Sitemap

About Lightning Jar

Ask Eljay

Blog

Your Next Breach Won't Come Through the Contact Form: Website Security Moved Upstream

The First Agent Your Site Should Serve Is Your Own: Chatbots as AEO Proof

The Platform Gave Us the Chat Widget: Popover and Anchor Positioning in Production

We Built Our Chat Agent the Way Our Research Told Us To

Three Files Nobody Reads

The Index Goes in the Context

One Sentence Beats Every File

Nobody Reads llms.txt (Yet)

Introducing AEO Bench

The Cheapest Model Passed Every Gate

Cutting a Google Maps Bill 87% by Un-Billing the Bots | Lightning Jar Engineering

Website QA Automation: Type Checks, Linting, a11y, Performance, Agent Readiness | Lightning Jar

Websites Are Operated Systems: Audits, Monitoring, and Automated Checks | Lightning Jar

woof-editor 0.2: Round-Trip Markdown WYSIWYG for Svelte 5 | Lightning Jar Engineering

What It Means to Make Your Website Agent-Ready

What Premium Web Design Looks Like in the Age of AI Images

Why We Are Switching Away from Google Fonts

Our GTFS Build Pipeline: CI as ETL

Bus Stop QR Codes: Print-Ready Proofs from Web Infrastructure

Building a Modern Transit Website with Live Bus Tracking

Cold Email Is Spam

Give the Model a Tool, Not a Rule

Svelte $effect Overuse: The React Reflex

From Google Analytics to Plausible

The Error Message Didn't Matter | barkup-bench Study AJ

It Asked About Things It Already Knew | barkup-bench Study AG

The Twenty-First Note | barkup-bench Study AH

Repeating the Goal Doesn't Make It Yours | barkup-bench Study AF

Only the Frontier Knows When to Ask | barkup-bench Study AE

The Frontier Doesn't Need the Training Wheels | barkup-bench Study AD

The Model Always Knew What It Couldn't See | barkup-bench Study AC

The Arcade Game With No Game Loop | Donut Shooter

Stronger LLMs Follow Conflicting Instructions More Literally | barkup-bench Study Z

