Productivity rain dances: why busy work isn't real work

Executive overview

Optimising your productivity system and clearing your inbox both feel like progress — but neither produces output. The unifying flaw is focusing on inputs (activity, tools, busyness) rather than outputs (what you actually ship).

Ditch the rain dance: identify the one thing your job produces that matters, then ruthlessly protect the time to do it.

What productivity rain dances are

• Term coined by Chris Williamson: rituals that feel productive but aren't linked to outcomes
• Two categories: elaborate systems (external brains, Pomodoro timers) and performative busyness (inbox zero, Slack clearing, desk-sitting)
• What unifies them: focus on input activity, not on measurable output
• Rain dances are attractive because they're easier and more satisfying than hard work
• Result: you feel busy and exhausted, but produce little of value
• From William Sim: "obsessing over process while being detached from outcomes gives you all the pain of hard work with none of the results"

Two wrong responses — and the right one

• Wrong response 1: demonise work itself ("hustle culture is a scam") — your business or job suffers
• Wrong response 2: abandon all organisation — other people fill the vacuum and direct your time
• Right response: shift attention from inputs to outputs
• Ask: what is the most valuable thing I produce? What actually helps me produce more of it?
• Output-focused answers are unglamorous, hard to make YouTube videos about, and they work

Low-tech practices that actually move the needle

• Work quotas: cap how many concurrent projects, committees, or reviews you take on — overload kills deep work quality
• Active vs. waiting lists: designate only a few projects "active"; everything else is waiting and generates no meetings or emails until it moves up
• Office hours: deflect ad hoc back-and-forth to a fixed window; protect the rest of the day
• Time-block planning: draw boxes on paper for when you'll do what — forces confrontation with overcommitment
• Deep work separation: when doing cognitively demanding work, no email, no Slack — full attention doubles quality

Meeting notes and post-meeting processing

• Use a "to discuss" column in a task board (e.g. Trello) per person or standing meeting
• If laptops are welcome: update the board live during the meeting
• If not: schedule a 15-minute processing block immediately after every meeting
• Goal: leave the meeting + processing block with zero open loops in your head

Evaluating AI tools without being a tech reporter

• You don't need to track AI developments unless that's literally your job
• Killer apps announce themselves — email and Google required no advance preparation to adopt
• AI impact will be industry-specific; the application layer matters more than new model releases
• Early productivity gains: helping novice users access capabilities that already exist in software
• Wait for the killer app to show up in your work; it will be obvious when it does

Navigating a return to the office with a side project

• Time-block both jobs explicitly: here are the hours for each
• Multi-scale planning prevents surprises — weekly and quarterly views keep you ahead
• Manage workload carefully; in-person culture makes it easy to informally take on too much
• Avoid hyperactive hive-mind collaboration: use shared documents, office hours, and fixed protocols instead of ongoing ad hoc messaging

Choosing a path toward college teaching (evidence-based planning)

• Lifestyle-centric planning sets the vision; evidence-based planning makes it real
• Don't guess or assume — go talk to actual instructors at the type of institution you want
• Ask: how did you get here, what do they look for, what is the pay and workload reality?
• Without real evidence, your plan is a fairy tale; with it, you gain an edge most candidates lack
• Adjunct vs. full-time distinction matters less than understanding the actual hiring path

Thriving as a grad student and parent

• The "grad student baby paradox": new parents often become more productive because constraints force focus
• Grad school rarely requires as many hours as students believe — it requires focused, purposeful hours
• Work a defined schedule (e.g. 9–5); most grad students waste enormous time on low-focus drifting
• Tell your advisor your constraints directly; you'll likely shift toward an autonomous working relationship
• Worst case: switch advisors — programs won't remove you for having a child

Managing high-volume communication workflows

• High-volume coordinator roles can't run on ad hoc email — that paradigm doesn't scale
• Replace ad hoc messages with structured queues: shared spreadsheets, shared folders emptied on fixed days
• Use a non-personal email address (e.g. "application-requests@") to reset sender expectations
• Office hours for queries: fixed call windows let you batch interactions and protect the rest of your day
• Retrain communicators away from "I tapped you on the shoulder" expectations toward "I entered a system"
• People want clarity and certainty their request won't be lost — not instant response

How large language models actually work (and why ChatGPT has no instincts)

• A large language model is a feed-forward network: input text is pushed through layers, each transforming it, then a token is output — nothing changes between runs
• Analogy: a Play-Doh factory — the factory doesn't remember the last batch or want anything
• Genuine instincts require four things: understanding, ongoing state, drives (goals), and actuation (ability to act)
• Current LLMs provide understanding only; the other three must be explicitly programmed
• Joe Rogan is wrong that ChatGPT spontaneously has survival drives — it can't, by architecture
• Joe is right that the risk is real: simple control loops + drive logic + actuation, connected to an LLM's world-understanding, could produce unsettling autonomous behaviour
• That combination is not what the big labs are building (the money is in Excel helper features)
• The real risk is a small actor assembling those pieces cheaply — not emergent sentience from scaling