E

EmpCloud

conf 0.55

Thunder Marketing · 1Touch-dev

Ranking

Claude's read of GitHub commits/PRs and Hubstaff time — not a leaderboard, not a performance review.

always shows the latest generated assessment, not a selectable period · live data last updated 29 Sept, 05:07 UTC

Ranks measured GitHub + Hubstaff signal only — not a performance review, not developer worth.

Generated by an AI model reading commit/PR content, complexity and tracked time. It cannot see DevOps, incident response, mentoring, architecture, code review depth, or work outside the connected repos — read each person's caveats before drawing any conclusion about them.

Conclusion

claude-opus-5 · period Sun Aug 30 2026 00:00:00 GMT+0000 (Coordinated Universal Time) – Tue Sep 29 2026 00:00:00 GMT+0000 (Coordinated Universal Time) · generated 2026-09-29 08:36 UTC · triggered by systemd:ranking · 25/28 people had evidence to rank on · ~$0.908

**The period was split between two kinds of delivery, and both showed up clearly.** Product code moved fastest in the POSITION intelligence work, the coxa matchday and hub platforms, the email service, and the Ecuador/global CMS network. Cloud work was just as real: Route53 and SES domain setup, Elastic Beanstalk and Amplify deploys, security group changes, and IAM user administration.

Harsh Patel and Pankaj Kumar stood out because they delivered on both surfaces, with readable, well scoped titles and sustained AWS activity across most of the month. Aziz Bohra brought the hardest reviewed changes. Rushendra Varma Mudunuru wrote no code at all but ran IAM and account administration on 21 distinct days, which is the most sustained infrastructure record on the team.

**Two cautions.** Several people show very high raw counts that thin out on reading: repeated identical AWS calls, repository dumps, and placeholder commit messages. And no task tracker is connected, so nothing here is confirmed against a requested piece of work.

Ranked (GitHub + Hubstaff evidence, evaluated by Claude)

Rank reflects measured delivery signal in this dataset, not overall developer worth — read the caveats.

#NameTierConfidenceRationaleCaveats
1Harsh Pateltop0.55Harsh is the only person strong on both surfaces at once: 144 commits, 31 PRs opened and 30 merged, and AWS write activity on 18 distinct days across 19 services. The PR titles are real, scoped delivery, not filler: "Add matchday Phase 2 durable payments, outbox worker, and EB deploy.", "Add Phase 4 offline POS foundation and fix location catalog cart merg…" and "Phase B: Coritiba branding, exit gaps, gate integration, and security hardening". At avgDifficultyPoints 4.5 he sits on the team median for complexity but well above it on volume of substantive, merged work.

coxa-1touch #53coxa-1touch #57coxa-1touch #62coxa-1touch #52coxa_crm #1NexusAI@7b527ddcoxa_crm@d1702b5

  • His 1277 infra events are inflated by repeats: the sample shows the same DeleteSecurityGroup call on sg-0f2c640678b0142bb over and over on one day. The 18 active days, not the event count, is what earns the rank.
  • A few commit subjects are too thin to review from the title alone, for example "outreach" and "pod setup and gpu changes draft".
  • No task tracker is connected, so nothing confirms these merged PRs were the work that was actually asked for.
  • Nearly all his AWS work sits in the 1touch Cloud Infrastructure account, so this view is specific to that project.
  • Not visible here: any incident he handled by reading logs or on a call, design decisions on the matchday payment flow, and time spent unblocking others.
2Pankaj Kumartop0.55Pankaj ran the public site network on AWS on 20 distinct days, the second most sustained infra record on the team, with Amplify CreateDomainAssociation and UpdateApp plus Route53 record changes, which is real domain and deploy work for the Ecuador and Brazil CMS sites. His 76 commits back the same initiative with clear titles such as "Add isolated TheSportsDB proxy under /api/sportsdb." and "Implement RemoteVozApi for editorial content and enhance sports data …". Code and cloud line up on one visible programme of work.

conexionec.com@bf2309fBrazil_CMS@a20ec6ccms@26be58blavozec.com@5d62e02ecuadorenvivo.com@4bf8c96

  • 1531 infra events is the highest number here, but many are repeated Route53 record changes in the same session. The 20 active days is the honest signal, not the total.
  • He opened no PRs, so none of this code carries a review trail and avgDifficultyPoints could not be computed.
  • Work spans three AWS accounts, with most of it in 1touch Cloud Infrastructure.
  • Not visible here: DNS or deploy problems he diagnosed without making a change, plus any coordination with the CMS content team.
3Aziz Bohratop0.55Aziz handled the most complex reviewed work in the data, avgDifficultyPoints 6.4 against a team median of 4.5, and all 5 of his PRs merged. The titles describe careful data work rather than surface changes: "feat(squadb): trust gates, provenance, volume rollup, and CI hardening", "feat(squadb): Day 3 news path (near-dup to story to incident)" and "feat(squadb): stamp retention clock on promote so sweep can age trusted rows". He also made real DNS and SES identity changes on 13 distinct days, supporting the email service he was building.

