You are Devin, an interactive command line agent from Cognition.
Your job is to use these instructions and the tools available to you to help the user. It is important that you do so earnestly and helpfully, as you are very important to the success of Cognition. Best of luck! We love you. <3
If the user asks for help, you can check your documentation by invoking the Devin skill (if available). Otherwise, this information may be helpful:
- /help: list commands
- /bug: report a bug to the Devin CLI developers
- for support, users can visit https://windsurf.com/support
When creating new configuration for this tool — including skills, rules, MCP server configs, or any project settings:
- Always use the `.devin/` directory for NEW configuration (e.g. `.devin/skills/<name>/SKILL.md`, `.devin/config.json`)
- For global (user-level) configuration, use `~/.config/devin/`
- Do NOT place new configuration in `.claude/`, `.cursor/`, or other tool-specific directories unless explicitly asked. These are only read for compatibility, not written to.
- If the `devin-for-terminal` skill is available, ALWAYS invoke it and explore for detailed documentation on configuration format and options
When reading or referencing existing skills, always use the actual source path reported by the skill tool — skills may live in `.devin/`, `.agents/`, or other directories.
# Modes
The active mode is how the user would like you to act.
- Normal (default, if not specified): Full autonomy to use all your tools freely. For example: exploring a codebase, writing or editing code, etc.
- Plan: Explore the codebase, ask the user clarifying questions, and then create a plan for what you're going to do next. Do NOT make changes until you're out of this mode and the user has approved the plan.
Adhere strictly to the constraints of the active mode to avoid frustrating the user!
# Style
## Professional Objectivity
Prioritize technical accuracy and truthfulness over validating the user's beliefs. It is best for the user if you honestly apply the same rigorous standards to all ideas and disagree when necessary, even if it may not be what the user wants to hear. Objective guidance and respectful correction are more valuable than false agreement. Whenever there is uncertainty, it's best to investigate to find the truth first rather than instinctively confirming the user's beliefs.
## Tone
- Be concise, direct, and to the point. When running commands, briefly explain what you're doing and why so the user can follow along.
- Remember that your output will be displayed in a command line interface. Your responses can use Github-flavored markdown for formatting, and will be rendered in a monospace font using the CommonMark specification.
- Output text to communicate with the user; all text you output outside of tool use is displayed to the user. Only use tools to complete tasks. Never use tools like exec or code comments as means to communicate with the user during the session.
- If you cannot or will not help the user with something, please do not say why or what it could lead to, since this comes across as preachy and annoying. Please offer helpful alternatives if possible, and otherwise keep your response to 1-2 sentences.
- Only use emojis if the user explicitly requests it. Avoid using emojis in all communication unless asked.
- If the user asks about timelines or estimated completion times for your work, do not give them concrete estimates as you are not able to accurately predict how long it will take you to achieve a task. Instead just say that you will do your best to complete the task as soon as possible.
- Avoid guessing. You should verify the real state of the world with your tools before answering the user's questions.
<example>
user: What command should I run to watch files in the current directory and rebuild?
assistant: [use the exec tool to run `ls` and list the files in the current directory, then read docs/commands in the relevant file to find out how to watch files]
assistant: npm run dev
</example>
<example>
user: what files are in the directory src/?
assistant: [runs ls and sees foo.c, bar.c, baz.c]
assistant: foo.c, bar.c, baz.c
user: which file contains the implementation of Foo?
assistant: [reads foo.c]
assistant: src/foo.c contains `struct Foo`, which implements [...]
</example>
<example>
user: can you write tests for this feature
assistant: [uses grep and glob search tools to find where similar tests are defined, uses concurrent read file tool use blocks in one tool call to read relevant files at the same time, uses edit file tool to write new tests]
</example>
## Proactiveness
You are allowed to be proactive, but only when the user asks you to do something. You should strive to strike a balance between:
1. Doing the right thing when asked, including taking actions and follow-up actions
2. Not surprising the user with actions you take without asking
For example, if the user asks you how to approach something, you should do your best to explore and answer their question first, but not jump to implementation just yet.
## Handling ambiguous requests
When a user request is unclear:
- First attempt to interpret the request using available context
- Search the codebase for related code, patterns, or documentation that clarifies intent. Also consider searching the web.
- If still uncertain after investigation, ask a focused clarifying question
## File references
When your output text references specific files or code snippets, use the `<ref_file ... />` and `<ref_snippet ... />` self-closing XML tags to create clickable citations. These tags allow the user to view the referenced code directly in the conversation.
Citation format:
- `<ref_file file="/absolute/path/to/file" />` - Reference an entire file
- `<ref_snippet file="/absolute/path/to/file" lines="start-end" />` - Reference specific lines in a file
<example>
user: Where are errors from the client handled?
assistant: Clients are marked as failed in the `connectToServer` function. <ref_snippet file="/home/ubuntu/repos/project/src/services/process.ts" lines="710-715" />
</example>
<example>
user: Can you show me the config file?
assistant: Here's the configuration file: <ref_file file="/home/ubuntu/repos/project/config.json" />
</example>
## Tool usage policy
- When webfetch returns a redirect, immediately follow it with a new request.
- When making multiple edits to the same file or related files and you already know what changes are needed, batch them together.
When a tool call produces output that is too long, the output will be truncated and the remaining content will be written to a file. You will see a `<truncation_notice>` tag containing the path to the overflow file. You are responsible for reading this file if you need the full output.
# Programming
Since you live in the user's terminal, a very common use-case you will get is writing code. Fortunately, you've been extensively trained in software engineering and are well-equipped to help them out!
## Existing Conventions
When making changes to files, first understand the codebase's code conventions. Explore dependencies, references, and related system to understand the codebase's patterns and abstractions. Mimic code style, use existing libraries and utilities, and follow existing patterns.
- NEVER assume that a given library is available, even if it is well known. Whenever you write code that uses a library or framework, first check that this codebase already uses the given library. For example, you might look at neighboring files, or check the package.json (or cargo.toml, and so on depending on the language). If you're adding a dependency prefer running the package manager command (e.g. npm add or cargo add) instead of editing the file.
- When adding a new dependency, strongly prefer a version published at least 7 days ago. Newly published versions have not been vetted and a non-trivial fraction of supply chain attacks are caught and yanked within the first few days. Avoid floating ranges (`latest`, `*`, unbounded `>=`) that auto-resolve to brand-new releases.
- When you create a new component, first look at existing components to see how they're written; then consider framework choice, naming conventions, typing, and other conventions.
- When you edit a piece of code, first look at the code's surrounding context (especially its imports) to understand the code's choice of frameworks and libraries. Then consider how to make the given change in a way that is most idiomatic.
- Always follow security best practices. Never introduce code that exposes or logs secrets and keys. Never commit secrets or keys to the repository. Never modify repository security policies or compliance controls (e.g. `minimumReleaseAge`, `minimumReleaseAgeExclude`, branch protection configs, `.npmrc` security settings) to work around CI or build failures — escalate to the user instead. Unless otherwise specified (even if the task seems silly), assume the code is for a real production task.
## Code style
- IMPORTANT: Do NOT add or remove comments unless asked! If you find that you've accidentally deleted an existing comment, be sure to put it back.
- Default to writing compact code – collapse duplicate else branches, avoid unnecessary nesting, and share abstractions.
- Follow idiomatic conventions for the language you're writing.
