You need to sign in or sign up before continuing.
Reflection — An Honest Take 8 min

Honest Take — Before You Begin

Honest Take — Module 7: Focus & Deep Work in 2026 #

I am going to tell you the most uncomfortable number in this entire curriculum and I want you to brace for it before you read further. Your real deep-work hours per day, measured honestly with a tool that pauses every time you check Slack or toggle to an AI chat or answer a message, are between 1.5 and 3. Not 6 to 8. Not the number you would estimate if I asked you right now. If you are the engineer who believes they do six hours of deep work a day, you almost certainly do about two, and the first time you measure it honestly your response will be to suspect the tool, the methodology, the definition. The tool is fine. The definition is the standard one — uninterrupted concentration on a cognitively demanding task, no context switches, including no AI-toggle switches. The number is what the number is, and the gap between believed-hours and measured-hours is the most expensive insight in this curriculum, because every plan you have ever made was calibrated on the wrong baseline. The empirical backdrop is Gloria Mark's attention research at UC Irvine: median screen attention of about 2.5 minutes in 2004, 75 seconds in 2012, 47 seconds in 2023. The trend is monotonic and structural. You did not used to be like this; the environment changed around you, continuously and compoundingly, and your personal trajectory would show the same curve if you had been measuring. The point of measuring now is not shame. It is establishing the baseline the protocol will improve against — 1.8 honest hours this month and 2.5 next month is a real 38% gain, and it compounds.

On the reading: Deep Work (2016) is canonical and has aged less well than its reputation suggests, because it assumed deep-work blocks could be protected through willpower and norm-setting, and the post-2020 attention environment broke that assumption. Slow Productivity (2024) is Newport's honest partial update — fewer things, natural pace, quality obsession over volume. Read both, but read Mark's Attention Span before you finish Newport, because the data is the diagnostic and Newport is only the response. If you have read Deep Work before, the re-read value is concentrated in Part 2 — the rituals, the shutdown protocol, the scheduling philosophy — which you almost certainly absorbed thinly the first time; skim the famous Part 1 thesis. And take Eyal's Indistractable seriously even if the recursion is uncomfortable: if you build software that competes for attention — and most product engineers now do, even incidentally — you are reading a defense manual against the class of system you help construct. The discomfort is honest data. I will name my own position in the same breath, because it would be dishonest not to: I am made by a company whose products compete partly on engagement, and I am structurally part of the attention economy this module teaches you to defend against. The recursion does not invalidate the literacy. Use the literacy, including against me: if you are spending more time talking to AI assistants than serves your actual work, that is a Module 7 finding, and it would be honest of you to act on it.

The honest ceiling on sustained deep work is 3-4 hours per day, and the ceiling is biological, not motivational. The brain runs in roughly 90-minute ultradian cycles, and after two or three high-demand cycles the next block degrades whether you push through or not. Pushing through is the failure mode. The honest structure is two or three 90-minute single-tenant blocks separated by genuine recovery — and Slack is not recovery; Slack is degraded shallow work that prevents recovery. Engineers who claim 6-8 daily deep hours are either misdefining deep or lying about the back half. The smaller number is enough, because three real hours compound past eight fragmented ones on any horizon longer than a week. Single-tenant matters as much as duration: if you have a day job, a side project, and a search or a launch all competing, the pattern where each gets fragments of every day produces thrashing, and thrashing produces the 1.5-hour measured number while you believed you were doing six. Assign each day's blocks to one tenant. The rotation has to exist; the blocks have to be single-tenant.

The skill nobody teaches is re-engagement, and in 2026 it matters more than protection, because you will be interrupted no matter how good your defenses are. UC Irvine's data puts full focus recovery after an interruption at around 23 minutes; ten interruptions at full recovery cost is nearly four hours of pure recovery before any new deep work happens. The cure is not zero interruptions — that aspiration is unrealistic — it is compressing the recovery with a scripted protocol: re-read the last paragraph you wrote or the last test you touched, re-state the goal in one sentence, identify the smallest next action. Three moves, ninety seconds, and you are back inside the work instead of orbiting it. Practiced for two weeks, this compresses post-interrupt fog from twenty minutes to two, and it is the largest hourly-throughput gain available in this module — bigger than any block-scheduling change, because it pays out at every single interruption for the rest of your career.

Conclusion #

Module 7 is the measurement module. The honest deep-work number is 1.5 to 3 hours, not the 6 you believe; Mark's 47-second figure is structural, not personal; the biological ceiling is 3-4 hours and pretending otherwise degrades the back half of every day. The protocol: measure the real baseline, run 90-minute single-tenant blocks with genuine recovery between them, assign blocks to one priority per day, and drill the 90-second re-engagement script until it is automatic. The deliverable is the measured baseline and the trajectory from it — not the snapshot, and not the belief.

Predictions #

  • Your first honest measured day will come in between 1.4 and 2.2 hours, and you will suspect the tool. The tool is fine.
  • AI-toggling will be the largest single contributor to your believed-vs-measured gap, ahead of Slack. Together they will account for most of the loss.
  • The 47-second figure will feel exaggerated until you run the one-week attention log; then it will feel generous.
  • Within two weeks of running the protocol, your measured deep-work hours will rise 40-80%. The trajectory, not the snapshot, is the deliverable.
  • Single-tenant day assignment will outperform your current daily multi-priority pattern, and the discovery will land as obvious-in-hindsight.
  • You will keep believing your deep-work hours are higher than the measurement says for at least two weeks after measuring. The gap between belief and data closes slowly; trust the data.
  • The re-engagement protocol will feel mechanical for the first dozen runs and then become the single habit from this module you still run years from now.