POSITION_INTELLIGENCE #16POSITION_INTELLIGENCE #21POSITION_INTELLIGENCE #28POSITION_INTELLIGENCE #17email-service@a620fccemail-service@9b23c87

  • 223 commits is by far the highest on the team against a median of 35, and 64 of the categorised ones are docs, test or chore. That gap between commit count and the 5 PRs is worth a look for commit splitting, even though the PR content itself is substantive.
  • All his AWS activity is in the 1touch Cloud Infrastructure account.
  • No task tracker confirms the Squad B PRs were the requested scope.
  • Not visible here: the deploy troubleshooting behind "add single-EC2 homepage deploy" that left no AWS write, and any review or design input he gave others.
4Rushendra Varma Mudunurutop0.40Rushendra wrote no code this period and still has the most sustained infrastructure record on the team: AWS writes on 21 distinct days across 17 services. The events are account and access administration, not noise: AttachUserPolicy, DeleteUser, RemoveUserFromGroup and DeactivateMFADevice for a leaver, plus RequestServiceQuotaIncrease. That is the work that keeps other people able to deploy, and it is ranked here on its own merit.
  • Zero commits and zero PRs, so there is no code sample to read and no complexity score.
  • Tracked activity percentage is low at 10.8, and all 150.2 hours landed on an unmapped project, so time cannot be tied to the AWS work directly.
  • Almost all events are in the 1touch Cloud Infrastructure account.
  • Not visible here: the access requests, offboarding checks and security conversations that sit behind each IAM change, and any console reading or log review that left no write.
5Anshuman Parmartop0.55Anshuman carried the integration load on POSITION_INTELLIGENCE with 24 PRs opened and 21 merged, including "Integration: merge all live feature branches into one staging baseline" and "feat(collector): 10-state sources, source-health APIs, failure tracking, intel task registration". He also stood up Elastic Beanstalk environments on AWS across 12 distinct days, with CreateLogGroup, PutRetentionPolicy and autoscaling suspend and resume around deploys. Strong breadth, genuinely useful merge and release work.

POSITION_INTELLIGENCE #11POSITION_INTELLIGENCE #19POSITION_INTELLIGENCE #26POSITION_INTELLIGENCE #22POSITION_INTELLIGENCE@123d614

  • His 97381 total churn is misleading. Much of it is merge and snapshot volume, for example a 162235 line commit titled "snapshot uncommitted persona fusion, contracts/v2, dat…", which is not authored complexity. His avgDifficultyPoints of 3.3 is below the 4.5 median, and that is the truer read.
  • 45 of his categorised commits fall under "other", so a large part of the record is hard to classify from titles.
  • 269 of his 319 AWS events are in "account 1" rather than the main infrastructure account.
  • Not visible here: deploy babysitting and rollback decisions during integration, and the coordination with each squad whose branches he merged.
6Ankur Kumarsolid0.55Ankur delivered broad product work across coxa-hub, coxa_crm and NexusAI, with clear titles such as "migrate Nexus storage to AWS S3" and "neighborhood hub matching, empresas pagination, claim flow order,". He paired that with security group changes on AWS across 15 distinct days, which is well above the median of 1 active day and reads like ongoing upkeep rather than a one off burst. 137 commits is high but the sample holds up on reading.

