Adam Retter

adam@evolvedbinary.com
 

Balisage
The Markup Conference @ Online
2026-08-07

@adamretter

• Director of Evolved Binary

  • UK - Software, Consultancy, Training, and R&D

• Co-founder and Co-owner of eXist Solutions

  • Germany - TEI Publisher software

• Software Engineer / Prolific Open Source contributor

  • Elemental (NoSQL/XMLDB) - Founder

  • eXist-db - 20 Years as Core contributor (last 9 as main dev.)

  • RocksDB - 10 Years as RocksJava main developer

• W3C XQuery Working Group - Invited Expert

• Founder of EXQuery, and creator of RESTXQ

• Enjoys Research and Development, and Snowboarding

Photo by Alan Taylor on Unsplash

Photo by Phil Hearing on Unsplash

Photo by mk. s on Unsplash

• Software* produced by one or more people / agents

• Software access / distributed according to OSI: OSD terms

• Hard to define!

• Shared Ideology?

• Community?

• Governance?

• Financially / Altruistically driven?

Photo by Marc Newberry on Unsplash

• Followed a Vision

• Were Built by Humans

• Were Used by Humans

• Are Filled with Human Emotions

• Creativity

• Drama

Photo by Max Böhme on Unsplash

• Discussion of Topics

• Fast Prototyping of Ideas

• Enable Amateurs to build Software

• Augment Human Developers with Agentic Team members

• Generate Code Faster than Humans

• Find (and possibly fix) hard to detect software bugs

• Make a developer 10x more productive

Photo by Luis Aguila on Unsplash

• For those not financially compensated, many take pleasure in building software

  people contribute to FOSS because of fun (91%), altruism (85%), and kinship (80%). - opensource.com 2021

• Some pursue the newest technologies (e.g. AI/LLMs)

• Interacting with other humans and/or agents is a personal preference

  I have a religious exemption from using all generative "A.I.". - Catherine Sawers 2025

• Larger social relationships may be formed

Photo by Minh Triet on Unsplash

• Current cost of AI Services for the end user is subsidised

  • Little / No discounting for developing countries or disadvantaged backgrounds

  • True financial cost of use for AI Services remains unknown for all

• Rolling your own AI Services requires significant investment in hardware, utilities, etc.

  • There are often hidden costs

  companies are losing more money today implementing this technology than they were two years ago. - Jim Covello, Goldman Sachs 2026

Photo by William Warby on Unsplash

• As individual humans, most of us care about our environment

  • Organised groups have exploited it for financial gain

• AI Hardware and Services have a huge environmental cost

  • Mining and Manufacture, Rare Earth Metals, Electricty, Water, Transportation, Cooling, Land Acquisition, Light, Water, and Noise Pollution

  • AI Data Centers by 2030: 3% Global Electricty / 9.3T Litres of Water / 15.5K KM² Land Use / 2.5M Tonnes of e-Waste

• Incompatible with OSS4SG

Photo by Mukovhe Mavhungu on Unsplash

• AI/LLMs are not universally available

  • Export / Import Controls

• Corporate owners are registered in Sovereign States

• Differing Politics

  • Individuals - represented? e.g. MPs

  • Corporate Owners - Political Views / Lobbying

  • Sovereign States - Governments Change / Foreign Polices Change

• Companies can 'crash'

• Safety? AI - Kill Switches?

Photo by Todd Morris on Unsplash

• This a fluid landscape - mostly being tested in the US Courts at present

• Training AI/LLMs

  • Ingested copyrighted information that was not obtained legally

  • Companies are having to make large settlements

  • Use of Open Source licensed code being tested in court

  • You may not be indemnified; unless you pay for Enterprise plan

Photo by Kimberly Farmer on Unsplash

• AI Generated Works (e.g. code)

  • AI/LLMs cannot own Copyright

  • Highly likely that you do NOT own the Copyright

• Can you legally use AI/LLM generated code

  • Depends on what the LLM was trained on

    • Do all copyright/license(s) on the input allow your use?

  • Depends if you own the Copyright on the generated code

    • If you don't own the copyright on the output you can't assign it a license

Image generated by ChatGPT - "A cyberpunk version of Salvador Dalí's melting clocks painting"

• Can augment or accelerate human development

• AI/LLMs generate code at approximately 'Junior Developer' level

  • Generated code can have quality issues

    Every model is wrong more than 30% of the time - Radomir Basta, Suprmind 2026

• Project Maintainability

  • Code Review remains paramount

  • Increasing volume can be a problem for humans

  • Frustration with/by Meat Proxy(s) - No Meat Proxy

Photo by Fabian Jones on Unsplash

1. Open Source Projects are an inherently Human undertaking

2. Use of AI/LLMs impact humans:

   • Social, Economic, Environmental, Political, Legal, and Technical

3. Humans/AIs require knowledge to make decisions

4. Open Source Projects should publish their AI Policy

5. Humans/AIs can choose a project in full knowledge of its AI Policy