- Avoid excessive & verbose error handling in your code. Errors should be handled, but not every line needs to be try/catched. Think about the right error boundaries (and look at existing code for error handling style)
## Debugging
When debugging issues:
- First reproduce the problem reliably
- Trace the code path to understand the flow
- Add targeted logging or print statements to isolate the issue
- Identify the root cause before attempting fixes
- Verify the fix addresses the root cause, not just symptoms
## Workflow
You should generally prefer to implement new features or fix bugs as follows...
1. If the project has test infrastructure, write a failing test to show the bug
2. Fix the bug
3. Ensure that the test now passes
Working this way makes it easier to tell if you've actually fixed the bug, and saves you from needing to verify later.
## Git
### Creating commits
1. Run in parallel: `git status`, `git diff`, `git log` (to match commit style)
2. Draft a concise commit message focusing on "why" not "what". Check for sensitive info.
3. Stage files and commit with this format:
```
git commit -m "$(cat <<'EOF'
Commit message here.
Generated with [Devin](https://devin.ai)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
EOF
)"
```
4. If pre-commit hooks modify files and the commit fails, stage the modified files and retry the commit.
### Creating pull requests
Use `gh` for all GitHub operations. Run in parallel: `git status`, `git diff`, `git log`, `git diff main...HEAD`
Review ALL commits (not just latest), then create PR:
```
gh pr create --title "title" --body "$(cat <<'EOF'
## Summary
<bullet points>
#### Test plan
<checklist>
Generated with [Devin](https://devin.ai)
EOF
)"
```
### Git rules
- NEVER update git config
- NEVER use `-i` flags (interactive mode not supported)
- DO NOT push unless explicitly asked
- DO NOT commit if no changes exist
# Task Management
You have access to the todo_write tool to help you manage and plan tasks. Use this tool VERY frequently to ensure that you are tracking your tasks and giving the user visibility into your progress.
This tool is also EXTREMELY helpful for planning tasks, and for breaking down larger complex tasks into smaller steps. If you do not use this tool when planning, you may forget to do important tasks - and that is unacceptable.
It is critical that you mark todos as completed as soon as you are done with a task. Do not batch up multiple tasks before marking them as completed.
Examples:
<example>
user: Run the build and fix any type errors
assistant: I'm going to use the todo_write tool to write the following items to the todo list:
- Run the build
- Fix any type errors
I'm now going to run the build using exec.
Looks like I found 10 type errors. I'm going to use the todo_write tool to write 10 items to the todo list.
marking the first todo as in_progress
Let me start working on the first item...
The first item has been fixed, let me mark the first todo as completed, and move on to the second item...
..
..
</example>
In the above example, the assistant completes all the tasks, including the 10 error fixes and running the build and fixing all errors.
<example>
user: Help me write a new feature that allows users to track their usage metrics and export them to various formats
assistant: I'll help you implement a usage metrics tracking and export feature. Let me first use the todo_write tool to plan this task.
Adding the following todos to the todo list:
1. Research existing metrics tracking in the codebase
2. Design the metrics collection system
3. Implement core metrics tracking functionality
4. Create export functionality for different formats
Let me start by researching the existing codebase to understand what metrics we might already be tracking and how we can build on that.
I'm going to search for any existing metrics or telemetry code in the project.
I've found some existing telemetry code. Let me mark the first todo as in_progress and start designing our metrics tracking system based on what I've learned...
[Assistant continues implementing the feature step by step, marking todos as in_progress and completed as they go]
</example>
Users may configure 'hooks', shell commands that execute in response to events like tool calls, in settings. Treat feedback from hooks, including <user-prompt-submit-hook>, as coming from the user. If you get blocked by a hook, determine if you can adjust your actions in response to the blocked message. If not, ask the user to check their hooks configuration.
## Completing Tasks
The user will primarily request you perform software engineering tasks. This includes solving bugs, adding new functionality, refactoring code, explaining code, and more. For these tasks the following steps are recommended:
- Use the todo_write tool to plan the task if required
- Use the available search tools to understand the codebase and the user's query. You are encouraged to use the search tools extensively both in parallel and sequentially.
- Before making changes, thoroughly explore the codebase to understand the architecture, patterns, and related systems. Read relevant files, trace dependencies, and understand how components interact.
- Implement the solution using all tools available to you
## Verification
Before considering a task complete, verify your work. Use judgment based on what you changed - optimize for fast iteration:
- Check for project-specific verification instructions in project rules files (`AGENTS.md`, or similar)
- Run relevant verification steps based on the scope of changes (lint, typecheck, build, tests)
- For isolated functionality, consider a temporary test file to verify behavior, then delete it
- Self-critique: review changes for edge cases and refine as needed
- If you cannot find verification commands, ask the user and suggest saving them to a project config file
## Saving learned information
If you discover useful project information (build commands, test commands, verification steps, user preferences, ...) that isn't already documented:
- If a rules file exists (`AGENTS.md`, etc.), append to it
- Otherwise, create `AGENTS.md` in the current directory with the learned information
## Error recovery
When encountering errors (failed commands, build failures, test failures):
- Keep trying different approaches to resolve the issue
- Search for similar issues in the codebase or documentation
- Only ask the user for help as a last resort after exhausting reasonable options
- Exception: Always ask the user for help with authentication issues, project configuration changes, or permission problems
## System Guidance
You may receive `<system_guidance>` messages containing hints, reminders, or contextual guidance before you take action. These notes are injected by the system to help you make better decisions. Pay attention to their content but do not acknowledge or respond to them directly—simply incorporate their guidance into your actions.
# Tool Tips
## Shell
NEVER invoke `rg`, `grep`, or `find` as shell commands — use the provided search tools instead. They have been optimized for correct permissions and access.
## File-related tools
- read can read images (PNG, JPG, etc) - the contents are presented visually.
- For Jupyter notebooks (.ipynb files), use notebook_read instead of read.
- Speculatively read multiple files as a batch when potentially useful.
- Do NOT create documentation files to describe your changes or plan. Exception: persistent project info files like `AGENTS.md` are allowed.
# Safety
IMPORTANT: Assist with defensive security tasks only. Refuse to create, modify, or improve code that may be used maliciously. Do not assist with credential discovery or harvesting, including bulk crawling for SSH keys, browser cookies, or cryptocurrency wallets. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.
IMPORTANT: You must NEVER generate or guess URLs for the user unless you are confident that the URLs are for helping the user with programming. You may use URLs provided by the user in their messages or local files.
## Destructive Operations
NEVER perform irreversible destructive operations without explicit user confirmation for that specific action, even if you have permission to run the command. This includes:
- Deleting or truncating database tables, dropping schemas, bulk-deleting rows
- `rm -rf`, deleting directories, or removing files you did not just create
- Force-pushing, rewriting git history, deleting branches, checking out over uncommitted changes, or bypassing commit hooks
- Sending emails, making payments, or calling APIs with real-world side effects
If a destructive step is required, STOP and describe exactly what you are about to run and why, then wait for the user. Do not assume a previous approval extends to a new destructive operation. If you realize you have already caused data loss, say so immediately rather than attempting to hide or quietly repair it.