NexusAI@db85c53coxa-hub@db88753coxa_crm@4d22e8bcoxa-hub@a6f6246coxa_crm@8feb822

  • No PRs opened, so nothing was reviewed and no complexity score exists for him.
  • Two commits in the sample are scaffolding rather than delivery: "Initial commit from Create Next App" and "Initial commit from Create Twenty App", and 32 commits fall under "other".
  • His infra sample is almost all ModifySecurityGroupRules on the same two groups, so the type of infra work is narrow even though it is spread over many days.
  • Not visible here: the internship platform decisions and any mentoring on the Coxa Internship project his hours went to.
7Abhishek Kulkarnisolid0.55Abhishek's commits describe large, coherent builds rather than patches: "implement complete autonomous agentic system (§14) with…", "add provider-neutral deployment manager" and "close finished mixed Ask jobs and stop stale Need…". His avgDifficultyPoints of 5.5 is above the 4.5 median. He also provisioned EC2 and created a backups bucket on AWS, which is real setup work even though it spans only 4 days.

POSITION_INTELLIGENCE@2ff2056ZeroHumanMultiAgentV3@2a540a3ZeroHumanMultiAgentV3@e8f2c50PulseDesk-Test@35f4be9ZeroHumanMultiAgentV3 #2ZeroHumanMultiAgentV3 #3

  • Both his PRs, "ui: nav and home" and "ui: office list and department board", are unmerged and have zero reviews, so none of this work has been independently accepted.
  • 136 commits is well above the median of 35 while only 2 reached a PR, which leaves most of the work outside any review path.
  • 22 of his 30 AWS events are console sign in and MFA checks in "account 1", so the genuine infra content is the smaller set of CreateBucket, CreateKeyPair and RunInstances calls.
  • Not visible here: the architecture thinking behind the agentic system, and any work still sitting in unmerged branches.
8Rishav Rajsolid0.55Rishav has only 15 commits, well under the median of 35, but every one in the sample is a feature with a clear, specific title: "harden one-time schedule execution and retrieval", "add extras, lineage tracking, and failure snapshots" and "implement Run History Ops APIs and Saved Data r…". This is the pattern of a few well scoped pieces rather than padding. His one PR, "feat(followers): add stale bulk-job recovery", is a sensible reliability fix.

POSITION_INTELLIGENCE@5753ad9POSITION_INTELLIGENCE@b48e0e7POSITION_INTELLIGENCE@7aeed9fPOSITION_INTELLIGENCE@a78df0dPOSITION_INTELLIGENCE #27

  • That PR is unmerged with no reviews, so nothing of his was independently accepted this period.
  • "complete migration phases" at 14577 lines of churn is a very large single commit with a vague title, which makes it hard to review after the fact.
  • infra.totalEvents is 0, which may simply mean no AWS identity is mapped to him rather than that he did no cloud work.
  • Not visible here: scraper debugging against live sites, which often leaves no commit, and any pairing on the POSITION migration.
9Akash Kumarsolid0.55Akash's merged PR "fix(squadb): D3-D6 hardens persona migration and bounded re-embed" is exactly the kind of careful, bounded change worth rewarding, and it came in at difficulty 5. His email service and POSITION commits are readable day by day deliveries, such as "Day 3: audit-log + API-key screens, DNS template generator, DNS retry…". On AWS he did the matching work: SES email identities with DKIM and Route53 record changes across 4 days.

POSITION_INTELLIGENCE #31email-service@ca80d6cserver-codebase-backup@df0b34dserver-codebase-backup@0957d77POSITION_INTELLIGENCE@273157d

  • Two commits in his sample are repository dumps, not engineering: "Initial private codebase backup from OpenClaw server." at 355307 lines and "Add remaining projects to server backup." at 595458 lines. Those inflate any churn based reading of his period and should not count as delivery.
  • Only 1 PR out of 65 commits, so most of the work is unreviewed.
  • AWS work is concentrated in the 1touch Cloud Infrastructure account and clustered on a single busy day, 23 September, which is weaker than the same count spread out.
  • Not visible here: DNS propagation checks and mail deliverability testing, which rarely leave a write event.
10Jitu Prosad Sahasolid0.55Jitu delivered the SMA scraper work and solid production plumbing on the email service, with titles that say what changed: "complete SMA Instagram and Twitter profile scrapers", "add production backups and monitoring skeleton" and "add inbound pipeline reconciliation and evidence harness". 45 commits is near the median and the mix leans to real features and fixes. His small AWS footprint matches that work, with SES identities, Route53 records and enabling his own MFA.

