I counted my notifications on a Tuesday in early March 2026. All of them. Phone, laptop, watch, email, Slack, monitoring dashboards. 217 notifications in a single workday. I responded to 23. The other 194 were noise — glanced at, mentally processed, dismissed. Each one stole 3–5 seconds of attention. That's 10–15 minutes of pure interruption, spread across the day in 217 tiny cuts.
But the real cost wasn't 15 minutes. It was context-switching — the mental penalty your brain pays every time it jumps from one thing to another. A 2005 study from the University of California, Irvine found it takes an average of 23 minutes to fully return to a task after an interruption. Even glancing at a notification and swiping it away costs 60–90 seconds of re-engagement. Multiply that by 217 and the math stops being theoretical.
So I went on a notification diet. Here's the exact system, layer by layer. ⚙️
Step 0: The audit
Before cutting anything, I spent one week logging every notification. No spreadsheet — just a tally in a plain text file. Every time my phone buzzed or my screen flashed, I added a line: source, type, whether I acted on it.
After seven days, the pattern was obvious:
| Source | Daily avg | Acted on | Verdict |
|---|---|---|---|
| Slack (all channels) | 78 | 8 | Mostly noise |
| 42 | 6 | 85% automated/marketing | |
| Phone (apps) | 35 | 3 | Almost entirely useless |
| Monitoring/alerts | 31 | 4 | Alert fatigue |
| Calendar | 12 | 12 | All useful — keep |
| SMS/calls | 9 | 5 | Mixed |
| Watch | 10 | 2 | Redundant |
Total acted on: 40 out of 217. An 18% signal-to-noise ratio. For every useful notification, 4.4 useless ones trained my brain to ignore everything — including the important stuff.
This is what engineers call alert fatigue: when you get so many warnings that you stop trusting any of them. Hospitals study this problem because missed alerts kill people. In your workday, missed alerts just kill your focus. Same mechanism, lower stakes.
Layer 1: Slash the sources
Phone apps. I turned off notifications for every app except: phone calls, SMS, Signal, and my uptime monitor. Every other app — news, social media, shopping, banking, weather — silent. Not "reduced." Silent. If I need to check the weather, I'll open the app. It doesn't need to interrupt me about clouds.
Email. I unsubscribed from everything that wasn't a human writing to me specifically. Newsletters go to a separate email address I check once daily. Automated notifications from services (GitHub, Stripe, AWS) get filtered into folders with no alert. I check those folders twice a day: 9 AM and 3 PM.
Smartwatch. Turned off all notifications except phone calls. The watch is a watch again. It tells me the time. That's its job.
Result after Layer 1: 217 daily notifications dropped to about 90.
Layer 2: Fix Slack
Slack was the biggest offender. 78 notifications per day, 8 acted on. The problem wasn't Slack — it was that I sat in 24 channels, all set to "notify for all messages."
New setup:
- Muted all channels. Yes, all of them. Default state: muted.
- Unmuted three: #incidents (production issues), #deploys (deployment status), #direct (messages addressed to me).
- Keyword alerts — Slack lets you set notification keywords that trigger across all channels. I set my name and "urgent." If someone mentions me or flags something critical, I see it. Everything else: I check manually twice a day.
- Threads only. I reply in threads, not in channels. This reduces @channel noise for everyone.
- Schedule send. If it's not urgent, I schedule the message for 9 AM tomorrow. No late-night pings to colleagues.
Slack notifications dropped from 78 to about 12. Of those 12, I act on 10. That's an 83% signal ratio, up from 10%. ⚙️
Layer 3: Fix monitoring alerts
My monitoring setup — UptimeRobot for uptime checks, Grafana for metrics dashboards (visual panels showing server health in real time), custom scripts for log analysis — generated 31 alerts per day. Most were "CPU usage above 70%" or "memory usage above 80%." These aren't incidents. These are normal fluctuations that resolve themselves in minutes.
New alert tiers:
Critical (immediate, phone call via PagerDuty free tier):
- Service is DOWN
- Response time exceeds 10 seconds
- Disk usage above 95%
- SSL certificate expires within 7 days
- Database connection failures
Not a push notification — an actual phone call. If it's critical enough to interrupt deep work, it's critical enough to ring.
Warning (batched daily digest at 8 AM):
- CPU above 90% for 30+ minutes
- Memory above 90% sustained
- Unusual traffic patterns
Info (logged, no notification):
- Everything else. CPU at 75%. Memory at 80%. Normal variance. Logged to Grafana dashboards. I glance at dashboards once a day during my morning review.
Monitoring alerts dropped from 31 to an average of 2.3 per day. Critical ones wake me up. Warnings I review with morning tea. Info I scan during my daily check. 🛁
Layer 4: The five that remain
After all three layers, my daily notification count settled at about 5 that demand immediate attention:
- Incident alerts — something broke in production. 0–1 per day.
- Direct messages — a human needs a response from me specifically. 2–3 per day.
- Calendar reminders — meetings starting in 5 minutes. These stay because I will forget otherwise.
- Deploy notifications — did the deploy succeed or fail? One per deploy.
- Phone calls — if someone calls, it's urgent enough to interrupt.
Everything else is batched, silenced, or eliminated. Not "snoozed." Not "reduced." Gone or scheduled.
What changed in a week
The first two days felt wrong. Quiet. Like something was broken. I kept reaching for my phone to check... nothing. The phantom buzzing — researchers call this phantom vibration syndrome — took about three days to fade.
By day five, something shifted. I finished a 3-hour deep work block without interruption for the first time in months. Read an entire technical document in one sitting. Had a conversation without glancing at my phone once.
By week two, I was less tired at the end of the day. Not because I worked less — same hours. But my brain wasn't processing 200 micro-interruptions. The cognitive load — the total mental effort your working memory handles at any moment — was simply lower.
By month one, I'd gained back roughly 90 minutes of productive time per day. Not from the notifications themselves — from the context-switching they eliminated. 90 minutes times 20 workdays is 30 hours per month. Almost a full work week. I spent some of those hours in the bath. No regrets. 🛁
The resistance you'll face
People will say you're "hard to reach." Good. Hard to reach means hard to interrupt. It doesn't mean unresponsive — you still check Slack twice a day, you still answer emails within 4 hours, you still pick up phone calls. You just don't let 200 pings dictate the rhythm of your day.
The people who genuinely need you urgently will call. Everyone else can wait for your next check-in. And here's what I found: almost everything can wait 2 hours. Almost nothing is as urgent as it feels in the moment of the ping.
The notification diet isn't about being unavailable. It's about being deliberately available — on your terms, on your schedule, with your full attention. 🫶
Five notifications a day. That's all that matters. Everything else is noise pretending to be signal. 🫶
