Izzy Hodnett

Things I've built, and how each one works.

Nine of them. Three were built for the organization I work for; six I built on my own time. Each one has a page explaining what it solves and roughly how it is put together — written for people who do not build software and have no reason to want to.

See the work
Izzy Hodnett
Built at work

Three things I built for my employer

Social Current accredits nonprofit organizations across the United States and Canada. Two of these are in production there — one used by the public, one by the organizations going through accreditation. The third is in beta testing now.

Accreditation support effectiveness tracker

Accreditation support effectiveness tracker

accreditation.izzyhodnett.com
What it isRecords how a support team actually handles incoming questions, before and after a change to the way they work, so the effect of that change can be measured rather than argued about.
Who uses itTen people at Social Current — waiting to start
Build timeAn afternoon
Read how it works →
Social Current Standards

Social Current Standards

standards.social-current.org
What it isThe accreditation standards, published as something an organization can actually navigate instead of a document it has to scroll.
Who uses itOrganizations in accreditation across the US and Canada
Build timeThree days
Read how it works →
Accreditation FAQ Library

Accreditation FAQ Library

scribehow.com
What it isOver 150 step-by-step guides, written to carry two groups of people through two system changes happening at the same time.
Who uses itEvery organization currently going through accreditation
Build timeFive minutes a guide
Read how it works →
Built on my own time

Six things I built for myself and other people

Tabletop and board game tools. They are small, and I am saying so — but they are real software with real users, not exercises. They are where I learned what these tools can and cannot do before I proposed any of it at work. The first one took two weeks. The ones after it took between one day and one week each.

Moonrakers Command

Moonrakers Command

moonrakersapp.org
What it isA companion for a published board game: keep track of a game while it is being played, then look at how the group actually plays.
Who uses itEleven people, on the web and on Android
Build timeTwo weeks
Read how it works →
TM Stats

TM Stats

tm-stats.com
What it isSomewhere for a group that plays the same game regularly to keep its results and get real answers out of them.
Who uses itFive people — one regular playing group
Build timeThree days
Read how it works →
Sigil City Atlas

Sigil City Atlas

lotus-blossom-district-atlas.fochizzy.workers.dev
What it isA navigable atlas of an invented city, built in layers so you can move from a whole district down to a single building.
Who uses itFive people — my own tabletop group
Build timeOne day
Read how it works →
D&D 4e Parser

D&D 4e Parser

dnd4eparse.org
What it isOpens character files saved by software that no longer runs, and turns them back into something readable.
Who uses itPublic; shared alongside the Heroic 4e character builder
Build timeOne day
Read how it works →
AD&D Bestiary

AD&D Bestiary

adnd-bestiary.fochizzy.workers.dev
What it isA reference you can search mid-conversation without losing your place.
Who uses itMy tabletop group, plus a community forum
Build timeFour days
Read how it works →
Valeria Scoring and Valeria Randomizer

Valeria Scoring and Valeria Randomizer

Google Play, closed testing
What it isTwo phone apps for the same board game: one works out the final score, the other sets up a fresh game.
Who uses itNine people, in closed testing on Google Play
Build timeOne week
Read how it works →
Practice

How I work

I use three different artificial intelligence tools on the same piece of work, and I give them different jobs. This is a working habit rather than a claim about quality.

  1. One plans. Before anything is built, I work through with the first tool what the thing should do, what it should deliberately not do, and what finished looks like. Most mistakes get caught here, while they are still an argument rather than a rebuild.
  2. A second builds it. A different tool does the actual construction, working to that plan.
  3. A third checks it — and it is made by a different company. The tool that reviews the work is never the tool that produced it.

Why the third one comes from a different company

A tool asked to check its own work tends to approve it. It made certain assumptions going in, and it still holds those assumptions coming out, so it cannot see them. A tool built by a different company, trained differently, does not share them — which is why its approval is worth something and self-approval is not.

This catches one particular kind of mistake: the confident, plausible, wrong kind. That is the sort that survives a careful read, because nothing about it looks wrong. It does not make the work correct. It makes a specific failure much less likely.

Testing and safety

How I check things work, and what happens when they do not

The failures that matter are the quiet ones — the page that loads but shows the wrong thing, the link that stopped working and told nobody.

Checked before anything is published

A short program inspects this site every time before it goes out, and refuses to publish it if something is wrong. It confirms that every link promised on a page is actually there, that nothing claims to have been built by something that did not build it, and that no personal contact detail has slipped into a page that should not carry one. That last check is written to look for the shape of a phone number rather than for my phone number, so the number itself never has to be stored anywhere in order to be checked for.

These are not tests I run when I remember. Publishing fails if a check fails.

When something does break

Every change is dated and kept. If a release causes a problem, the previous version goes back in about a minute, which means fixing something badly is recoverable rather than expensive.