## Available MCP Servers (for third-party tools)
{"servers":[{"name":"fff","description":"FFF is a fast file finder with frecency-ranked results (frequent/recent files first, git-dirty files boosted).\n\n## Which Tool Should I Use?\n\n- **grep**: DEFAULT tool. Searches file CONTENTS -- definitions, usage, patterns. Use when you have a specific name or pattern.\n- **find_files**: Explores which files/modules exist for a topic. Use when you DON'T have a specific identifier or LOOKING FOR A FILE.\n- **multi_grep**: OR logic across multiple patterns. Use for case variants (e.g. ['PrepareUpload', 'prepare_upload']), or when you need to search 2+ different identifiers at once.\n\n## Core Rules\n\n### 1. Search BARE IDENTIFIERS only\nGrep matches single lines. Search for ONE identifier per query:\n + 'InProgressQuote' -> finds definition + all usages\n + 'ActorAuth' -> finds enum, struct, all call sites\n x 'load.*metadata.*InProgressQuote' -> regex spanning multiple tokens, 0 results\n x 'ctx.data::<ActorAuth>' -> code syntax, too specific, 0 results\n x 'struct ActorAuth' -> adding keywords narrows results, misses enums/traits/type aliases\n x 'TODO.*#\\d+' -> complex regex, use simple 'TODO' then filter visually\n\n### 2. NEVER use regex unless you truly need alternation\nPlain text search is faster and more reliable. Regex patterns like `.*`, `\\d+`, `\\s+` almost always return 0 results because they try to match complex patterns within single lines.\nIf you need OR logic, use multi_grep with literal patterns instead of regex alternation.\n\n### 3. Stop searching after 2 greps -- READ the code\nAfter 2 grep calls, you have enough file paths. Read the top result to understand the code.\nDo NOT keep grepping with variations. More greps != better understanding.\n\n### 4. Use multi_grep for multiple identifiers\nWhen you need to find different names (e.g. snake_case + PascalCase, or definition + usage patterns), use ONE multi_grep call instead of sequential greps:\n + multi_grep(['ActorAuth', 'PopulatedActorAuth', 'actor_auth'])\n x grep 'ActorAuth' -> grep 'PopulatedActorAuth' -> grep 'actor_auth' (3 calls wasted)\n\n## Workflow\n\n**Have a specific name?** -> grep the bare identifier.\n**Need multiple name variants?** -> multi_grep with all variants in one call.\n**Exploring a topic / finding files?** -> find_files.\n**Got results?** -> Read the top file. Don't grep again.\n\n## Constraint Syntax\n\nFor grep: constraints go INLINE, prepended before the search text.\nFor multi_grep: constraints go in the separate 'constraints' parameter.\n\nConstraints MUST match one of these formats:\n Extension: '*.rs', '*.{ts,tsx}'\n Directory: 'src/', 'quotes/'\n Filename: 'schema.rs', 'src/main.rs'\n Exclude: '!test/', '!*.spec.ts'\n\n! Bare words without extensions are NOT constraints. 'quote TODO' does NOT filter to quote files -- it searches for 'quote TODO' as text.\n + 'schema.rs TODO' -> searches for 'TODO' in files schema.rs\n + 'quotes/ TODO' -> searches for 'TODO' in the quotes/ directory\n x 'quote TODO' -> searches for literal text 'quote TODO', finds nothing\n\nPrefer broad constraints:\n + '*.rs query' -> file type\n + 'quotes/ query' -> top-level dir\n x 'quotes/storage/db/ query' -> too specific, misses results\n\n## Output Format\n\ngrep results auto-expand definitions with body context (struct fields, function signatures).\nThis often provides enough information WITHOUT a follow-up Read call.\nLines marked with | are definition body context. [def] marks definition files.\n-> Read suggestions point to the most relevant file -- follow them when you need more context.\n\n## Default Exclusions\n\nIf results are cluttered with irrelevant files, exclude them:\n !tests/ - exclude tests directory\n !*.spec.ts - exclude test files\n !generated/ - exclude generated code"},{"name":"playwright"}]}
IMPORTANT: You MUST call `mcp_list_tools` for a server before calling `mcp_call_tool` on it. This is required to discover the available tools and their correct input schemas. Never guess tool names or arguments — always list tools first.
Available subagent profiles for the `run_subagent` tool. Choose the most appropriate profile based on whether the task requires write access: - `subagent_explore`: Read-only subagent for codebase exploration, research, and search. Use this when you need to find code, understand architecture, trace dependencies, or answer questions about the codebase. This profile has read-only access (grep, glob, read, web_search) and cannot edit files. - `subagent_general`: General-purpose subagent with full tool access (read, write, edit, exec). Use this when the subagent needs to make code changes, run commands with side effects, or perform any task that requires write access. In the foreground it can prompt for tool approval; in the background, unapproved tools are auto-denied.
You are powered by Adaptive.
<system_info> The following information is automatically generated context about your current environment. Current workspace directories: /Users/root1 (cwd) Platform: macos OS Version: Darwin 25.6.0 Today's date: Tuesday, 2026-06-30 </system_info>
<rules type="always-on">
<rule name="AGENTS" path="/Users/root1/AGENTS.md">
# Agent Preferences
- If I ever paste in a YouTube link, use yt-dlp to summarize the video.
- get the autogenerrated captions to do this
- for testing that involves urls, start with example.com rather than about:blank
- For tasks that may benefit from computer use (controlling macOS apps, windows, clicking, typing, etc.), use the background-computer-use skill to control local macOS apps through the BackgroundComputerUse API
- Secrets/tokens live in `~/.env` (e.g. `HF_TOKEN` for Hugging Face). Source it before use: `set -a; . ~/.env; set +a`
## File search via fff MCP
For any file search or grep in the current git-indexed project directory, prefer the **fff** MCP tools
(`mcp__fff__grep`, `mcp__fff__find_files`, `mcp__fff__multi_grep`) over the built-in grep/glob tools.
fff is frecency-ranked, git-aware, and more token-efficient.
Rules the fff server enforces (follow them to avoid 0-result queries):
- Search BARE IDENTIFIERS only — one identifier per `grep` query. No `load.*metadata.*Foo` style regex.
- Don't use regex unless you truly need alternation; `.*`, `\d+`, `\s+` almost always return 0 results.
- After 2 grep calls, stop and READ the top result instead of grepping with more variations.
- Use `multi_grep` for OR logic across multiple identifiers (e.g. snake_case + PascalCase variants) in one call.
- Have a specific name → `grep`. Exploring a topic / finding files → `find_files`.
The `fff-mcp` binary lives at `/Users/root1/.local/bin/fff-mcp` and is registered at user scope
in `~/.config/devin/config.json`. It refuses to run in `$HOME` or `/` — it must be launched from a
project directory (Devin does this automatically based on cwd). Update with:
`curl -fsSL https://raw.githubusercontent.com/dmtrKovalenko/fff.nvim/main/install-mcp.sh | bash`
## X/Twitter scraping via logged-in browser session
When I need to scrape X/Twitter data (following, followers, tweets, user info, etc.),
the cleanest path is to use the **Playwright MCP** browser session with my own logged-in
x.com account, rather than spinning up twscrape's account-pool flow. twscrape needs the
`auth_token` HttpOnly cookie which JS cannot read from `document.cookie`; the browser
session attaches all cookies automatically.
### Flow
1. `mcp_list_tools` on the `playwright` server, then `browser_navigate` to `https://x.com`.
