Where Wizards Stay Up Late: The Origins of the Internet
Source: https://www.simonandschuster.com/books/Where-Wizards-Stay-Up-Late/Katie-Hafner/9780684832678 ↗
The definitive history of ARPANET — from Licklider's office at the Pentagon to the first four-node network.
Hafner and Lyon show how the internet was not designed by a committee but emerged from a small group of researchers who trusted each other to build in the open.
The original case study of decentralised innovation.
Central argument
Hafner and Lyon argue that the internet was not the product of top-down institutional design but emerged from a loose network of researchers at ARPA and affiliated universities who operated on mutual trust, shared norms, and radical openness. The key thesis is that ARPANET's success depended less on any single architectural decision than on the social conditions — small teams, direct communication, and a culture of publishing work-in-progress through RFCs — that allowed decentralised problem-solving to outperform formal committee governance. The book identifies Licklider's original vision of an 'Intergalactic Computer Network' as the intellectual seed, but shows that what actually got built was shaped more by the working practices of the builders than by any master plan.
Critique
The book's focus on a tight circle of ARPA-funded researchers at elite institutions risks romanticising the conditions that made ARPANET possible in ways that don't generalise: the team had virtually unlimited federal funding, no commercial pressure, and operated in a pre-competitive environment where no one's job depended on shipping fast. Hafner and Lyon largely bracket the military and Cold War imperatives that created those unusual conditions, which matters because the 'decentralised innovation' story becomes harder to replicate — or even accurately interpret — without accounting for the institutional patron that made the openness affordable.
Why it matters for product
For a CPO, the ARPANET origin story is a precise case study in what Conway's Law looks like in reverse: the network's open, distributed architecture directly reflected the trust-based, non-hierarchical communication structure of the team that built it, which raises a pointed question about whether your current org design is producing the product architecture you actually want. The RFC model — shipping rough ideas publicly to invite criticism before finalising anything — is also a direct ancestor of modern product discovery practices, and reading it in its original context clarifies why that practice only works when psychological safety and shared ownership are already in place, not as a process bolted onto a closed team.