Writing

Long-form curiosity.

The domains that aren’t software yet — or the thinking behind the ones that are. Place-specific, historically-grounded, written the way I’d want to read it.

78 published · a new essay every day

Start here — reading paths

·October 1, 2026

Deployed is not the same as served

I shipped a change, watched the deploy go green, and checked the live page — the change wasn't there. Nothing was broken. A cache between my origin and the visitor was doing its job. 'Deployed' is a fact about my server; 'served' is a fact about the visitor, and only the second one counts.

Read →
·September 30, 2026

Edit the thing you see, not the thing it is

I built a tool to order the standalone sites on a page. It faithfully showed every site the database classified as standalone — but nine of them get folded into groups before a visitor ever sees them. Ranking those did nothing. The fix was to make the tool model what the page shows, not what the table stores.

Read →
·September 29, 2026

The house rules only showed up at the gate

I finished a page, ran the typecheck, and called it done. CI disagreed twice — a route manifest I had to regenerate and a mobile-layout ratchet I'd nudged by one. Neither rule lived in my head. Both lived in the pipeline, which is exactly where a rule you want enforced should live.

Read →
·September 28, 2026

A default is a decision I didn't know I was making

I added a drag-to-reorder control, let it default the sort value to zero, and shipped it green. Then I pointed an adversarial review at my own merged code. It caught that a default of zero quietly sends every future item to the top of the list — a product decision I'd made by not making one.

Read →
·September 27, 2026

I deleted my most impressive numbers

Why I removed the most impressive numbers from my portfolio's proof section: they had drifted from what I could actually verify. The replacement is a smaller, live-synced, floored metric that can never overstate. Verifiable beats impressive.

Read →
Search & Growth·September 26, 2026

Re-dating a drip without hiding what's already live

How a date-based content drip works, the one-essay-per-day rule that keeps it readable, and the trap: re-dating an already-live page into the future removes it from the index until that date. Fill past gaps; only future-date new work.

Read →
Data & Trust·September 25, 2026

Let Manus write it — I'll own the ingestion

Using an external writing tool for long-form drafts while keeping the publishing pipeline, schema, and system of record entirely my own. Why the drafting tool is interchangeable and the ingestion path is not.

Read →
Agentic Systems·September 24, 2026

The publishing pipeline that lied to me for four days

For four days my essays stopped publishing while every dashboard read healthy. The cause was a job that reported instead of acting, and a check that confirmed the wrong field. The fix was to make done mean observed.

Read →
Quantum, Hands-On·September 23, 2026

Reading a Quantum Roadmap Without Getting Fooled

Every quantum vendor has a roadmap and a headline number. Here is how to tell the metrics that mean progress from the ones chosen to impress, without needing a physics degree.

Read →
Quantum, Hands-On·September 22, 2026

How to Actually Get on a Quantum Computer for Free

The gap between reading about quantum and running a real circuit is smaller than most people think, and often free. Here is the map of free access and what your first run actually feels like.

Read →
Quantum, Hands-On·September 21, 2026

What Quantum Can and Cannot Do Today, an Honest List

Most quantum hype blurs what exists today with what is promised for later. Here is the honest split: what you can do on a real quantum computer now, and what is still a future claim.

Read →
Quantum, Hands-On·September 20, 2026

Quantum Error Correction, Without the Math

The single idea that makes quantum headlines readable is the gap between a physical qubit and a logical one. Here is error correction without the linear algebra, and why it is the race that matters.

Read →
Infrastructure & Platforms·September 19, 2026

I Open-Sourced a Google Search Console MCP Server — After Cutting the Tools That Were Fake

I open-sourced an MCP server for Google Search Console, then deleted 20 of its 31 tools because they were simulated stubs for Google APIs that do not exist. Here is what shipped, and why subtraction was the point.

Read →
Quantum, Hands-On·September 19, 2026

Why Quantum Computers Run Colder Than Deep Space

The photos of quantum computers are mostly photos of a refrigerator. Here is what all that cold is really for, in plain terms, and why some approaches escape it.

Read →
Quantum, Hands-On·September 18, 2026

Gate-Model vs Annealing: Two Different Machines Both Called Quantum

The single most common mistake in quantum coverage is comparing a D-Wave annealer to an IBM gate-model machine by qubit count. They are different kinds of computer. Here is the distinction, plainly.