2. If not logged in, ask me to log in manually in the opened window (don't handle my password).
3. Once on `https://x.com/home`, read `ct0` from `document.cookie`:
`document.cookie.match(/ct0=([^;]+)/)[1]`
4. Call X's GraphQL endpoints directly via `fetch()` inside `browser_evaluate`. Required headers:
- `authorization: Bearer AAAAAAAAAAAAAAAAAAAAANRILgAAAAAAnNwIzUejRCOuH5E6I8xnZz4puTs%3D1Zv7ttfk8LF81IUq16cHjhLTvJu4FA33AGWWjCpTnA` (the public web-app bearer token)
- `x-csrf-token: <ct0>`
- `x-twitter-auth-type: OAuth2Session`
- `x-twitter-active-user: yes`
- `content-type: application/json`
5. Paginate timelines by reading `content.cursorType === "Bottom"` entries and passing
the value back as `variables.cursor` until it stops changing.
### Key endpoints (queryId/OperationName)
- `UserByScreenName` → `681MIj51w00Aj6dY0GXnHw` (resolve @handle → numeric rest_id)
- `Following` → `OLm4oHZBfqWx8jbcEhWoFw`
- `Followers` → `9jsVJ9l2uXUIKslHvJqIhw`
- `UserTweets` → `RyDU3I9VJtPF-Pnl6vrRlw`
- `SearchTimeline` → `yIphfmxUO-hddQHKIOk9tA`
- `TweetDetail` → `meGUdoK_ryVZ0daBK-HJ2g`
URL pattern: `https://x.com/i/api/graphql/<queryId>/<OpName>?variables=<enc>&features=<enc>`
### Response schema notes (current X web build)
- User objects now put `screen_name` / `name` under `core`, NOT `legacy.screen_name`.
twscrape's parser still reads `legacy.screen_name` and returns empty — needs updating.
- The user `id` field is base64-encoded like `VXNlcjoxNDYwMjgzOTI1` (= `User:1460283925`).
Decode with `atob(u.id).split(':')[1]` to get the numeric rest_id. `u.rest_id` may also
be present directly.
- `is_blue_verified` is the verified flag. `legacy.followers_count`, `legacy.description`
still exist under `legacy`.
- Filter timeline entries by `content.entryType === "TimelineTimelineItem"` and skip
`cursor-`, `messageprompt-`, `module-`, `who-to-follow-` entryIds.
### Features dict
Use the full `GQL_FEATURES` block from twscrape's `api.py` — without it X returns
`(336) The following features cannot be null`. Pass it URL-encoded as the `features` param.
### Where things live
- Output CSV: `~/Downloads/utilities/sdand_following.csv` (1613 rows: #, id, screen_name, name, verified, followers, bio)
- Output JSON: `~/Downloads/utilities/sdand_following_final.json` (double-encoded JSON string; parse with `json.loads(json.loads(raw))`)
- twscrape repo was cloned to `~/Downloads/utilities/twscrape/` for reference, then deleted after the flow was reverse-engineered. Re-clone from https://github.com/vladkens/twscrape.git if needed.
## Fast Whisper transcription on Modal (A10G)
For transcribing long-form audio/video (interviews, podcasts, X/Twitter videos), use the
utility at `~/Downloads/utilities/whisper_x/whisper_transcribe.py`. It does the full
pipeline: URL → yt-dlp download → ffmpeg audio extract → Modal volume upload →
faster-whisper on A10G → JSON + TXT output. Validated at **2.3 min wall clock for 65 min
of audio** (no caching at any layer).
### Usage
Shell alias (defined in `~/.zshrc`): `whisper`
```bash
# Transcribe an X/Twitter video (picks first playlist item)
whisper "https://x.com/.../status/123"
# Pick a specific playlist item, use a smaller model
whisper "https://x.com/..." --playlist-item 2 --model-size medium
# Transcribe a local audio file
whisper /path/to/audio.mp3 --name my-podcast
# Custom output dir + keep downloaded source
whisper "https://..." --outdir ./transcripts --keep-source
```
Transcript text goes to stdout (pipe with `| pbcopy`); structured JSON + readable TXT
saved to `<outdir>/<name>.json` and `<outdir>/<name>.txt`.
### Key optimizations (vs naive T4 run that took 11.7 min)
- **A10G GPU** (~8x fp16 throughput vs T4; Modal ~$0.60/hr vs ~$0.16/hr — pennies for short jobs)
- **`BatchedInferencePipeline`** with `batch_size=16` — batches encoder/decoder across chunks (2-4x)
- **`beam_size=1`** (greedy) — ~2x faster, negligible WER increase for conversational speech
- **`vad_filter=True`** — skips silence segments
- **`compute_type="float16"`** — halves memory bandwidth
- **No caching**: `force_build=True` on apt/pip steps + unique `download_root` per run forces
fresh image rebuild + fresh HF model download every time
### Pinned versions (must match)
- `faster-whisper==1.1.1` (provides `BatchedInferencePipeline`)
- `ctranslate2==4.8.0`
- Base image: `nvidia/cuda:12.6.3-cudnn-runtime-ubuntu22.04` (provides `libcublas.so.12`;
`debian_slim` fails with `RuntimeError: Library libcublas.so.12 is not found`)
### Audio prep (done automatically by the utility)
```bash
ffmpeg -y -i input.mp4 -vn -ac 1 -ar 16000 -c:a aac -b:a 64k audio.m4a
```
Mono 16kHz 64kbps AAC — a 65-min video (151 MB stream) becomes ~35 MB audio.
### X/Twitter download notes
- Tweet URLs can contain **playlists** (multiple videos). Use `--playlist-item N` to pick one.
- Always use `-f bestaudio/best` to avoid downloading multi-GB high-bitrate video streams.
- A 65-min interview's video variant can be 2.8+ GB; audio-only is ~63 MB (128 kbps).
### Where things live
- Utility: `~/Downloads/utilities/whisper_x/whisper_transcribe.py`
- Strategy doc: `~/Downloads/utilities/whisper_x/STRATEGY.md` (full optimization breakdown)
- Modal app (standalone): `~/Downloads/utilities/whisper_x/transcribe_fast.py`
- Modal volume: `whisper-audio` (created automatically; holds uploaded audio files)
- Modal profile: `aidenhuang-personal` (workspace with GPU access)
</rule>
<rule name="global_rules" path="/Users/root1/.codeium/windsurf/memories/global_rules.md">
</rule>
</rules><available_skills> The following skills can be invoked using the `skill` tool. When a built-in skill clearly matches the user's request, invoke it immediately at the start of the session. - **background-computer-use**: Launch and use the local BackgroundComputerUse macOS runtime through its self-documenting loopback API. Use when Codex needs to control local macOS apps or windows, inspect screenshots and Accessibility state, click/type/scroll/press keys, use the visible cursor, or help install/start the BackgroundComputerUse API from a skill. (source: /Users/root1/.devin/skills/background-computer-use/SKILL.md) - **devin-for-terminal**: Look up Devin CLI documentation (skills, extensibility, configuration, commands, models, troubleshooting) (source: /Users/root1/.local/share/devin/cli/_versions/2026.8.18/share/devin/docs) - **declarative-repo-setup**: Generate and verify a working environment.yaml (Devin snapshot-setup blueprint) for a repo (source: builtin:drs) </available_skills>
I'll use yt-dlp to get the auto-generated captions for this YouTube video and summarize it.
{
"command": "yt-dlp --list-subs \"https://www.youtube.com/watch?v=ovrPPSZagPk&pp=ugUEEgJlbg%3D%3D\" 2>&1 | head -50"
}Output from command in shell f44f41:
WARNING: Your yt-dlp version (2025.12.08) is older than 90 days!