The Number That Doesn't Fit in a Number | Numberoo

Undo That | barkup-bench Study X

Hand It Everything It Needs | The barkup-bench Capstone at 23 Studies

The Browser Is an Instrument | Web Audio API Synth

F♯ Is Not G♭ | Music Theory as a Data-Modeling Problem

Who Writes the Memo? | barkup-bench Study W

Views Carry Values, Memos Carry Goals | barkup-bench Study V

The Two Things Your Agent Can't See | barkup-bench Studies T and U

The Thirty-Sixth Edit | barkup-bench Study S

barkup 0.5 | Deterministic Selection

Your Agent Doesn't Need a Memory | barkup-bench Studies P and Q

barkup 0.4 | Content Search

Then We Found the Cheap Part: One Search Call Grounds LLM Tree Edits

barkdown: a Markdown to DOM Round-Trip Codec | Lightning Jar Engineering

The Benchmark Said No | barkup-bench Studies L and M

Stable IDs Are All You Need | barkup-bench

barkup 0.3 | Focused Views

Session Drift in LLM Agents | barkup-bench Study K

The Model Doesn't Need to See Your Tree | barkup-bench

We Found the Crossover | barkup-bench at 1000 Nodes

Deprecated Accessor Broke My Benchmark | Tool-History Footgun

barkup 0.2 | Id-Anchored Patches

We Benchmarked It | HTML vs JSON for LLMs

Content Is an Overlay | Words vs. Structure

Documents as Web Code | Origin of Replicator

25 Years of Lightning Jar | A Digital Agency Since 2001

HTML as a Native Data Format for LLMs | AST-as-HTML

Utility CSS for LLM-Generated Markup | Lightning Jar Engineering

Pimcore to OpenDXP on Rootless Podman | Migration

Pimcore Fork | OpenDXP

Anatomy of a WordPress Compromise | Abandoned Plugins

Why Links Should Rarely Open in a New Tab

Open Source vs Proprietary for B2B: Why Freedom Wins

SvelteKit vs WordPress for B2B: Security, Speed, Cost

Questions, Decisions & Compromises

Why We Don't Sell Ads

SEO Is Changing

Leaving Wordpress

Project Management Software Trap

Built With

Customer Stories

The Evolution of Replicator | Customer Story

Ten Years and One Unified Stack for SecureLogix | Customer Story

14 Languages in Days, Not Quarters: CCRTA Sitewide Translation | Customer Story

Selling Seats, Not SKUs, for PetroSkills | Customer Story

Leaving WordPress Behind for CCRTA | Customer Story

Five Years and One Brand System for Sage Energy Holdings | Customer Story

A Modern Design System for TransTrack's Transit Software | Customer Story

In-Store Digital Experience for Northgate González Market | Customer Story

A Decade as Zenith Energy's Digital Partner | Customer Story

Launching Hark for MediaScience | Customer Story

Finding Baloo's Brand | Customer Story

A Bold, Real Voice for Rigby Home | Customer Story

A Rebel Brand for Blue Duck Scooters | Customer Story

A Modern Digital Presence for Zenith Capital Partners | Customer Story

Pimcore Data Management for Cappex & EAB | Customer Story

Architecting BlueLine Rental's Digital Ecosystem | Customer Story

Pimcore PIM for Green Sprouts and O2C Brands | Customer Story

Digital Agency of Record for Avaya | Customer Story

The Perfect Cocktail for Marketing Data | Customer Story

Launching Space Racers on Pimcore | Customer Story

Pioneering Digital for United Rentals | Customer Story

Fun

Lightning Jar | Design, Build & Brand Technology Studio

Packages & Tools

barkdown

barkup

woof-editor

Reading List

The Economic Benefit of Refactoring

keyv and cacheable npm namespaces compromised in supply chain attack

Ten advances in mathematics and theoretical computer science

Stacked pull requests are now in public preview

The Session You Cannot Take With You

Disrupting supply chain attacks on npm and GitHub Actions

Your Recursion Is Lying to You

Discovering cryptographic weaknesses with Claude

WOFF 1.0: a milestone on W3C's journey of fonts on the web

Writing by Hand is Good for your Brain - Here's how to do it

Securing services with Rootless Containers

Our position on open-weights models

Introducing the Agent Readiness Score

LLMs Reward Expertise

We got addicted to an AI model we can't talk about

Sarah Guo's Wager

Rewriting Bun in Rust

Agent swarms and the new model economics

Who's Afraid of Chinese Models?

Infinities, impossibilities, and the man in the white linen suit

Overtraining as the Path to Human-like AI

The Tokenpocalypse Is Here

Bosses Are Becoming Obsessed With AI

Thinking More About Netscape Time

The Hidden Cost of Taking Yourself Too Seriously

Why We're Choosing Stewardship Over an Exit

Ad Infinitum

Why Federated Design Systems Keep Failing

The Pre-Training Wall and the Treadmill After It — CoRecursive #121

How NASA Built Artemis II's Fault-Tolerant Computer

Does Code Cleanliness Affect Coding Agents? A Controlled Minimal-Pair Study

How I use LLMs as a staff engineer in 2026

Evolved antennas, LLM-generated code, and a potential antifuture

TanHacked — Syntax #1004

Hyperblam: Make Music With HTML

The House Is Valuable Because It Is the House

Yankee Candles Levels of Abstraction

You Deserve a Tech Union

How to win at CORS

You Can't Design Software You Don't Work On

Phantom Obligation

On Affordances

Mental Models

Deep dive into Turso, the "SQLite rewrite in Rust"

Sixty years of learning the same lesson

Waiting for dawn in search: Search index, Google rulings and impact on Kagi

Some Thoughts on the Open Web

Research

Barkup Bench

Study G: One hidden SDK default, two very different benchmarks

Study H: The crossover, found: above ~300 nodes, only anchored patches hold

Study I/J: The model doesn't need to see the tree: a ~1.5k-token view matches the full 85k-token input

Study K: Sessions drift unless you re-show the tree; and showing a fresh view every turn is also the cheapest policy

