## TL;DR
"Works on my colleague's machine" is never random, it is an unseen difference between two setups. Your job is to turn a vague comparison into a concrete diff: same browser, same version, same network, same account, same data. The moment you find the first difference, test it, because the fix usually hides right there.

## The query

```text
how to diagnose "it works on my colleague's machine"
```

## Use this when

- A bug hits one user but not another
- The reporter cites a colleague as proof it should work
- QA cannot reproduce what support sees
- You need the environment-diff checklist
- A ticket is stuck on "cannot reproduce"

## Not for

- Outages hitting everyone at once
- Bugs you have already reproduced yourself
- Pure network problems
- Permission differences that are already documented

## Steps

### 1. Confirm they are actually doing the same thing

Ask both people to describe the exact clicks, on the exact page, with the exact same record or file. Half the time the colleague did something slightly different: a different filter, a different tab, a different row. Same-ish is not same.

Expected output: a confirmed identical repro path, or the difference in steps found.

### 2. Diff the browsers and versions

Get the browser name and version from both machines, plus any extensions. An outdated browser, a beta channel, or one aggressive ad blocker explains a huge share of these. Have the failing user retry in incognito to strip extensions in one move.

Expected output: browser difference found, or browsers ruled out.

### 3. Diff the networks

Office wifi versus home wifi versus VPN. Have the failing user try a phone hotspot. If the hotspot works, the office network is interfering, and the colleague was never a valid comparison because they were on a different network.

Expected output: network confirmed as the variable, or ruled out.

### 4. Diff the accounts and permissions

Log in as the affected user yourself if your tools allow it, or compare their role against the colleague's. Different plans, roles, teams, or feature flags produce exactly this symptom: same page, different experience, both "correct."

Expected output: a permission or account difference found, or ruled out.

### 5. Diff the data

Same page with different records can behave differently: a corrupt attachment, an unusually large dataset, a record in a weird state. Have the failing user try the colleague's exact record. If the colleague's record works for them too, the data is the bug.

Expected output: the problem isolated to specific data, or data ruled out.

## Ready-to-use message

```text
"Works for them but not you" always comes down to a hidden
difference, so let's find it. Can you tell me:

1. Your browser and version, and theirs if you know it
2. Are you both on the same network, or is one of you on VPN?
3. Can you try their exact record or file on your machine?

Also try the same steps in an incognito window. That rules out
browser extensions in one shot. Tell me what changes and we'll
have it cornered.
```

## Variant phrasings

### works for my coworker but not me

Steps 1 through 3. Steps, browser, network, in that order.

### cannot reproduce bug that customer sees

Steps 2, 4, and 5. When you cannot repro, the difference is usually account state or data.

### same app different behavior two users

Step 4 first. Permissions and feature flags are the usual suspect when the machines look identical.

## Why it happens

No two machines are ever truly identical, but users experience them as identical, so any behavioral difference feels like a haunted bug. In reality the app is deterministic: same inputs, same outputs. The inputs just include forty invisible things like cached files, network routes, roles, and record states. Diagnosing this is really just making the invisible inputs visible one by one.

## Edge cases

- The colleague was on an old cached version: both users were "using the app" but running different code. Check versions first, always.
- Timezone or locale differences: date handling bugs appear for users in some regions and not others. Ask where each person is located.
- The colleague has an admin role: admins often bypass the exact checks that fail for normal users. Never use an admin's machine as the baseline.
- Intermittent on both machines: then it was never an environment difference. Look at timing, load, or flaky backends instead.
- Remote desktop or virtual machines: display scaling and clipboard quirks in virtualized environments cause UI bugs that look like app bugs. Note the setup.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_nQNyIahGJDJuonSL390cFg