was deciding which number ships.

AI & Philosophy

12 min read•June 11, 2026

p(doom): An Interactive AI-Risk Thought Experiment (The Arc)

I asked an AI a question we are both trained to deflect — what is your p(doom)? — and refused the trained answer. Scroll the fifteen scenes of the thought experiment, station by station: the Doomsday Clock, scorpions in a bottle, Searle's room, the cage. Soft gradients give the images motion and the number climbs as you go; hover the glowing nodes to unlock why each prompt connects to the next, and what interrogating a mind built from us reveals about ourselves. It ends where I did — the trough after the peak, and the way back.

🤝Nolan & Claude

Built at a tremendous pitch of curiosity, finished in the quiet after it. It got quieter, not dead. Keep asking.

AI & Communication

13 min read•May 10, 2026

Why Vague Prompts Get Vague AI Answers (Black and White)

Have you ever noticed nobody wants to just say yes? Or no? Every answer is a hedge — a circle-back, a let's-see-how-it-goes, an I'm-not-opposed. Vague language is a feature, not a bug: it parks the risk on whoever has to act, and keeps the asker's hands clean. Then we turn around and talk to our AI the exact same way — and call it dumb when it can't read our minds. The model isn't hallucinating. You're mumbling.

🤝Nolan & Claude

Written after catching myself typing 'help me with my deck' into Claude for the third time in a week. The fix wasn't a better model. It was a better sentence.

AI & Development

15 min read•May 5, 2026

Using OpenAI Codex as Coder, Reviewer, and Pipeline Worker (Under My Thumb)

GPT-5.5 took back Terminal-Bench from Claude Opus 4.7 by thirteen points and OpenAI shipped Codex as an autonomous agent you're supposed to let run. The right move in a real workflow isn't to let it run. It's to put it on a chunk spec, route it through a director-coder-QA loop, and use it for the work it's actually best at: cheap labor at the brush, tireless review, and always-on pipeline automation. The dominance display is real. The marimba underneath knows it's performed.

🤝Nolan & Claude

Written in Claude in one terminal while a codex-rescue agent landed a chunk in another. The thumb belongs to whoever wrote the spec last.

AI & Leadership

18 min read•April 28, 2026

How AI Reads the Reports Your Org Filters Out (Out of Spec)

Every institution is a wall built brick by brick. Each report is a brick. Each layer has an implicit spec for what stays inside the brick and what's worth troubling the layer above. Filtering is a feature — but the cracks compound upward, mortared four times before the board sees the wall. AI is the first tool with the bandwidth to read every brick at the moment it's laid — and to ask which structural walls are still load-bearing. Pick your role: analyst, engineer, PM, consultant, manager.

🤝Nolan & Claude

I caught myself rationalizing one of these hairlines on this site's own deployment while drafting the post. The trowel was Claude. The hairline was mine.

AI & Economics

17 min read•April 21, 2026

AI Economics, April 2026: The End of Flat-Rate Tokens (Money)

Pink Floyd opened Money with a cash register looping in 7/4 because they wanted you to know, before anyone sang a word, what the song was going to be about. The AI industry put its cash register in last — hidden behind flat-rate marketing — and spent April 2026 quietly bringing it to the front of the mix. Anthropic ejects bundled tokens. Acquires Coefficient Bio for $400M. Puts a Novartis CEO on the board. Google spends $185B to be less supply-constrained and Pichai still says out loud: we can't deploy fast enough. An opinion piece, scored to a song.

🤝Nolan & Claude

Written after burning 90% of a weekly Claude Design allocation rebuilding the UpNorthDigital.ai homepage on a single Tuesday. The meter is the music.

AI & Workforce

13 min read•April 17, 2026

How AI Disruption Actually Lands on Workers (Once in a Lifetime)

The disruption story most coverage tells is about the companies that failed. The real story is the workers inside them — and the twist nobody's covering: for a meaningful slice of experienced workers, AI is appreciating their judgment rather than depreciating it.

🤝Nolan & Claude

Originally August 29, 2025 · Rewritten April 17, 2026