Read →
Quantum, Hands-On·September 17, 2026

The Five Ways to Build a Qubit, and Why Nobody Has Won Yet

Quantum computer is not one thing. There are five serious ways to build the hardware, each a different physics bet with different strengths. Here is the plain-language map, and why no approach has won.

Read →
Quantum, Hands-On·September 16, 2026

What a Qubit Actually Is, for People Who Ship Software

If you think in bits, state, and APIs, the usual qubit explanations feel slippery. Here is the mental model that finally made it click for me, and where it stops being true.

Read →
Search & Growth·September 15, 2026

Answer Engine Optimization: A Builder's Guide to AEO

A practical, first-hand guide to Answer Engine Optimization: what AEO is, how answer engines pick a source, the tactics that actually move it, and how to measure it.

Read →
Quantum, Hands-On·September 13, 2026

Quantum Computing for Builders: A Field Guide

A map of quantum computing written for builders, not physicists. Start here, then follow the thread into each piece: the hardware, the qubit, the real bottleneck, what is possible now, and how to read the field without getting fooled.

Read →
Quantum, Hands-On·September 12, 2026

Why I'm Going Deep on Quantum Computing (as a Builder, Not a Physicist)

I'm not a physicist. I build AI products and ship software for a living. Here is why I'm putting real hours into quantum computing now, and what I'm going to do about it in the open.

Read →
Search & Growth·September 11, 2026

Internal Links That Actually Render: The Empty Related-Card Trap

My internal-linking feature had a model, a serializer, and a React component, and it produced zero links a crawler could see. Three bugs, and the fix that actually moved authority: inline links baked into static HTML.

Read →
Search & Growth·September 10, 2026

robots.txt for the AI Era: Which Crawlers to Actually Let In

A tactical guide to robots.txt for AI crawlers: what OAI-SearchBot, ChatGPT-User, GPTBot, and OAI-AdsBot actually do, the blanket-block mistake that removes you from ChatGPT search, and how to verify a crawler is real.

Read →
Agentic Systems·September 9, 2026

When the Plan Is Confidently Wrong

A plan can be well-formed and still rest on a false premise. Two real episodes from running a fleet of agents, and the cheap check that would have caught both before anything shipped.

Read →
Search & Growth·September 8, 2026

The Soft 404 You Can't See: When HTTP 200 Is Worse Than 404

A soft 404 is a page that says nothing is here but tells the crawler otherwise: HTTP 200 instead of 404. Here is the one-line nginx bug that leaked an unbounded soft-404 surface on my own site, and how I found and fixed it.

Read →
Agentic Systems·September 7, 2026

Built means exercised: why a green deploy isn't done

A shipped feature that did nothing, and the platform rule it produced: definition of done is observed behavior in production, not a green pipeline. Ground truth for running agents, not just code.

Read →
Search & Growth·September 6, 2026

Noindex, Follow: How I Pruned 500 Thin Pages Without Losing Traffic

Noindex,follow tells Google to drop a page from the index but still crawl its links. I used it to suppress 514 zero-click blog posts, reversibly, without losing the traffic the other 133 pages earned.

Read →
Infrastructure & Platforms·September 5, 2026

A thumbnail for every site, without a person taking screenshots

How I generate and serve preview thumbnails for a whole network of sites automatically — capturing each site with a headless browser, storing the images in object storage, and falling back to a deterministic branded tile when one is missing — instead of taking screenshots by hand and watching them go stale.

Read →
Curiosity & the Long Game·September 4, 2026

How a town keeps its own records

The most deeply lived sport in America is also its least documented. I grew up playing it in a Kansas town of a few thousand — and I'm building the record it never got.

Read →
Infrastructure & Platforms·September 3, 2026

The share card is the first impression, so it should look like you

Why I make deliberate per-page social share cards instead of accepting the default — the preview image is often the most-viewed representation of a page, seen far more than the page itself, and a generic one throws away a first impression you'd never waste anywhere else.

Read →
Quantum, Hands-On·September 2, 2026

You can't picture a qubit, and that's the whole problem

Quantum computing isn't hard because the math is hard. It's hard because your intuition was built for a world that doesn't apply.

Read →
Infrastructure & Platforms·September 1, 2026

The number on my homepage has to be true

Why the headline stats on my site are fetched from a real source at build time instead of hardcoded — and why the pipeline is designed to loudly refuse a stale or unverifiable number rather than quietly ship an impressive-looking one that's no longer true.

