It's true now, and always has been: healthy constraints beget growth.
This principle is especially relevant when you want to re‑engineer your core workflows so you can capture the full benefits of today's ever‑improving "word‑crunching" computers. Their capabilities are expanding far faster than even the smartest engineers can retrofit 20th‑century tools and processes to take advantage of them (interpret that last bit however you'd like 😉).
I have adopted a set of constraints (first with discomfort, now with relief) that keep pace with this expanding scope of capability. Below, I outline them in roughly the order I applied them. The first is a hard rule; the rest were soft at first but are solidifying over time.
1. The Hard Stop
Every day at 5:30 p.m. I completely unplug. I leave my phone, computer, and @LimitlessAI pendant in the office and do not touch them again until the next morning.
- Physical distance removes the temptation to redirect attention away from the people and matters that deserve it at home.
- My sleep quality has improved dramatically; I now wake naturally between 5:30 a.m. and 7:00 a.m., fully refreshed.
Whatever digital trivia accumulates in my inboxes overnight simply doesn't bother me.
2. The Sunday Blackout
On Sundays I am totally unreachable-except via literal word of mouth. This has proved invaluable for me, my family, and, I believe, everyone who works with me.
3. No Feeds First Thing
I avoid any feed, e.g., Slack, email, X, at the start of the day.
How many of us have begun on the wrong foot after glancing at a Slack ping or email subject line?
Those cul‑de‑sacs devour our four most valuable hours of attention and return us to where we started, no closer to our true goals.
4. AI as First Pass Filter
Instead of diving straight into raw data:
1. I use Claude Max or @ChatGPTapp Pro to mediate my initial interaction with inboxes (the connectors are good enough to be useful now, and they're getting better).
2. "Together," we separate opportunities (the serendipitous stuff) from commitments (the handful of meetings that truly matter).
3. Only then do I decide where to invest my finite attention.
5. Dictation Before Typing
Before I "write" a line of code:
- I dictate my intentions for 15-20 minutes into Claude Code via @superwhisperapp + a buttery‑smooth Ghostty terminal window.
- Claude 4 Opus plans before implementing, and so I end up spending most of my time on "negotiating" the right outcome, formalizing it as best I can (while providing sufficient context to perform the task), and then verifying the plan + final output; with enough focus, you can generally "two-shot," i.e., plan and execute, a relatively complex task in 5 or fewer keystrokes.
The rule of thumb: as little typing as possible, and the words you do speak are worth the model's limited attention. Clear, verbal thought directs the machine toward something worthwhile.
6. The 30‑Minute Meeting Buffer
High‑quality meetings demand preparation and swift follow‑up. I therefore:
- Block 30 minutes before every meeting for review and mindset shifting.
- Block 30 minutes after for immediate action on commitments.
This slack lets me stay attentive, calm, and present, while ensuring that insights become progress instead of backlog; @howie_ai knows this and incorporates my preferences into coordination with prospects, clients, and collaborators.
The Upshot
When you use AI to preserve and extend your attention toward the one that truly deserves it, you must ruthlessly prioritize:
- Delegate where possible.
- Say "no" when necessary.
- Build qualification and filtering mechanisms that vaporize digital trivia before it enters your "context window.
These practices are still evolving; they're "good," but not yet "best.
As I learn more, I'll share details on using AI for delegation, asynchronous (and liquid) team management, and the twin arts of negotiation and verification-especially under uncertainty, in the "out‑of‑distribution" spaces where our labor will increasingly complement AI's capabilities.