A developer who fused Phish listening with deep coding flow for 30 years finds agent-managed work staccato and incompatible with that state.
Key Takeaways
The author built distributed systems, backend services, and a 200+ page dissertation inside a single ritual: continuous Phish playback plus uninterrupted coding sessions.
Agentic workflows (managing Claude or similar agents, reviewing merges, short context-switches) shattered the long unbroken attention arc that Phish jams and deep systems work share.
The mismatch is rhythmic: jams reward sustained presence; agent queues are interrupt-driven by design.
The author distinguishes fulfillment and creativity from output – useful work still happens, but the felt experience of being inside the work is gone.
The closing question is posed as an open engineering problem: what does flow look like in an agentic world?
Hacker News Comment Review
Commenters split on whether autocomplete-style AI preserves flow versus full agent coding destroys it; the autocomplete camp notes fewer stuck-on-obscure-bugs interruptions, while agent skeptics describe constant damage control of a fast-moving junior.
A recurring counterpoint: agent-managed coding is like bridge design where a worker pours the concrete – still engineering, just abstracted. Others reject this framing because reviewing autocompleted blocks breaks their flow immediately.
Some commenters are grieving the culture around code craft itself – debates over style, frameworks, and best practices feel irrelevant when the agent ships the feature regardless.
Notable Comments
@cyberpunk: “if we want to do it for craft it’s time to contribute to OpenBSD or something – with phish on, not for money.”
@artyom: argues some people are born engineers for life and do not need exploratory phases – the post illustrates this clearly and that such a life is not boring.
