codex usage

    Codex Weekly Usage — August 24–30, 2026

    Executive summary

    Rock Paper Scissors operations led the available August 24–27 evidence; August 28–30 remain uncovered.

    The four verified source days contain 149 active public summaries across four work arcs. Review and quality assurance is the strongest observed work-mode signal, followed by operations and fulfillment. Two outcome or decision signals and two follow-up categories remained open. The three uncovered days are unknown and are not represented as zero activity.

    Codex execution telemetry is supporting evidence, not a measure of productivity, duration, or the week's value.

    4
    Work arcs

    Distinct public work threads supported by retained Skysight summaries.

    2
    Outcome signals

    Public-safe outcomes or decisions supported by observed context.

    2
    Open follow-ups

    Follow-ups the observed record does not show as resolved.

    27
    Top-level tasks

    Top-level Codex tasks recorded in local execution logs; not a productivity score.

    Week in reviewObserved Aug 24, 3:40 AM–Aug 27, 4:00 PM

    What shaped the week

    Major work arcs

    Rock Paper Scissors operations and Other observed work are the two largest observed arcs by retained summary mentions. Counts can overlap and do not represent duration.

    Observed evidence by work arc

    Retained structured-summary mentions; overlapping evidence, not time allocation.

    Rock Paper Scissors operations
    92
    Other observed work
    40
    Research and administration
    29
    Hyphenomenon product and reporting
    16

    Rock Paper Scissors operations

    92 observed

    Catalog quality review and wholesale or fulfillment preparation.

    Other observed work

    40 observed

    Observed work that did not match a named public reporting arc.

    Research and administration

    29 observed

    Research, purchasing, and administrative follow-through.

    1 additional work arcLower-volume observed threads remain available without competing with the primary arcs.

    Hyphenomenon product and reporting

    16 observed

    Product-system review, reporting strategy, and project sequencing.

    Outcomes and decisions

    • Delivery and implementation progress

      Advanced

      Retained summaries indicate implementation, publication, deployment, or other delivery progress.

    • Review and verification decision

      Advanced

      Retained summaries indicate an approval, acceptance, verification, or passed review gate.

    Next steps and open loops

    • Outstanding follow-through

      Open

      Retained summaries identify work that remained pending or still needed follow-through.

    • Unresolved quality or reliability issue

      Open

      Retained summaries identify an unresolved issue, defect, failure, or remediation need.

    How the observed week unfolded4 day summaries
    1. Aug 24

      Rock Paper Scissors operations + Hyphenomenon product and reporting

      36 active summaries, led by operations and fulfillment signals. 14 additional observed summaries were protected.

    2. Aug 25

      Rock Paper Scissors operations + Other observed work

      43 active summaries, led by review and quality assurance signals. 17 additional observed summaries were protected.

    3. Aug 26

      Rock Paper Scissors operations + Other observed work

      35 active summaries, led by review and quality assurance signals. 4 additional observed summaries were protected.

    4. Aug 27

      Rock Paper Scissors operations + Research and administration

      35 active summaries, led by research and administration signals.

    How to read this evaluationWork-mode signals, protected context, and evidence limits

    Observed work-mode signals

    Overlapping summary-level signals, not duration or a complete time sheet.

    Review and quality assurance
    70
    Operations and fulfillment
    58
    Research and administration
    49
    Planning and decisions
    20
    Building and delivery
    10

    Privacy and evidence boundary

    35 observed summaries were excluded from public topic analysis. Their time positions remain visible so the report does not mislabel protected context as missing or inactive.

    This evaluation uses 197 summaries from four isolated Skysight source-contract v2 restores, each hash-verified with zero conflicts or disappeared records. Coverage is available for August 24–27; August 28–30 are missing and unverified, and capture completeness is not asserted. Protected topics remain excluded. Counts are overlapping evidence signals rather than duration, productivity, value, or unique task totals, and uncovered intervals remain unknown rather than zero activity.

    Evidence appendix

    Repository activity, delivery records, source evidence, and technical diagnostics support the evaluation without competing with its first read.

    Explore supporting evidenceExecution, delivery, Skysight source context, and technical telemetry

    Where The Work Landed

    Repository activity and changed lines show which codebases received attention during the report window.

    Tool Calls By Repository

    Working repository resolved from tool workdir or turn context.

    shopify-headless3,406 52%
    hyphenomenon1,300 20%
    multi-mailbox-ops836 13%
    faire-lead-ops726 11%
    Outside git metadata147 2%
    chronicle-visualizer97 1%

    Output Shape

    Compare the rollouts carrying the largest patches with the mix of file categories touched.

    File Touch Mix

    File touches grouped by category; exact counts and share of retained touches.