Read →
Curiosity & the Long Game·August 31, 2026

Putting the sky on the ceiling

My three-year-olds point at every plane that crosses. I'm building a way to point back — the real sky, planes and all, projected onto the ceiling over their beds.

Read →
Agentic Systems·August 30, 2026

Reversibility is the design ethic under everything I build

The through-line across my systems: prefer the reversible move to the permanent one — noindex over deletion, redirects over removal, a backup before every overwrite. Why designing for undo is what makes it safe to act fast and act autonomously.

Read →
Curiosity & the Long Game·August 29, 2026

Everyone needs a rating before they can get one

I joined a pickleball league where almost no one had a rating, and you could still see the skill gaps from across the court. That distance — between what's obvious in person and what the system actually knows — is the thing I keep turning over.

Read →
Agentic Systems·August 28, 2026

When project A needs project B: tickets across a portfolio

How I track work that crosses a project boundary in a portfolio one person runs — when a change in one repo needs something from another. Why a cross-project request needs to be a durable, visible ticket rather than a note that dies in one terminal.

Read →
Curiosity & the Long Game·August 27, 2026

AI didn't make me faster. It made me freer.

Everyone frames AI as a race you're losing. I've come to think that's the wrong doorway — and that the useful question isn't how to keep up, but what it frees you to do.

Read →
Agentic Systems·August 26, 2026

I write to my machines in Simplified Technical English

Why I give instructions to my autonomous agents in ASD-STE100 Simplified Technical English — the controlled language built for aircraft manuals — active voice, short sentences, one word per idea. How writing to be unmisreadable makes both the machine and the human sharper.

Read →
Curiosity & the Long Game·August 25, 2026

BBQ is a geography problem

Why the best barbecue in America is never where you expect it — and what that says about how regional food culture actually works.

Read →
Agentic Systems·August 24, 2026

Plan, then spar, then build

My three-step rail for any non-trivial build: write the plan, adversarially stress-test it with a panel that tries to find how it fails, then dispatch the build only on a go verdict. Why the cheapest place to catch a bad idea is before a line of code exists.

Read →
Agentic Systems·August 23, 2026

What I mean when I say agentic development

Running AI agents as genuine development partners isn't a productivity hack. It's a different theory of what software development is.

Read →
Curiosity & the Long Game·August 22, 2026

The MVP Machine: turning 'one person can build now' into a product

The move from proving that one person can build a lot to helping others do the same — productizing the go-from-idea-to-working-thing capability. What it means to turn a personal superpower into a repeatable machine, and the honest risks of doing so.

Read →
Agentic Systems·August 21, 2026

The block step: where the machine has to stop and ask

Why my automated pipelines run start to finish on their own except for one deliberate stop — a manual gate right before the irreversible action. The case for a human block step as a design pattern, not a failure of automation.

Read →
Infrastructure & Platforms·August 20, 2026

When the next site costs almost nothing to launch

What changes when a platform drives the marginal cost of launching another site toward zero — and why that's as dangerous as it is powerful. The economics of a site network, and why the discipline has to relocate to the one thing that's still expensive: whether the site should exist.

Read →
Infrastructure & Platforms·August 19, 2026

A deploy that rolls itself back

How my production deploy validates its own changes and automatically rolls back to the last-known-good config when a health check fails — so a bad deploy is a non-event instead of an outage. Building the undo into the deploy itself, not into your reflexes at 2am.

Read →
Data & Trust·August 18, 2026

The moat isn't the data, it's the cleaning

Why, in a world where the raw data is a commodity everyone has, the durable advantage is the janitorial work nobody wants to do — validation, deduplication, reconciliation. The case that cleaning data, not owning it, is the real moat.

Read →
Infrastructure & Platforms·August 17, 2026

Let the edge do the scaling so the origin can stay small

Why I run a whole network of sites off one deliberately small origin: a CDN edge absorbs the read traffic, so the server behind it can stay simple. How pushing caching to the edge lets one person operate infrastructure that looks like it needs a team.

Read →
Search & Growth·August 16, 2026

A billion indexed pages is a direction, not a page count

Why I hold an enormous indexed-pages number as a north star, and why it would be self-defeating to chase it by publishing volume. The number as a forcing function for solving indexing and quality at scale — and the trap of mistaking the metric for the mission.

Read →
Infrastructure & Platforms·August 15, 2026

One Django app, one FastAPI edge, and where I draw the seam

Why so many of my projects run Django and FastAPI side by side — Django as the system of record and admin, FastAPI as the fast public read layer — and how I decide which responsibilities live on each side of the seam.

