The open source index for ZyxWorks.
One page, listing the tools the studio built for its own work and gave away. Live at zyxworks.github.io.
| MurmurFlow | Press-to-talk dictation that runs on your own machine. Hold a key, say it, let go, and the words land at your cursor in any app. macOS and Windows. Source |
| Agent Office | Four coding agents in one tmux window, each in its own git worktree. The one that has stopped and is waiting on you says so on its border. Source |
Both came out of Zyx, the OS the studio runs on.
Each one keeps its own product page in its own repo under docs/, so a tool and
the page selling it ship together and can never drift apart. This repo owns the
index, nothing else. It used to own the host's robots.txt as well, because it
used to serve the host root; see the note below for what the repo rename changed.
index.html plus three self-hosted woff2 files. That is the whole site: 54 kB,
no build step, no framework, no package manager, no Google Fonts request, nothing
to run before deploying and nothing that can rot between deploys. GitHub Pages
serves the repo root on every push to main.
The same rule holds on the two product pages, and it is the same promise the products themselves make: a page selling a tool that phones home to nobody should not phone home either.
Design decisions live in comments inside index.html, next to the code they
explain: why the site is dark only, why the backdrop field has a separate narrow
placement for phones, why it is served from a personal account rather than an
organisation.
The site is deliberately not indexed yet, and both halves of that are on purpose:
robots.txtlets Googlebot and Bingbot crawl. Blocking a crawler cannot remove a URL from an index, it only freezes whatever is already in there.- the
noindexmeta tag in the HTML is what actually removes the page, and a crawler has to be let in to read it. - everyone else, AI crawlers included, is disallowed on all paths.
Lift this once there is a proper Impressum. Any new page in this repo needs the same two tags, or it becomes the one indexed page here.
This file is back at the host root. A user site has to be the repo named
<owner>.github.io, and under the personal account this repo was a project page
whose robots.txt the host never read. Owned by the ZyxWorks organization it
is that organization's user site, so zyxworks.github.io/robots.txt is served and
the "everyone else, stay out" rule covers the host and its sibling project pages
again (/murmurflow, /agent-office).
ZyxWorks is a product studio and forward deployed engineering practice. These are the things we built for ourselves and gave away.
Tools: MurmurFlow · Agent Office
Studio: what we do for companies · Zyx · working with us · GitHub