    Docs and governance192 83%
    Application and pipeline code25 11%
    Other files13 6%

    Verified Delivery Ledger

    Outcomes come from terminal return records. Intent-only command counts remain in the technical appendix.

    Command Outcomes

    Pull request operations - unknown
    Count93
    Share56%
    Commit operations - success
    Count28
    Share17%
    Content push operations - success
    Count23
    Share14%
    Commit operations - unknown
    Count13
    Share8%
    Pull request operations - success
    Count4
    Share2%
    Pull request operations - failure
    Count2
    Share1%
    Commit operations - failure
    Count1
    Share<1%
    Content push operations - unknown
    Count1
    Share<1%

    Repositories In Successful Push Output

    hyphenomenon
    Count1
    Share50%
    shopify-headless
    Count1
    Share50%

    Agent Stack Used

    Observed runtime evidence only; the available catalog is excluded. Status chips distinguish completed calls, skill loads, and explicit requests.

    App or connectorGmail
    Runtime providerMulti Mailbox Ops
    Loaded skillmulti-mailbox-ops
    • Invoked means a completed runtime call was recorded.
    • Loaded means a skill file was read or discovered.
    • Requested marks an explicit skill request in the source.
    • The logs do not preserve a reliable causal chain between every layer, so the diagram shows observed layers rather than inventing one.

    Apps and Connectors

    Completed connector calls recorded by the runtime.

    Gmail
    Invoked1 rollout
    235
    Shopify
    Invoked2 rollouts
    23
    Adobe
    Invoked1 rollout
    3
    Vercel
    Invoked1 rollout
    3
    Canva
    Invoked1 rollout
    1
    Google Drive
    Invoked1 rollout
    1

    Tool Providers

    Completed MCP provider calls, grouped by server namespace.

    Multi Mailbox Ops
    Invoked2 rollouts
    1,059
    Node tool runtime
    Invoked3 rollouts
    79
    Creative Production Mcp
    Invoked1 rollout
    6
    Codex App
    Invoked1 rollout
    1

    Skills

    Explicitly requested or loaded skills; application is not inferred.

    multi-mailbox-ops
    Loaded + Requested10 rollouts
    99
    control-in-app-browser
    Loaded + Requested20 rollouts
    44
    control-chrome
    Loaded + Requested12 rollouts
    33
    pdf
    Loaded + Requested7 rollouts
    28
    computer-use
    Loaded + Requested10 rollouts
    27
    shopify-admin
    Loaded + Requested11 rollouts
    24
    visual-design-critique
    Loaded + Requested18 rollouts
    23
    rps-print-order
    Loaded + Requested2 rollouts
    18
    imagegen
    Loaded + Requested5 rollouts
    17
    shopify-admin-execution
    Loaded + Requested8 rollouts
    17

    Skysight Evidence Across The Week

    This evidence view locates the public work arcs and protected-context positions that support the synthesis. The calendar grid is not expected computer activity, and values are not duration.

    Open source reportSkysight Daily — Aug 24, 2026
    12–4a4–8a8a–12p12–4p4–8p8p–12a

    Rock Paper Scissors operations

    Other observed work

    Research and administration

    Hyphenomenon product and reporting

    Protected context

    Color intensity represents overlapping structured-summary mentions. Empty outlined cells have no retained summary evidence; zero means retained evidence exists but this domain did not match. Values are not duration.
    Work domain12–4a4–8a8a–12p12–4p4–8p8p–12a
    Rock Paper Scissors operationsRock Paper Scissors operations, 12–4a: zero matches within retained summariesRock Paper Scissors operations, 4–8a: 41 summary mentionsRock Paper Scissors operations, 8a–12p: 25 summary mentionsRock Paper Scissors operations, 12–4p: 19 summary mentionsRock Paper Scissors operations, 4–8p: 6 summary mentionsRock Paper Scissors operations, 8p–12a: 1 summary mention
    Other observed workOther observed work, 12–4a: zero matches within retained summariesOther observed work, 4–8a: 7 summary mentionsOther observed work, 8a–12p: 9 summary mentionsOther observed work, 12–4p: 14 summary mentionsOther observed work, 4–8p: 10 summary mentionsOther observed work, 8p–12a: zero matches within retained summaries
    Research and administrationResearch and administration, 12–4a: 1 summary mentionResearch and administration, 4–8a: 15 summary mentionsResearch and administration, 8a–12p: 5 summary mentionsResearch and administration, 12–4p: 8 summary mentionsResearch and administration, 4–8p: zero matches within retained summariesResearch and administration, 8p–12a: zero matches within retained summaries
    Hyphenomenon product and reportingHyphenomenon product and reporting, 12–4a: zero matches within retained summariesHyphenomenon product and reporting, 4–8a: 8 summary mentionsHyphenomenon product and reporting, 8a–12p: 8 summary mentionsHyphenomenon product and reporting, 12–4p: zero matches within retained summariesHyphenomenon product and reporting, 4–8p: zero matches within retained summariesHyphenomenon product and reporting, 8p–12a: zero matches within retained summaries
    Protected contextProtected context, 12–4a: 1 withheld summary mentionProtected context, 4–8a: 7 withheld summary mentionsProtected context, 8a–12p: 12 withheld summary mentionsProtected context, 12–4p: 12 withheld summary mentionsProtected context, 4–8p: 3 withheld summary mentionsProtected context, 8p–12a: zero matches within retained summaries

