Cowork Telemetry Brief

Scott Moran Operator Surface Report

Thu, Mar 19 to Mon, Mar 23 plus Tue, Mar 24 so far

This machine looks less like a laptop and more like an AI control room. Across the last five full days, 84.9% of non-ambient focus stayed in build surfaces and the keyboard produced 124,691 chars. The cleanest execution day was Sat, Mar 21, while the highest-output day by raw keyboard volume was Thu, Mar 19. The main risk showing up now is not laziness or lack of depth. It is mode collision: build, comms, publishing, revenue, and webmaster work all share the same operating surface. Today so far that leak is visible in app.slack.com, search.google.com, bing.com.

84.9%
Build-stack share
Non-ambient focus that stayed inside build surfaces across the five full days.
124,691
Chars typed
Raw keyboard output across the five full days, which is a better leading indicator than app hours.
1,717
AI paste events
Paste events into Codex, Claude, and OpenWork across the five full days.
97.2% / Sat, Mar 21
Cleanest build day
The day with the highest build-share and enough keyboard output to count as real execution.
4.0h
Today so far
Observed non-ambient focus minutes already captured today.
42m
Today admin pull
Communication plus publishing/SEO minutes already showing up today.
Prescription

Protect the pure build pattern

The best operating mode in this dataset is obvious: Sat, Mar 21 ran at 97.2% build-share with 31,246 chars typed. Across the full five-day window, build surfaces still held 84.9% of non-ambient focus. The prescription is simple: keep the first serious block in Codex/Claude/Terminal territory before Slack, SEO, or GTM tabs open.

Prescription

Separate admin sweeps from creation sweeps

Today so far already spent 41m inside Slack + Search Console + Bing Webmaster surfaces. That is legitimate work, but it feels like motion because it shares the same Chrome window as build work. Treat distribution/admin work as a named sweep with a hard stop, ideally after the core build block lands.

Prescription

You are paying a real prompt-shuttle tax

Full days in this window logged 1,717 paste events into Codex, Claude, and OpenWork alone, peaking on Mon, Mar 23. That pattern says you are operating as an orchestration hub. The leverage move is to put more briefs, comparison notes, and intermediate outputs in canonical files so less value gets trapped in copy-paste hops.

Prescription

Judge days by mode mix, not raw minutes

Thu, Mar 19 was the highest-output day by keyboard volume at 39,670 chars, while other heavy-focus days had much cooler keyboards. That means raw app minutes are not enough. Track whether a day was build, synthesis, distribution, or GTM, and score success against the intended mode.

Mode Mix

Top Surfaces Across The Five Full Days

Claude
43.3h
Codex
18.8h
Google Chrome
17.1h
QuickTime Player
3.6h
OpenWork
111m
Terminal
96m
Browser Gravity

Domains That Actually Pulled The Day Around

larry.moran.bot
93m
scottpedia0.github.io
69m
github.com
62m
meet.google.com
39m
gemini.google.com
38m
deckmanager.go2.io
30m
app.slack.com
29m
aistudio.google.com
22m
Recurring Topics

What The Keyboard Kept Returning To

Agents + orchestration
357 keyword hits
Infra + auth
167 keyword hits
Reports + artifacts
64 keyword hits
Revenue + pipeline
54 keyword hits
Publishing + web
40 keyword hits

This is a lightweight keyword pass over real keystroke chunks after stripping control tokens. It is not a semantic model, but it is good at surfacing recurring obsessions.

Hourly Shape

Engagement Heatmap

12a
1
2
3
4
5
6
7
8
9
10
11
12p
1
2
3
4
5
6
7
8
9
10
11
Mar 19
Mar 20
Mar 21
Mar 22
Mar 23
Mar 24

Cell intensity combines non-ambient focus minutes and typing volume for that hour. The point is rhythm, not exact accounting.

Five-Day Readout

Day-By-Day Interpretation

Full day
Thu, Mar 19

Build + GTM crossover

23.4h focus 39,670 chars 358 AI pastes

You were still mostly building, but revenue and publishing work clearly bled into the same day. That is productive when intentional, but it is expensive if it happens before the core build work is done. Browser gravity: scottpedia0.github.io, deckmanager.go2.io.

Apps: Claude, Google Chrome, Codex
Domains: scottpedia0.github.io, deckmanager.go2.io, app.slack.com
Full day
Fri, Mar 20

Build + GTM crossover

21.1h focus 30,827 chars 189 AI pastes

You were still mostly building, but revenue and publishing work clearly bled into the same day. That is productive when intentional, but it is expensive if it happens before the core build work is done. Browser gravity: github.com, larry.moran.bot.

Apps: Claude, Google Chrome, QuickTime Player
Domains: github.com, larry.moran.bot, meet.google.com
Full day
Sat, Mar 21

Pure build day

13.5h focus 31,246 chars 254 AI pastes

This is the cleanest execution pattern in the window: 97.2% of non-ambient focus in the build stack, 31,246 chars typed, and very little browser sprawl. Main surfaces: Claude, Codex.

Apps: Claude, Codex, Google Chrome
Domains: larry.moran.bot, gemini.google.com, scottpedia0.github.io
Full day
Sun, Mar 22

Thinking / review day

17.8h focus 11,711 chars 395 AI pastes

The machine still says build, but the keyboard cooled off. This usually means design, reading, review, or longer-form AI steering rather than straight output. Main surfaces: Claude, Codex.

Apps: Claude, Codex, Google Chrome
Domains: n/a
Full day
Mon, Mar 23

Synthesis / review day

13.1h focus 11,237 chars 521 AI pastes

This reads like orchestration rather than raw production: research-model surfaces rose, paste volume was high, and the day spread across multiple copilots. Main surfaces: Codex, Google Chrome. Browser gravity: aistudio.google.com, larry.moran.bot.

Apps: Codex, Google Chrome, Claude
Domains: aistudio.google.com, larry.moran.bot, meet.google.com
Today so far
Tue, Mar 24

Distribution / admin leak

4.0h focus 9,354 chars 193 AI pastes

Today is still productive, but the browser is pulling hard toward coordination and webmaster surfaces. 33m of communication time and 9m of publishing/SEO already showed up before the day closed. Browser gravity: app.slack.com, search.google.com.

Apps: Claude, Codex, Google Chrome
Domains: app.slack.com, search.google.com, bing.com
Browser Evidence

Recurring Windows

Claude Code overview - Claude Code Docs
6.6h
Relay
90m
Go2 — End of Day Report — March 19, 2026
54m
██████/README.md at main · Scottpedia0/██████
51m
Meet - ██████ and Scott Moran
42m
Google Gemini
37m
access.moran.bot
21m
Go2 — End of Day Brief | Mar 18, 2026
15m
Method

How This Report Was Built

Raw source tables: activity_sessions, keystroke_chunks, and clipboard_events inside Scott's local cowork.db.

Ambient/system surfaces like Universal Control, Finder, and notification center are excluded from focus totals so the report does not confuse machine state with work.

Chrome is classified by domain to distinguish build support from comms, research, publishing/SEO, and revenue work.

The right way to use this report is as an operating mirror: decide what mode the day is supposed to be, then compare the actual mix against that intent.