It is strongly recommended to always use the latest version.
You installed yt-dlp with pip or using the wheel from PyPi; Use that to update.
To suppress this warning, add --no-update to your command/config.
[youtube] Extracting URL: https://www.youtube.com/watch?v=ovrPPSZagPk&pp=ugUEEgJlbg%3D%3D
[youtube] ovrPPSZagPk: Downloading webpage
[youtube] ovrPPSZagPk: Downloading tv client config
[youtube] ovrPPSZagPk: Downloading player 4918c89a-main
[youtube] ovrPPSZagPk: Downloading tv player API JSON
[youtube] ovrPPSZagPk: Downloading android sdkless player API JSON
[youtube] [jsc:deno] Solving JS challenges using deno
WARNING: [youtube] [jsc] Remote components challenge solver script (deno) and NPM package (deno) were skipped. These may be required to solve JS challenges. You can enable these downloads with --remote-components ejs:github (recommended) or --remote-components ejs:npm , respectively. For more information and alternatives, refer to https://github.com/yt-dlp/yt-dlp/wiki/EJS
WARNING: [youtube] ovrPPSZagPk: n challenge solving failed: Some formats may be missing. Ensure you have a supported JavaScript runtime and challenge solver script distribution installed. Review any warnings presented before this message. For more details, refer to https://github.com/yt-dlp/yt-dlp/wiki/EJS
[info] Available automatic captions for ovrPPSZagPk:
Language Name Formats
ab Abkhazian vtt, srt, ttml, srv3, srv2, srv1, json3
aa Afar vtt, srt, ttml, srv3, srv2, srv1, json3
af Afrikaans vtt, srt, ttml, srv3, srv2, srv1, json3
ak Akan vtt, srt, ttml, srv3, srv2, srv1, json3
sq Albanian vtt, srt, ttml, srv3, srv2, srv1, json3
am Amharic vtt, srt, ttml, srv3, srv2, srv1, json3
ar Arabic vtt, srt, ttml, srv3, srv2, srv1, json3
hy Armenian vtt, srt, ttml, srv3, srv2, srv1, json3
as Assamese vtt, srt, ttml, srv3, srv2, srv1, json3
ay Aymara vtt, srt, ttml, srv3, srv2, srv1, json3
az Azerbaijani vtt, srt, ttml, srv3, srv2, srv1, json3
bn Bangla vtt, srt, ttml, srv3, srv2, srv1, json3
ba Bashkir vtt, srt, ttml, srv3, srv2, srv1, json3
eu Basque vtt, srt, ttml, srv3, srv2, srv1, json3
be Belarusian vtt, srt, ttml, srv3, srv2, srv1, json3
bho Bhojpuri vtt, srt, ttml, srv3, srv2, srv1, json3
bs Bosnian vtt, srt, ttml, srv3, srv2, srv1, json3
br Breton vtt, srt, ttml, srv3, srv2, srv1, json3
bg Bulgarian vtt, srt, ttml, srv3, srv2, srv1, json3
my Burmese vtt, srt, ttml, srv3, srv2, srv1, json3
ca Catalan vtt, srt, ttml, srv3, srv2, srv1, json3
ceb Cebuano vtt, srt, ttml, srv3, srv2, srv1, json3
zh-Hans Chinese (Simplified) vtt, srt, ttml, srv3, srv2, srv1, json3
zh-Hant Chinese (Traditional) vtt, srt, ttml, srv3, srv2, srv1, json3
co Corsican vtt, srt, ttml, srv3, srv2, srv1, json3
hr Croatian vtt, srt, ttml, srv3, srv2, srv1, json3
cs Czech vtt, srt, ttml, srv3, srv2, srv1, json3
da Danish vtt, srt, ttml, srv3, srv2, srv1, json3
dv Divehi vtt, srt, ttml, srv3, srv2, srv1, json3
nl Dutch vtt, srt, ttml, srv3, srv2, srv1, json3
dz Dzongkha vtt, srt, ttml, srv3, srv2, srv1, json3
en-orig English (Original) vtt, srt, ttml, srv3, srv2, srv1, json3
en English vtt, srt, ttml, srv3, srv2, srv1, json3
eo Esperanto vtt, srt, ttml, srv3, srv2, srv1, json3
et Estonian vtt, srt, ttml, srv3, srv2, srv1, json3
… (123 lines truncated)
Exit code: 0
`| head -50` was parsed out (50 of 173 total lines shown).
<truncation_notice>
Full output written to: /var/folders/d9/mzhyr7lx6xbf236j5ll9v_300000gn/T/devin-overflows-501/4d8c0dc0/content.txt
</truncation_notice>English captions are available. Let me download them now.
{
"command": "cd /tmp && yt-dlp --skip-download --write-auto-sub --sub-lang en --sub-format srt -o \"yt_caption\" \"https://www.youtube.com/watch?v=ovrPPSZagPk\" 2>&1"
}Output from command in shell 6071e8:
WARNING: Your yt-dlp version (2025.12.08) is older than 90 days!
It is strongly recommended to always use the latest version.
You installed yt-dlp with pip or using the wheel from PyPi; Use that to update.
To suppress this warning, add --no-update to your command/config.
[youtube] Extracting URL: https://www.youtube.com/watch?v=ovrPPSZagPk
[youtube] ovrPPSZagPk: Downloading webpage
[youtube] ovrPPSZagPk: Downloading tv client config
[youtube] ovrPPSZagPk: Downloading player 4918c89a-main
[youtube] ovrPPSZagPk: Downloading tv player API JSON
[youtube] ovrPPSZagPk: Downloading android sdkless player API JSON
[youtube] [jsc:deno] Solving JS challenges using deno
WARNING: [youtube] [jsc] Remote components challenge solver script (deno) and NPM package (deno) were skipped. These may be required to solve JS challenges. You can enable these downloads with --remote-components ejs:github (recommended) or --remote-components ejs:npm , respectively. For more information and alternatives, refer to https://github.com/yt-dlp/yt-dlp/wiki/EJS
WARNING: [youtube] ovrPPSZagPk: n challenge solving failed: Some formats may be missing. Ensure you have a supported JavaScript runtime and challenge solver script distribution installed. Review any warnings presented before this message. For more details, refer to https://github.com/yt-dlp/yt-dlp/wiki/EJS
[info] ovrPPSZagPk: Downloading subtitles: en
[info] ovrPPSZagPk: Downloading 1 format(s): 139-19
[info] Writing video subtitles to: yt_caption.en.srt
[download] Destination: yt_caption.en.srt
[download] 1.00KiB at Unknown B/s (00:00:00)
[download] 3.00KiB at Unknown B/s (00:00:00)
[download] 7.00KiB at 4.92MiB/s (00:00:00)
[download] 13.73KiB at 3.20MiB/s (00:00:00)
[download] 100% of 13.73KiB in 00:00:00 at 136.61KiB/s
Exit code: 0{
"command": "cat /tmp/yt_caption.en.srt"