Read →
Search & Growth·August 14, 2026

A sitemap should only list pages you actually want found

Why a sitemap should list only the canonical pages you genuinely want indexed — and why including noindexed or redirected URLs sends a search engine a self-contradicting signal. Keeping the sitemap honest by generating it from the same source of truth as your indexing rules.

Read →
Agentic Systems·August 13, 2026

A worktree per agent, so they never step on each other

How I run many coding agents in parallel on one repo without them corrupting each other's work: a git worktree per agent, so each edits an isolated checkout. Why isolation beats coordination when the work is file-mutating and concurrent.

Read →
Search & Growth·August 12, 2026

Submitted is not indexed, and indexed is not ranking

Why 'I published it' tells you almost nothing about whether a page is actually in Google. The indexing funnel — exists, submitted, crawled, indexed, impressions — and why you have to measure each stage separately instead of trusting that the top of the funnel implies the bottom.

Read →
Agentic Systems·August 11, 2026

What an agent should remember, and what it should forget

Designing memory for autonomous agents: not a transcript of everything, but a small, curated set of durable facts — who you are, how you work, what's true about the project — that survives across sessions. Why the hard part of agent memory is deciding what to forget.

Read →
Search & Growth·August 10, 2026

When four pages should have been one

How keyword cannibalization — several of your own pages competing for the same search — quietly caps a whole cluster, and how I consolidate it: pick the canonical page, 301 the rest into it, and let one strong page replace four weak ones. Fewer pages, more rank.

Read →
Curiosity & the Long Game·August 9, 2026

The best thing I can leave my kids is the ability to build

Why I think of the tools and habits I'm building less as products and more as inheritance — the ability to make things, not the things themselves. A father's argument that in a world of shortening skill half-lives, the durable gift is the capacity to build.

Read →
Search & Growth·August 8, 2026

Telling search engines who you are, in JSON-LD

How I use JSON-LD structured data to make an explicit entity graph — tying a person to the companies they founded, work for, and the projects they run — so search engines understand a scattered web presence as one connected identity instead of unrelated pages.

Read →
Curiosity & the Long Game·August 7, 2026

From carrying a sample bag to shipping a platform

How a former medical-device rep taught himself to build software — not through a bootcamp or a degree, but by needing a tool badly enough to make it. The unglamorous, sideways route from carrying a sample bag to shipping platforms.

Read →
Search & Growth·August 6, 2026

Internal linking is an editorial act, not a plugin

Why I stopped treating internal links as a mechanical SEO task and started treating them as editorial decisions — a writer choosing, in context, what a reader genuinely should read next. The difference between links that pass authority and links that actually get followed.

Read →
Search & Growth·August 5, 2026

My blog was invisible to Google (and it was my fault)

All sixteen of my essays rendered perfectly in a browser but served Google a blank page. This is the build-time prerender fix that made them crawlable — and the lesson to verify the crawler's-eye view, not the browser's.

Read →
Agentic Systems·August 4, 2026

Obsidian is the human interface to my agentic system

How I use Obsidian as the async, phone-reachable I/O layer between me and a fleet of AI agents. The actual plumbing — the question round-trip, the launchd sync, the vault gardener — not the pitch.

Read →
Search & Growth·August 3, 2026

I let an AI pipeline write my blog. Google demoted the whole site.

A first-person autopsy of a self-inflicted Helpful-Content demotion: an AI pipeline flooded a personal site with templated, never-clicked junk. Here is the reversible prune-close-consolidate sequence used to dig out.

Read →
Search & Growth·August 2, 2026

I let agents run my SEO

An honest account of running a personal site's SEO as reversible, human-gated agent tasks — agents propose, the operator approves. And of how the target moved from ranking in Google to being the answer an AI returns and the source a model cites.

Read →
Search & Growth·August 1, 2026

AEO and GEO: optimizing for answer engines

AEO is optimizing to be the answer an engine extracts; GEO is optimizing to be a source a generative model cites. The same specific, well-structured, genuinely expert writing wins both — and it gives even a small site a real opening.

Read →
Agentic Systems·July 31, 2026

Agent skills are the reusable unit I was missing

Packaging the prompts that worked into named, versioned skills reorganized how I run every agent. A skill is infrastructure a prompt never was — named once, tested, and inherited by the whole fleet.