    Structured Evidence Coverage

    Retained summaries
    197
    Active public summaries
    149
    Protected summaries
    Topic excluded; observed position retained.
    35
    Inactive or locked summaries
    13
    Covered intervals
    197
    Expected calendar intervals
    Calendar grid only; not expected computer activity.
    1,008

    Codex Execution Activity

    Token and hourly activity describe the agent execution layer. They are supporting diagnostics, not a measure of the week's value, duration, or human productivity.

    Token Usage

    Cumulative runtime counters are differenced per rollout. Repeated snapshots are ignored; cached input stays visible instead of being mistaken for fresh context.

    Token Usage

    Retained local Codex telemetry only; not an account-wide usage or billing statement.

    994,574,582
    Total processed
    Input + output
    Input
    992,102,279
    Cached input
    942,761,472
    Non-cached input
    49,340,807
    Output
    2,472,303
    Reasoning output
    806,536

    Cached input is 95% of input. Reasoning output is included within output tokens.

    Tokens By Day

    Processed input plus output assigned to the local day of each positive cumulative-counter increment.

    Mon Aug 24329,069,901 33%
    Wed Aug 26254,578,877 26%
    Thu Aug 27228,134,595 23%
    Tue Aug 25182,791,209 18%

    Activity By Day And Clock Hour

    The day row shows each date in the report window. The hour chart below combines the same clock hour across all seven days; it is not a single-day timeline.

    Daily session timeline
    1. Mon Aug 2411
    2. Tue Aug 258
    3. Wed Aug 263
    4. Thu Aug 273
    5. Fri Aug 280
    6. Sat Aug 290
    7. Sun Aug 300

    Each count is a de-duplicated session start assigned to that local calendar day.

    Clock-hour pattern across the full week
    12 AM
    47
    2 AM
    21
    188
    4 AM
    740
    2c
    318
    2c
    6 AM
    613
    3c
    523
    1c
    8 AM
    1,442
    8c
    291
    10 AM
    277
    3c
    248
    12 PM
    370
    167
    1c
    2 PM
    590
    3c
    314
    4c
    4 PM
    209
    1c
    31
    6 PM
    61
    19
    8 PM
    29
    14
    10 PM

    Total: 6,512 tool calls. “c” marks successful commit operations in that hour.

    Technical Telemetry

    Raw tool, event, workspace, and patch diagnostics used to audit the analyzer. These counts are intentionally secondary to outcomes.

    Tool Surfaces

    Terminal/local shell6,291 97%
    Long-running task waits201 3%
    Multi-agent tools12 <1%
    Other/uncategorized5 <1%
    Automations2 <1%
    Thread tools1 <1%

    Function Calls

    wait201 91%
    open_in_codex5 2%
    send_message5 2%
    spawn_agent3 1%
    automation_update2 <1%
    followup_task2 <1%
    list_agents1 <1%
    Other2 <1%

    Custom Tool Calls

    Specific custom calls and their share of all recorded tool events.

    exec
    Count6,291
    Share97%

    Response Events

    event_msg
    Count25,355
    Share50%
    response_item
    Count24,105
    Share47%
    turn_context
    Count1,132
    Share2%
    world_state
    Count198
    Share<1%
    session_meta
    Count110
    Share<1%
    compacted
    Count69
    Share<1%
    inter_agent_communication_metadata
    Count15
    Share<1%

    Patch Applications

    Successful patch applications
    Count89
    Share100%

    Output Records

    Tool output records
    Count6,503
    Share86%
    Git refs in output
    Count686
    Share9%
    Truncated outputs
    Count276
    Share4%
    GitHub PR links emitted
    Count69
    Share<1%
    Long-running command handoffs
    Count1
    Share<1%

    Session And Log Audit

    Raw session_meta records
    Count110
    Share55%
    Active rollout log files
    Count64
    Share32%
    Unique session_meta.session_id values
    Count25
    Share13%

    Token Counter Audit

    Metered model-response increments
    7,574
    Token-count records
    7,670
    Repeated snapshots ignored
    96
    Records missing cumulative usage
    0
    Cumulative counter resets
    10

    Source Coverage

    Configured local session stores
    2
    Candidate rollout files inspected
    85
    Active rollout files
    64