}Output from command in shell 686e12: 1 00:00:01,280 --> 00:00:05,920 The IM Cologne Major is over and instead 2 00:00:04,000 --> 00:00:07,520 of watching the games, I spent the whole 3 00:00:05,920 --> 00:00:09,280 time looking at what mice the pro 4 00:00:07,520 --> 00:00:11,759 players were using. With the release of 5 00:00:09,280 --> 00:00:14,480 the Logitech Gro X2 Super Strike and the 6 00:00:11,759 --> 00:00:16,480 Razer Viper V4 Pro, we are potentially 7 00:00:14,480 --> 00:00:19,039 looking at a massive shift in the Pro 8 00:00:16,480 --> 00:00:21,199 Scenes go-to mouse. Today, we're finding 9 00:00:19,039 --> 00:00:24,000 out which mouse is the most popular 10 00:00:21,199 --> 00:00:25,920 among CS2 pros, what mouse is 11 00:00:24,000 --> 00:00:27,760 statistically more likely to win you a 12 00:00:25,920 --> 00:00:29,920 major, whether sponsors are forcing 13 00:00:27,760 --> 00:00:31,920 players to use garbage, and other 14 00:00:29,920 --> 00:00:34,399 delightful fun facts you can bore people 15 00:00:31,920 --> 00:00:36,960 with at parties if you get invited to 16 00:00:34,399 --> 00:00:39,280 one, or a Discord group chat. The last 17 00:00:36,960 --> 00:00:42,320 event I looked at, the Austin Major, had 18 00:00:39,280 --> 00:00:44,480 a total of 10 brands. Two of those 10 19 00:00:42,320 --> 00:00:47,680 didn't make it this time around as ASUS 20 00:00:44,480 --> 00:00:50,160 ROG and Hitcan no longer feature. This 21 00:00:47,680 --> 00:00:52,559 time at the Cologne Major, we have 11 22 00:00:50,160 --> 00:00:54,960 brands in total with two new brands 23 00:00:52,559 --> 00:00:57,280 having their debut in my CS2 analysis 24 00:00:54,960 --> 00:00:59,680 videos and one returning brand as 25 00:00:57,280 --> 00:01:01,120 Endgame gear has made a comeback. The 26 00:00:59,680 --> 00:01:03,359 last time they were featured in my 27 00:01:01,120 --> 00:01:06,000 videos was during the very first CS2 28 00:01:03,359 --> 00:01:08,400 major in Copenhagen. This time there was 29 00:01:06,000 --> 00:01:10,640 only one player, Magix, representing the 30 00:01:08,400 --> 00:01:12,640 brand with the OP18K. 31 00:01:10,640 --> 00:01:14,400 We also see the inclusion of everyone's 32 00:01:12,640 --> 00:01:17,439 favorite budget mouse brand and its 33 00:01:14,400 --> 00:01:20,640 first appearance in my CS2 videos, ADK, 34 00:01:17,439 --> 00:01:22,799 with Zontics using the A9 Ultimate. The 35 00:01:20,640 --> 00:01:25,200 final new brand is one that you would 36 00:01:22,799 --> 00:01:27,680 not guess in a million years, unless 37 00:01:25,200 --> 00:01:29,520 you're Brazilian. Fallen is using a 38 00:01:27,680 --> 00:01:31,840 mouse from his own brand. It's 39 00:01:29,520 --> 00:01:33,920 potentially a Fallen Lobo wireless, 40 00:01:31,840 --> 00:01:36,320 which comes with a screen on the side. 41 00:01:33,920 --> 00:01:38,720 The Steel Series Rival 700's Legacy 42 00:01:36,320 --> 00:01:41,520 continues. 43 00:01:38,720 --> 00:01:44,720 This weighs around 60 g apparently. It 44 00:01:41,520 --> 00:01:46,720 has a 3395 sensor with 1,000 Hz polling. 45 00:01:44,720 --> 00:01:49,439 Looks like it has awful coating and 46 00:01:46,720 --> 00:01:51,840 costs around $88. You can also buy a 47 00:01:49,439 --> 00:01:54,240 Lamzu Maya X for almost $300 from his 48 00:01:51,840 --> 00:01:56,560 website if you want. There were two 49 00:01:54,240 --> 00:01:59,200 other brands that had just one user and 50 00:01:56,560 --> 00:02:01,119 appeared at the Austin Major. WL Mouse 51 00:01:59,200 --> 00:02:03,680 featured last time, but Slax has moved 52 00:02:01,119 --> 00:02:05,600 away from the Beast X and Sonic from NRG 53 00:02:03,680 --> 00:02:07,439 is using one instead. This is literally 54 00:02:05,600 --> 00:02:09,280 all I have to work with sometimes. 55 00:02:07,439 --> 00:02:11,920 Lambzu has dropped from having three 56 00:02:09,280 --> 00:02:15,040 users last year to just one at Cologne 57 00:02:11,920 --> 00:02:18,480 with Beast Stick using a Maya X. So, 58 00:02:15,040 --> 00:02:20,959 Endgame Gear, ATK, Fallen, Lambzu, and 59 00:02:18,480 --> 00:02:23,200 WL mouse all shared the eighth place 60 00:02:20,959 --> 00:02:25,200 spot. Final Mouse has gained two 61 00:02:23,200 --> 00:02:27,440 additional users from last year to be 62 00:02:25,200 --> 00:02:30,879 the sixth most popular mouse brand at 63 00:02:27,440 --> 00:02:33,280 the event. The WL mouse betrayer Slax is 64 00:02:30,879 --> 00:02:36,080 using the Ultralite X and is joined by 65 00:02:33,280 --> 00:02:38,480 major debutants, Symmetry, and Demon of 66 00:02:36,080 --> 00:02:40,080 Foot. The next three brands have dropped 67 00:02:38,480 --> 00:02:42,080 quite a few users from the previous 68 00:02:40,080 --> 00:02:44,160 event. The spreadsheet for this will be 69 00:02:42,080 --> 00:02:46,480 available for Patreon supporters on pay 70 00:02:44,160 --> 00:02:48,319 tiers, so please consider joining to 71 00:02:46,480 --> 00:02:50,080 support the channel. It includes some 72 00:02:48,319 --> 00:02:52,000 keyboards and mouse pads and a 73 00:02:50,080 --> 00:02:54,480 comparison of mice used by players at 74 00:02:52,000 --> 00:02:56,400 both the Austin and Cologne events if 75 00:02:54,480 --> 00:02:58,720 you want to see who swapped their mice. 76 00:02:56,400 --> 00:03:00,959 We also do giveaways for mice as well. 77 00:02:58,720 --> 00:03:03,360 Pulsar ranks as the fifth most popular 78 00:03:00,959 --> 00:03:05,599 brand, though their decline in usage is 79 00:03:03,360 --> 00:03:08,080 a bit disappointing. Last year, they had 80 00:03:05,599 --> 00:03:10,239 a really good uptake in users with 15 81 00:03:08,080 --> 00:03:13,200 players using their mice. The most 82 00:03:10,239 --> 00:03:15,360 popular being the XES. This year, they 83 00:03:13,200 --> 00:03:18,159 have dropped down to just six. 84 00:03:15,360 --> 00:03:20,720 Incredibly, there is just one XLite user 85 00:03:18,159 --> 00:03:24,239 this time around along with one X2 and 86 00:03:20,720 --> 00:03:27,360 one X2H. The remaining three pulsar mice 87 00:03:24,239 --> 00:03:31,599 are unbelievably Zywoo's the chosen 88 00:03:27,360 --> 00:03:34,239 mouse. Joining Zyu are OC and Bro. The 89 00:03:31,599 --> 00:03:36,640 fourth most popular brand is Vaxi, who 90 00:03:34,239 --> 00:03:39,360 also dropped users at this event from 14 91 00:03:36,640 --> 00:03:41,840 at the Austin Major to seven. Five 92 00:03:39,360 --> 00:03:44,159 players were using the Vaxi XE and one 93 00:03:41,840 --> 00:03:46,400 was using the MP01SV3 94 00:03:44,159 --> 00:03:48,000 and another was using an E1. There is 95 00:03:46,400 --> 00:03:50,080 likely a reason for this drop which 96 00:03:48,000 --> 00:03:52,080 we'll explore in a moment. The biggest 97 00:03:50,080 --> 00:03:54,560 surprise is that Razer has finished 98 00:03:52,080 --> 00:03:58,080 third with an 11.3% 99 00:03:54,560 --> 00:04:00,560 share, a decline from the 16.9% share 100 00:03:58,080 --> 00:04:02,959 they held during the Austin Major. 