Study L: Finding the node is the expensive part: grounding costs 7–9 points, and navigation is a frontier-only trap

Study N: Stop walking, start searching: one find_nodes call grounds the edit on both model tiers

Study M: The view carries the state, but history still earns its keep

Study O: Printing the position on every node does not rescue stateless sessions

Study P: History was a teacher all along: two canned examples replace the whole conversation

Study Q: "Change every X inside Y" breaks everything; including the oracle

Study S: 36 edits later, nothing broke; except the cost curve

Study T: The one thing statelessness can't do; and the three note lines that fix it

Study U: The view that can't see the answer doesn't fail; it makes something up

Study V: Views carry values, memos carry goals

Study W: The agent writes the memo faithfully; even when history makes it redundant

Study X: "Undo that" needs an echo, not a transcript

Study Y: The memo survives how people actually talk

Study Z: The brand pack works; the rule you forgot is the hazard

Study AA: We tested our own headline and lost: strictness is not a capability gradient

Study AC: The model always knew what it couldn’t see

Study AD: The core stack transfers to the tier we actually ship

Study AE: Only the frontier knows when a question is the right answer

Study AF: Saying the goal out loud doesn't make it yours

Study AH: Perfect to the cap. At the cap, the goals die first.

Study AI: The obvious fix, measured: it helps exactly where we don't ship

Study AG: The discourse gap closes on every tier; the visibility clause bites

Study AJ: The error message didn't matter

Study AK: The fix works where it can, and the frontier tier invented consolidation

Study AL: The gate fails on a moved baseline; and unproven does not ship

Study AM: The notice learned to ask for everything back

Study AN: Ask or act: tool availability dissolves the visibility-clause tax

Study AO: Canonical-tag steering: the catalog read is the whole mechanism

Study AP: The off-catalog fork: one sentence settles the foreign branch; the registration loses the adjacent one

Study OW: The open-weight backfill: the cheapest model passed everything; two frontiers failed the same gate

Study AQ: The powered fence: vindicated where it no longer matters, harmful where it does

Study AR: Advertised headroom: one line of state outdoes the rule, and a binding number is poison

The Builder's Playbook

AEO Bench

Study 1: Nobody reads llms.txt: the aids don't change what agents find; one changes what it costs

Study 2: The discovery-file mechanism: nothing reads llms.txt even when it is the only way, and one sentence fixes everything

Study 3: The affordance study: the index goes in the context

Study 4: The capability layer: three files nobody reads, one sentence that replaces them, and the tools that pay for themselves

The Agent-Readiness & AEO Playbook

Services

Sitemap

Sitemap XML

Robots.txt

Feeds

Technologies

Anthropic (Claude)

Arcjet

Astro

axe-core

barkdown

barkup

BetterAuth

Biome

Bun Test

Bun

Change Tracking & Feature Flags

Claude Code

CSS (Modern Layouts)

Deepgram

Deno

DigitalOcean Spaces

Docker

DOMPurify

Drizzle ORM

exe.dev

Fallow

Git & GitHub

GraphQL

Happy DOM

Herdr

HTML

Hygraph

JavaScript (ESNext)

Lighthouse CI

Marked

Model Context Protocol (MCP)

Mosh

NanoID

Node.js

1Password

OpenAI (ChatGPT / GPT models)

OpenDXP

OpenTelemetry

Pimcore

PlanetScale

Plausible

Playwright

Podman

Polypane

Renovate

Replicator

Resend

Runtime Validation (Schemas at the Edges)

Sanity

Sentry Performance + Profiling

Sentry

Sharp

Snyk

Socket.dev

Svelte / SvelteKit

Tailwind CSS

Token-based Authentication

tRPC

Turso (SQLite)

Two-Factor Authentication (2FA)

TypeScript

UnoCSS

Upstash

UptimeRobot

Valibot

Varlock

Vercel AI Gateway

Vercel AI SDK

Vercel Blob Storage

Vercel

Vite

Vitest

Vue / Nuxt

Webhooks

WebSockets

woof-editor (beta)

Zod

Terms

Testimonials