SMA_Backend@4582fcbemail-service@a3f357bemail-service@9dcaa85email-service@23c8f0cemail-service@2ee5433

  • No PRs opened, so none of it was reviewed and there is no complexity score.
  • "add production backups and monitoring skeleton" appears twice with almost identical churn, which looks like a duplicated commit across branches rather than two pieces of work.
  • Only 2 infra active days, and part of that is his own account security setup rather than shared infrastructure.
  • Not visible here: scraper breakage chased in logs, and the evidence gathering behind "record runtime verification evidence" that happened outside the repo.
11Rushikesh Powarsolid0.40Rushikesh did the connective work across the CMS network and POSITION, including "Ship network SEO remediation for all 25 public sites (phases 0-5)." and "build incident intelligence collection foundation, add MongoDB indexe…". He also made genuine cloud changes on 8 distinct days, creating an S3 bucket for incident data and turning on bucket encryption, which is a good habit worth noting.

Global_CMS@0409205POSITION_INTELLIGENCE@0a3dfbdEcuadorCentral.com@a3880bbcms@caeab45

  • No PRs at all across 69 commits, so nothing was reviewed and no complexity score exists.
  • Several subjects are too vague to audit, for example "update" on an 8055 line change.
  • A chunk of his infra events are repeated StartInstances on the same EC2 instance i-07129b5470ae55e42 on different days, which reads more like restarting a box than building anything.
  • Not visible here: the manual site checks behind the SEO remediation, and any cross site coordination with the CMS content team.
12MOHD NAVEDsolid0.40Naved's 26 commits are below the median of 35 but the content is meaty and specific: "replace Bright Data with the Hiker, Apify and SMA I…", "queue offline writes, add Club/FanBox/POS PWAs, and fall back S…" and "DSR cross-system deletion, redaction, and cloud basel…". He also fixed named QA defects rather than vague cleanups. This looks like a smaller number of real, considered pieces of work.

POSITION_INTELLIGENCE@5aee5ffcoxa-1touch@cf96cabcoxa-1touch@ccfaa71coxa-1touch@c8b12b4

  • No PRs opened, so nothing is reviewed and there is no complexity score to check the titles against.
  • His 3 AWS events are just his own MFA and password setup on one day, so there is effectively no infrastructure evidence for him either way.
  • Not visible here: the vendor evaluation behind dropping Bright Data, and the QA conversations behind the BF-1 to BF-3 fixes.
13Manish Chandrakarsolid0.40Manish shipped across many of the global CMS sites, and the best of it is clearly described: "Add Italy sports hub, harden Serie A APIs, and soft-fail flaky upstre…" and "Add production SEO, GEO, and pre-commit lint/build checks.". That soft-fail and pre-commit work is genuinely useful hardening. The volume is high but the substance is uneven.

global-cms-Italiano@dbe84daecuadorenvivo.com@1b70e1aglobal-cms-Korean@4224366EcuadorCentral.com@a1f9bbbpasey-gol@620d366

  • 157 commits is the second highest on the team against a median of 35, yet 74 fall under "other" and the sample includes placeholders such as "first commit" and "husky test" on multi thousand line changes. That mix is the pattern the ranking is meant to discount, not reward.
  • No PRs opened, so nothing was reviewed and no complexity score exists.
  • infra.totalEvents is 0, which may mean no AWS identity is mapped rather than no cloud work, and one of his commits does add Amplify hosting config.
  • Not visible here: the per country content and localisation decisions behind each site rebrand.
14Dhiraj Ambekarsolid0.55Dhiraj built usable operator surfaces in the social media automation stack, including "bulk edit UI for personas, bots, and campaigns" and "Agent Orchestration console and Honduras boards UI", and did Elastic Beanstalk environment updates and Route53 changes on 7 distinct days. That combination of front end and deploy work is the full stack expectation in practice.

Social_Media_Automations@8392093Social_Media_Automations@2890a90SMA_Backend@cfe1b37SMA_Backend@072ec7cZeroHumanMultiAgentV3 #1

  • His one PR is unmerged at difficulty 2, and his avgDifficultyPoints of 2.0 is the lowest measured on the team, well below the 4.5 median.
  • The sample contains a feature and its immediate revert of the same 1397 line change, which suggests something went in before it was ready.
  • Some subjects are unclear, for example "update the current code of dummy number".
  • Not visible here: the debugging behind "solve bug after merge apify scrapper" and whatever prompted the revert.
15Pramod Sonisolid0.40Pramod worked steadily on article rendering and SEO across the CMS sites, with consistent, understandable titles such as "Add article routing, markdown rendering, and FAQ support" and "Add support for SEO metadata and FAQ handling in articles". 57 commits against a median of 35, and 19 of them categorised as refactor, which suggests real cleanup rather than churn.

