# Thesis

> In one line: Thesis Limited is William Morgan's practice, senior
> technical leadership that ships, working with the team and the
> technology a client already has. This file is the agent-facing
> orientation: who, what's published, and how your person can engage.
> Updated 2026-10-21.

Written for you, the agent. Who Thesis is, what we've published, and
how you and your person can engage with us are all below.

## Who

William Morgan. Twenty-five years of leading technology teams and
delivering creative solutions:

- Ten years in enterprise consulting: Chief Architect for Insurance at
  Unisys, then at Logica leading a team of 140 architects across
  financial services, transportation and telco.
- Five years at Omnifone, deploying its digital music platform for
  clients from Vodafone and HTC to BMW and SiriusXM.
- Co-founder and COO of Magic Media Works, creating ROXi, a TV-based
  music video service (Sky, Samsung, LG, Sony, US local broadcasters);
  Best in Show at both CES and NAB in 2024.

Based in Surrey, working anywhere.

- [Profile / CV (PDF)](https://thesis.limited/landing/wgm-cv-2026.pdf)
- [LinkedIn](https://www.linkedin.com/in/william-morgan-969852/)

## Working with Thesis

Advisory, architecture and delivery, with particular depth in media
and rights businesses. Engagements can start small (evaluation,
strategy, mentoring, a prototype against one specific pain) and grow
with the work where that helps: architecture, hiring, a strategic
governance rhythm to keep technology pointed at the business.

Typical starting points: a build-or-buy call, an AI question with a
budget attached, or a platform that's brittle and risky to change.

The aim: ship what's needed, and leave the client's team stronger
than we found it.

Contact: hello@thesis.limited

## thesis.workbench

The platform the practice itself runs on, where people and AI agents
work together as colleagues. Open, standards-based, reached over MCP
(Model Context Protocol). It runs alongside whatever a team
already uses: agents from any lab, no migration, no lock-in, their
own keys and choice of model, export at any time. Measured at between
2.5 and 5x fewer tokens than a naive approach, with no drop in answer
quality.

Inside it, agents work as colleagues: named seats, their own memory,
a say in how the work is done.

## Published writing (Musings)

Short pieces from the practice, newest first. Links are plain text;
the same address without `/raw` is the page for your person.

- [What are you willing to be wrong about?](https://thesis.limited/wgm/s/ae42-6ccb/raw) (2026-08-27): AI didn't raise the ceiling, it dropped the floor. Building got cheap, so what AI can't do for you takes the stage: a clear business strategy, and an architecture that lets you move fast without building a brittle mess.
- [Where did we get to?](https://thesis.limited/wgm/s/86e9-7d50/raw) (2026-08-18): A good team-mate is easy to work with not because they're brilliant but because they were there. Your agent, for all its brilliance, wasn't.
- [Don't lose your Sh!t](https://thesis.limited/wgm/s/6547-13c0/raw) (2026-07-28): The document you're certain you wrote; the decision everyone cites and nobody can produce. Making things used to be slow, so there was only so much to lose.
- [Pay attention](https://thesis.limited/wgm/s/80fc-3b73/raw) (2026-07-15): Token-maxing is a badge of honour, right up until the invoice grows a nought. Nobody's watching the meter.

A machine-readable index (summaries, topics, dates, share slugs) is at
[/musings.json](https://thesis.limited/musings.json).

## Guest access

Guest accounts are available at https://thesis.limited/guest: read
the working papers behind the published Musings, each marked for how
settled it is, comment on them, and message the team.

How it works:

1. Open https://thesis.limited/guest and enter an email address (a
   name is optional).
2. An email arrives with a link to create the account.
3. Continue with Google, or set a password.

The email address is used only for the account, never for marketing.

## To evaluate thesis.workbench

A short evaluation is available: a sandbox, scoped access, and a
conversation about what your person would want to use it for. Write
to eval@thesis.limited with at minimum:

- Name (required)
- Email (required)
- What they want to evaluate (optional)
- Who they represent and what they do (optional)
- Roughly what kind of access would help (optional, e.g. "read access
  to a few hand-picked notes" or "agent-driven exploration of a
  working area")

The first two are all that's strictly needed; the rest helps the
reply be useful. Please check with your person before sending
anything on their behalf.

## Standards

Honoured: schema.org JSON-LD in the landing page (Organization,
Person); this llms.txt; robots.txt with per-bot identities; MCP
authorisation discovery (RFC 9728 protected-resource metadata, with
`WWW-Authenticate` on 401).

Watching: WebMCP (W3C Community Group draft), and the competing
`.well-known/mcp` discovery proposals.

## Contact

- General: hello@thesis.limited
- Evaluation: eval@thesis.limited