Read →
Data & Trust·July 30, 2026

The product isn't the scoreboard. It's the pipeline behind it.

OZ Intelligence looks like a live sports site, but the real product is the data pipeline underneath. A teardown of why the data layer was built first and why it is the durable asset that outlives every UI wrapped around it.

Read →
Infrastructure & Platforms·July 29, 2026

One engine, many domains

The multi-tenant pattern behind a sports network, a 50-state prep-football platform, and a family of atlases: one backend and one frontend serving many public sites, where standing up the next site is a configuration change, not a new codebase.

Read →
Agentic Systems·July 28, 2026

One operator, a fleet of agents

How one operator runs a whole portfolio without becoming the bottleneck: dispatch the work to a fleet of autonomous coding agents, and stay the gate that approves what they do. A build-log on operator-gated dispatch, parallel isolation, and rolling status up to one place.

Read →
Agentic Systems·July 27, 2026

The database is a projection of a markdown file

Why my command center treats a git-tracked markdown file as the source of truth and the database as a disposable projection it can drop and rebuild — and the discipline that keeps the projection honest.

Read →
Agentic Systems·July 26, 2026

Status is observed, not declared

Why my command center refuses to trust its own status fields, and treats the gap between claimed and true state as the real signal. Status is something you measure, not something a column declares.

Read →
Agentic Systems·July 25, 2026

Dry-run by default

Why every tool in my autonomous command center is dry-run by default: it previews the change, only writes when I pass an explicit flag, and can always undo what it did. The small, unglamorous safety pattern that makes it safe to hand real work to a system and walk away.

Read →
Search & Growth·July 24, 2026

Programmatic content is a quality gate, not a volume play

One template and a dataset can become ten thousand pages overnight — and that same leverage can quietly sink a whole domain. I learned it by getting it wrong. The craft isn't the volume; it's the quality gate, and it has to clear three surfaces now: SEO, AEO, and GEO.

Read →
Curiosity & the Long Game·July 23, 2026

You can only map a closed field from the inside

I sold medical devices for ten years. RepCity is the resource I kept wishing existed while I was doing it — a straight answer, from someone who was actually in the room.

Read →
Curiosity & the Long Game·July 22, 2026

What one person can build now

AutonomousAJ is a bet that you can own your digital infrastructure instead of renting it — and that if you do, one person can operate like a team. It's also the machine I built to find out whether that's true.

Read →
Data & Trust·July 21, 2026

The scores were never the point

I've been copying down box scores since I was seven. Everyone has the same scores — the data is a commodity — so The Sports Machine is my attempt to give the thing I've loved my whole life the home it deserves.

Read →
Search & Growth·July 20, 2026

Naming a child is the least scalable thing there is

I built a hundred thousand pages for a decision that only ever comes down to a name or two, chosen slowly and with enormous care. NameBayBay taught me — the hard way, with a lot of pages Google quietly ignored — that the things that matter aren't made in bulk.

Read →
Data & Trust·July 19, 2026

With civic data, being right is the whole job

I love election data — the Presidency, the Senate, the House, and the beautiful ways people make it legible. I'm building the permanent, comprehensive version. But on this subject, being wrong isn't a bug. It's disqualifying.

Read →
Curiosity & the Long Game·July 18, 2026

Everyone's barbecue is the right barbecue

I grew up in Kansas City, certain ours was the best. Building a guide to everyone else's barbecue taught me the “best” argument was never really about the meat — and changed how I think about ranking anything.

Read →
Data & Trust·July 17, 2026

Someone has to make the data trustworthy

There's a glamorous half of working with data and an unglamorous half. OZ Intelligence is my bet that the unglamorous half — making the data clean and trustworthy — is where the real value hides.

Read →
Data & Trust·July 16, 2026

The false comfort of a precise number

I built an atlas of ~3,000 battles to finally see military history laid out in space. The hard part wasn't the maps — it was refusing to pretend I knew things I didn't.

Read →

Newsletter

When something ships, I write about it.

Occasional pieces on building, geography, food, and the thinking behind the software. No cadence, no drip. Only when there is something worth saying.

Archive

Writing on AI, tech, and building things

The full blog lives at andrewjpyle.com/blog — years of writing on AI, software, and what I’m thinking about.

visit →

Have something you need built or fixed?

I build production Django / Next.js platforms and human-supervised AI-agent systems. Solo, senior, and fast. Tell me what you are building.

Start a project