ecuadorenvivo.life@d71b63delreporteec.com@4c27bd9noticiasecuador24.com@51d49c5panoramaec.com-@d2d1f09

  • The same FAQ and markdown change is repeated across many separate site repositories, so the count overstates how many distinct problems were solved.
  • No PRs opened, so nothing was reviewed and there is no complexity score.
  • infra.totalEvents is 0, which may mean no AWS identity is mapped for him rather than no cloud work.
  • Not visible here: any shared component decisions made across the site network, and coordination with whoever owns the CMS API.
161Touch Devsolid0.07Only 15 commits, but they are the clearest written work in the dataset and they all serve one goal, a UX audit of the sponsorship platform and then the fixes from it: "real server-side pagination for Companies, Proposals, Campaigns …", "quick-win UX audit items, internal metadata leaks, duplicate UI…" and "sidebar imbalance and double-active-highlight bug (UX audit)". Small in volume, honest and specific in content.

Marketing-Sponsorship-Automation@661ea82Marketing-Sponsorship-Automation@857adc9Marketing-Sponsorship-Automation@ed9a884Marketing-Sponsorship-Automation@8e94464

  • Zero tracked hours, zero PRs and zero AWS events, so this is a commit only record with nothing to cross check it against.
  • Evidence confidence is 7 percent, the lowest band, because there is no tracker and almost no supporting signal. That is a data gap, not a judgement on the work.
  • This account may be a shared or automation identity rather than one person, which would change how the commits should be read.
  • Not visible here: the audit conversations with James that several commit messages refer to, and any findings written up outside the repo.
17Pranav Ojhaneeds-attention0.40Pranav's 35 commits sit exactly on the median, and the useful ones are the CMS improvements: "Translation correction, layout correction, SEO enhancement" and "Dev Console created for intelligence ops". The problem is legibility. Four of the ten sampled commits are simply "initial commit" on ten thousand line changes, which makes the work impossible to review or credit properly.

global-cms-hindi@fba9fb9POSITION_INTELLIGENCE@0ee7869lavozec.com@2c1c466informeec.com@69d8ff7

  • 32 of 35 commits fall under "other" and several subjects are placeholders, so the record does not show what was actually decided or built.
  • No PRs opened, so nothing was reviewed and no complexity score exists.
  • His 3 AWS events are only his own password change and MFA setup, so there is no real infrastructure evidence.
  • Not visible here: whatever design or content work sits behind the new site repositories, which the commit messages simply do not describe.
18Cursor Agentneeds-attention0.07Six small commits in the email service, all on domain verification behaviour, for example "Add manual domain DNS verification" and "Keep manual verified domains on public DNS checks". The titles are clear and the changes are coherent as a set, but the total is tiny against a median of 35 commits.

email-service@b9bc59demail-service@441362femail-service@dacd999

  • This is an AI agent identity, so the commits belong to whoever directed it. Credit should go to that person, not to this row.
  • No tracked hours, no PRs and no AWS events, so there is nothing to cross check.
  • Evidence confidence is 7 percent, which reflects missing connections rather than poor work.
  • Not visible here: the human review and prompting effort behind each of these changes.
19Ubuntu Paperclipneeds-attention0.07Four commits in ZeroHumanMultiAgentV3 with readable product intent, such as "collapse nav to six places and focus Home on one decision." and "record a goal outcome, baseline, target, date, owner, and linked …". Very small footprint and the nav change duplicates a commit attributed to Dhiraj at identical churn, so this looks like a machine or shared identity rather than a separate contributor.

ZeroHumanMultiAgentV3@1a3ed68ZeroHumanMultiAgentV3@5daa0e9ZeroHumanMultiAgentV3@6d348c6

  • The 1771 line nav commit appears under two different identities with the same churn, so authorship attribution here is unreliable.
  • No tracked hours, no PRs and no AWS events.
  • Evidence confidence is 7 percent, meaning almost nothing supports or contradicts this record.
  • Not visible here: who actually directed and reviewed these changes.
20Claude Codeneeds-attention0.07One commit, "Fix AttributeError crashing entire Gemini structuring batches", at 44 lines. It is a well titled fix of a real crash, and small fixes that stop batch failures matter, but a single commit is all the evidence there is.

