Agents made my retro tech safe to use again and showed their real value as testers of ideas
The author describes using an AI agent to revive unsupported, ageing computers that were previously unsafe or impractical to use online. By compiling modern software and applying system fixes, the agent made obsolete devices usable again, illustrating how agents can reduce the effort needed to test and implement technical ideas.
One example involved a 15-year-old Arch Linux device whose browser could not meet modern encryption requirements; after several hours, the agent compiled an up-to-date cURL version and later improved performance with a maths library. It also updated a 12-year-old iMac Pro, a 10-year-old virtual-reality PC and an underpowered Surface Go, while the author argues that similar agent-led experimentation could accelerate scientific research, provided humans retain responsibility for judging meaningful measures and results.
In short
- AI agents revived several unsupported computers.
- They make technical experimentation faster and cheaper.
- Humans must still define and assess meaningful outcomes.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
AI agents can dramatically reduce the time and effort required to maintain and update technical systems, as these practical examples demonstrate. By automating labour-intensive compilation and optimisation work, agents free human experts to focus on higher-level design decisions and research priorities. This acceleration of experimentation could yield faster scientific progress whilst humans retain clear judgment over what questions matter and what results genuinely mean.
The case against
Outsourcing technical problem-solving to AI agents risks eroding human expertise and the institutional understanding that develops through hands-on technical work. Once automated systems prove reliable, maintaining genuine human oversight becomes difficult in practice, and solutions generated by agents may be adequate rather than truly optimal. There is also a reasonable question of whether the resources devoted to agent-led experimentation might be better directed towards higher-impact technical priorities.
Explained simply
Older computers can become difficult to use online when their software no longer supports current security and web standards. Updating them may mean replacing old system components or compiling newer software, which can take time and technical knowledge.
An AI agent is software that can carry out multi-step tasks, such as finding, building and installing updated programs. These tools may make it easier to try out fixes on ageing devices and to test technical ideas.
The same approach could help researchers explore possible solutions more quickly. People still need to decide which questions are worth asking and whether the results are meaningful.