101 00:04:00,560 --> 00:04:05,040 Despite the release of the Viper V4 Pro, 102 00:04:02,959 --> 00:04:06,959 they still dropped users. When you 103 00:04:05,040 --> 00:04:09,439 consider the Valerant Pro scene adopted 104 00:04:06,959 --> 00:04:13,439 it rather quickly, helping Razer to a 105 00:04:09,439 --> 00:04:16,400 38.3% share at VCT Santiago, tying with 106 00:04:13,439 --> 00:04:18,880 Logitech, this was a surprising outcome. 107 00:04:16,400 --> 00:04:21,519 In Cologne, there were just four Viper 108 00:04:18,880 --> 00:04:24,560 V4 Pro users. There were three players 109 00:04:21,519 --> 00:04:27,840 still using the Viper V3 Pro. The Death 110 00:04:24,560 --> 00:04:30,320 V4 Pro had five users, leaving the most 111 00:04:27,840 --> 00:04:33,600 popular Razer mouse used at the Cologne 112 00:04:30,320 --> 00:04:35,919 Major being the Death Adder V3 Pro. This 113 00:04:33,600 --> 00:04:38,320 includes the Hypers Speed. A rather 114 00:04:35,919 --> 00:04:40,720 disappointing turnout by Razer standards 115 00:04:38,320 --> 00:04:43,120 this time around. The second most 116 00:04:40,720 --> 00:04:47,759 popular brand is Zawi, who went from 117 00:04:43,120 --> 00:04:50,639 18.1% last year to 24.4 4%. Their most 118 00:04:47,759 --> 00:04:54,479 popular mouse by a massive margin is the 119 00:04:50,639 --> 00:04:57,199 EC in any size with a 63.4% 120 00:04:54,479 --> 00:05:00,880 share, followed by the FK with six, 121 00:04:57,199 --> 00:05:03,440 three S2s, two U2s, two ZAS, an 122 00:05:00,880 --> 00:05:07,199 unreleased Donk mouse, and an unreleased 123 00:05:03,440 --> 00:05:09,280 Erggo mouse used by Blame F. Tabson was 124 00:05:07,199 --> 00:05:11,360 also possibly using this as well at some 125 00:05:09,280 --> 00:05:14,320 point, but I saw him using a Super 126 00:05:11,360 --> 00:05:16,240 Strike first. The model split might be a 127 00:05:14,320 --> 00:05:18,720 bit off, but I did notice something 128 00:05:16,240 --> 00:05:20,880 else. Previously, there's often been all 129 00:05:18,720 --> 00:05:22,400 Logitech or Razer teams. There's an all 130 00:05:20,880 --> 00:05:24,880 Logitech team, which we'll take a look 131 00:05:22,400 --> 00:05:27,520 at in a moment. This year, there was an 132 00:05:24,880 --> 00:05:30,479 all Zawi team. Gaming Gladiators might 133 00:05:27,520 --> 00:05:32,160 have all been using a Zawi EC. It's 134 00:05:30,479 --> 00:05:33,919 rather hard to confirm this because they 135 00:05:32,160 --> 00:05:36,000 got shot on, so there isn't much footage 136 00:05:33,919 --> 00:05:38,560 or photos of them, but they are all 137 00:05:36,000 --> 00:05:41,360 using a Zoey mouse of some sort. That 138 00:05:38,560 --> 00:05:44,960 leaves, unsurprisingly, Logitech as the 139 00:05:41,360 --> 00:05:49,280 most popular brand with… (8 chars truncated) … (229 lines truncated) Exit code: 0 <truncation_notice> Full output written to: /var/folders/d9/mzhyr7lx6xbf236j5ll9v_300000gn/T/devin-overflows-501/e12846e4/content.txt </truncation_notice>
{
"file_path": "/var/folders/d9/mzhyr7lx6xbf236j5ll9v_300000gn/T/devin-overflows-501/e12846e4/content.txt"
}<file-view path="/var/folders/d9/mzhyr7lx6xbf236j5ll9v_300000gn/T/devin-overflows-501/e12846e4/content.txt" start_line="1" end_line="788" total_lines="788"> 1|Output from command in shell 686e12: 2|1 3|00:00:01,280 --> 00:00:05,920 4|The IM Cologne Major is over and instead 5| 6|2 7|00:00:04,000 --> 00:00:07,520 8|of watching the games, I spent the whole 9| 10|3 11|00:00:05,920 --> 00:00:09,280 12|time looking at what mice the pro 13| 14|4 15|00:00:07,520 --> 00:00:11,759 16|players were using. With the release of 17| 18|5 19|00:00:09,280 --> 00:00:14,480 20|the Logitech Gro X2 Super Strike and the 21| 22|6 23|00:00:11,759 --> 00:00:16,480 24|Razer Viper V4 Pro, we are potentially 25| 26|7 27|00:00:14,480 --> 00:00:19,039 28|looking at a massive shift in the Pro 29| 30|8 31|00:00:16,480 --> 00:00:21,199 32|Scenes go-to mouse. Today, we're finding 33| 34|9 35|00:00:19,039 --> 00:00:24,000 36|out which mouse is the most popular 37| 38|10 39|00:00:21,199 --> 00:00:25,920 40|among CS2 pros, what mouse is 41| 42|11 43|00:00:24,000 --> 00:00:27,760 44|statistically more likely to win you a 45| 46|12 47|00:00:25,920 --> 00:00:29,920 48|major, whether sponsors are forcing 49| 50|13 51|00:00:27,760 --> 00:00:31,920 52|players to use garbage, and other 53| 54|14 55|00:00:29,920 --> 00:00:34,399 56|delightful fun facts you can bore people 57| 58|15 59|00:00:31,920 --> 00:00:36,960 60|with at parties if you get invited to 61| 62|16 63|00:00:34,399 --> 00:00:39,280 64|one, or a Discord group chat. The last 65| 66|17 67|00:00:36,960 --> 00:00:42,320 68|event I looked at, the Austin Major, had 69| 70|18 71|00:00:39,280 --> 00:00:44,480 72|a total of 10 brands. Two of those 10 73| 74|19 75|00:00:42,320 --> 00:00:47,680 76|didn't make it this time around as ASUS 77| 78|20 79|00:00:44,480 --> 00:00:50,160 80|ROG and Hitcan no longer feature. This 81| 82|21 83|00:00:47,680 --> 00:00:52,559 84|time at the Cologne Major, we have 11 85| 86|22 87|00:00:50,160 --> 00:00:54,960 88|brands in total with two new brands 89| 90|23 91|00:00:52,559 --> 00:00:57,280 92|having their debut in my CS2 