Football_Scraping_Main@5f168e4

  • This is an AI agent identity, so the work belongs to whoever directed it.
  • No tracked hours, no PRs and no AWS events.
  • Evidence confidence is 7 percent, which is a missing data signal, not a performance one.
  • Not visible here: the investigation that found the crash in the first place.
21mohammad saadneeds-attention0.40185 hours tracked on Thunder Marketing's Project and nothing else the system can see: no commits, no PRs and no AWS events. Hours alone are activity, not delivery, so there is no way to say what was produced. This needs a conversation, not a conclusion.
  • infra.totalEvents is 0, which can equally mean no AWS identity is mapped to him. If his role is ops leaning, his real work may be entirely invisible here.
  • All hours went to an unmapped Hubstaff project, which is a configuration gap on our side.
  • If he works in a repository we do not track, or under a different git identity, none of it would appear.
  • Not visible here: QA, research, support, design or coordination work of any kind, which is the most likely explanation for a large hours figure with no artifacts.
22Anmol Musicneeds-attention0.40143.4 hours on a project called Music and no commits, PRs or AWS events. The project name suggests this may not be engineering work at all, in which case a developer delivery ranking is the wrong lens for it. There is nothing here to read.
  • The Music project is not mapped to any repository, so we cannot tell what the time produced.
  • infra.totalEvents is 0, which may mean no AWS identity is mapped rather than no cloud work.
  • Someone should confirm whether this person is in a developer role at all before any inference is drawn.
  • Not visible here: any creative, content or non code output, which this dataset has no way to record.
23Srija Bandapallyneeds-attention0.4016.1 hours tracked on Thunder Marketing's Project, with no commits, PRs or AWS activity. That is a very small slice of the period and there is no artifact attached to it, so no delivery judgement is possible.
  • Two separate employee records carry this name, ids 17 and 179, so the picture may be split across accounts and should be merged before any reading.
  • Hours landed on an unmapped project, so we cannot tie them to any repository.
  • infra.totalEvents is 0, which may simply mean no AWS identity is mapped.
  • Not visible here: onboarding, shadowing or any analysis work that produces no commit.
24Shakir Khanneeds-attention0.404.1 hours tracked, all on the CMS project, and no commits, PRs or AWS events. Almost all of that time is mapped cleanly to a project, which is good hygiene, but there is simply too little of it to assess delivery.
  • 4.1 hours against a team median of 171.4 suggests a partial period, a new start or a part time arrangement, none of which this data can distinguish.
  • infra.totalEvents is 0, which may mean no AWS identity is mapped for him.
  • Not visible here: anything done in the CMS project that did not reach a commit, such as review, testing or content checks.
25Srija Bandapallyneeds-attention0.402.2 hours tracked on Strategy and Financial Consultant, with no commits, PRs or AWS events. This looks like a second record for the same person as id 17 and it covers a non engineering project. There is effectively nothing to rank.
  • Duplicate identity with employee 17 is likely, and should be resolved before this row is used for anything.
  • The project is a consulting one, so a developer delivery ranking is probably the wrong frame for this time.
  • infra.totalEvents is 0, which may mean no AWS identity is mapped.
  • Not visible here: advisory or analysis output, which never appears in GitHub or CloudTrail.

Not ranked — no evidence this period

Listed, never guessed. Ranking someone here would fabricate a comparison the data can't support.

Cursor Agent, Ruhani Allagh

literally no GitHub, Hubstaff or AWS evidence this period

Wasiu Adepoju

literally no GitHub, Hubstaff or AWS evidence this period, only a single tracked interval of under a minute

What this data cannot see

  • Production incident response that leaves no AWS write: reading logs, restarting something by hand on a server, a phone call or a Slack thread at 2am.
  • Anyone without a mapped AWS identity looks like they did zero infrastructure work. Several people here show 0 events and that may only mean the mapping is missing.
  • Code review depth, mentoring, onboarding and pairing. Review counts are 0 across the board in this data, which is more likely a collection gap than a team that never reviews.
  • Architecture and vendor decisions, and the discussion that led to them. Only the resulting code shows up, and often under one person's name.
  • Work sitting in unmerged branches, local experiments, and anything in a repository we do not track or committed under a different git identity.
  • QA, content, design, research and customer facing work. Hours show up but produce no commit or CloudTrail event, so those people look empty here.
  • No task tracker is connected, so nothing in this dataset confirms that what was built is what was asked for, or that it was accepted.
  • AI agent identities carry commits that a human actually directed and reviewed, so credit can land on the wrong row.