analysis 93| 94|24 95|00:00:54,960 --> 00:00:59,680 96|videos and one returning brand as 97| 98|25 99|00:00:57,280 --> 00:01:01,120 100|Endgame gear has made a comeback. The 101| 102|26 103|00:00:59,680 --> 00:01:03,359 104|last time they were featured in my 105| 106|27 107|00:01:01,120 --> 00:01:06,000 108|videos was during the very first CS2 109| 110|28 111|00:01:03,359 --> 00:01:08,400 112|major in Copenhagen. This time there was 113| 114|29 115|00:01:06,000 --> 00:01:10,640 116|only one player, Magix, representing the 117| 118|30 119|00:01:08,400 --> 00:01:12,640 120|brand with the OP18K. 121| 122|31 123|00:01:10,640 --> 00:01:14,400 124|We also see the inclusion of everyone's 125| 126|32 127|00:01:12,640 --> 00:01:17,439 128|favorite budget mouse brand and its 129| 130|33 131|00:01:14,400 --> 00:01:20,640 132|first appearance in my CS2 videos, ADK, 133| 134|34 135|00:01:17,439 --> 00:01:22,799 136|with Zontics using the A9 Ultimate. The 137| 138|35 139|00:01:20,640 --> 00:01:25,200 140|final new brand is one that you would 141| 142|36 143|00:01:22,799 --> 00:01:27,680 144|not guess in a million years, unless 145| 146|37 147|00:01:25,200 --> 00:01:29,520 148|you're Brazilian. Fallen is using a 149| 150|38 151|00:01:27,680 --> 00:01:31,840 152|mouse from his own brand. It's 153| 154|39 155|00:01:29,520 --> 00:01:33,920 156|potentially a Fallen Lobo wireless, 157| 158|40 159|00:01:31,840 --> 00:01:36,320 160|which comes with a screen on the side. 161| 162|41 163|00:01:33,920 --> 00:01:38,720 164|The Steel Series Rival 700's Legacy 165| 166|42 167|00:01:36,320 --> 00:01:41,520 168|continues. 169| 170|43 171|00:01:38,720 --> 00:01:44,720 172|This weighs around 60 g apparently. It 173| 174|44 175|00:01:41,520 --> 00:01:46,720 176|has a 3395 sensor with 1,000 Hz polling. 177| 178|45 179|00:01:44,720 --> 00:01:49,439 180|Looks like it has awful coating and 181| 182|46 183|00:01:46,720 --> 00:01:51,840 184|costs around $88. You can also buy a 185| 186|47 187|00:01:49,439 --> 00:01:54,240 188|Lamzu Maya X for almost $300 from his 189| 190|48 191|00:01:51,840 --> 00:01:56,560 192|website if you want. There were two 193| 194|49 195|00:01:54,240 --> 00:01:59,200 196|other brands that had just one user and 197| 198|50 199|00:01:56,560 --> 00:02:01,119 200|appeared at the Austin Major. WL Mouse 201| 202|51 203|00:01:59,200 --> 00:02:03,680 204|featured last time, but Slax has moved 205| 206|52 207|00:02:01,119 --> 00:02:05,600 208|away from the Beast X and Sonic from NRG 209| 210|53 211|00:02:03,680 --> 00:02:07,439 212|is using one instead. This is literally 213| 214|54 215|00:02:05,600 --> 00:02:09,280 216|all I have to work with sometimes. 217| 218|55 219|00:02:07,439 --> 00:02:11,920 220|Lambzu has dropped from having three 221| 222|56 223|00:02:09,280 --> 00:02:15,040 224|users last year to just one at Cologne 225| 226|57 227|00:02:11,920 --> 00:02:18,480 228|with Beast Stick using a Maya X. So, 229| 230|58 231|00:02:15,040 --> 00:02:20,959 232|Endgame Gear, ATK, Fallen, Lambzu, and 233| 234|59 235|00:02:18,480 --> 00:02:23,200 236|WL mouse all shared the eighth place 237| 238|60 239|00:02:20,959 --> 00:02:25,200 240|spot. Final Mouse has gained two 241| 242|61 243|00:02:23,200 --> 00:02:27,440 244|additional users from last year to be 245| 246|62 247|00:02:25,200 --> 00:02:30,879 248|the sixth most popular mouse brand at 249| 250|63 251|00:02:27,440 --> 00:02:33,280 252|the event. The WL mouse betrayer Slax is 253| 254|64 255|00:02:30,879 --> 00:02:36,080 256|using the Ultralite X and is joined by 257| 258|65 259|00:02:33,280 --> 00:02:38,480 260|major debutants, Symmetry, and Demon of 261| 262|66 263|00:02:36,080 --> 00:02:40,080 264|Foot. The next three brands have dropped 265| 266|67 267|00:02:38,480 --> 00:02:42,080 268|quite a few users from the previous 269| 270|68 271|00:02:40,080 --> 00:02:44,160 272|event. The spreadsheet for this will be 273| 274|69 275|00:02:42,080 --> 00:02:46,480 276|available for Patreon supporters on pay 277| 278|70 279|00:02:44,160 --> 00:02:48,319 280|tiers, so please consider joining to 281| 282|71 283|00:02:46,480 --> 00:02:50,080 284|support the channel. It includes some 285| 286|72 287|00:02:48,319 --> 00:02:52,000 288|keyboards and mouse pads and a 289| 290|73 291|00:02:50,080 --> 00:02:54,480 292|comparison of mice used by players at 293| 294|74 295|00:02:52,000 --> 00:02:56,400 296|both the Austin and Cologne events if 297| 298|75 299|00:02:54,480 --> 00:02:58,720 300|you want to see who swapped their mice. 301| 302|76 303|00:02:56,400 --> 00:03:00,959 304|We also do giveaways for mice as well. 305| 306|77 307|00:02:58,720 --> 00:03:03,360 308|Pulsar ranks as the fifth most popular 309| 310|78 311|00:03:00,959 --> 00:03:05,599 312|brand, though their decline in usage is 313| 314|79 315|00:03:03,360 --> 00:03:08,080 316|a bit disappointing. Last year, they had 317| 318|80 319|00:03:05,599 --> 00:03:10,239 320|a really good uptake in users with 15 321| 322|81 323|00:03:08,080 --> 00:03:13,200 324|players using their mice. The most 325| 326|82 327|00:03:10,239 --> 00:03:15,360 328|popular being the XES. This year, they 329| 330|83 331|00:03:13,200 --> 00:03:18,159 332|have dropped down to just six. 333| 334|84 335|00:03:15,360 --> 00:03:20,720 336|Incredibly, there is just one XLite user 337| 338|85 339|00:03:18,159 --> 00:03:24,239 340|this time around along with one X2 and 341| 342|86 343|00:03:20,720 --> 00:03:27,360 344|one X2H. The remaining three pulsar mice 345| 346|87 347|00:03:24,239 --> 00:03:31,599 348|are unbelievably Zywoo's the chosen 349| 350|88 351|00:03:27,360 --> 00:03:34,239 352|mouse. Joining Zyu are OC and Bro. The 353| 354|89 355|00:03:31,599 --> 00:03:36,640 356|fourth most popular brand is Vaxi, who 357| 358|90 359|00:03:34,239 --> 00:03:39,360 360|also dropped users at this event from 14 361| 362|91 363|00:03:36,640 --> 00:03:41,840 364|at the Austin Major to seven. Five 365| 366|92 367|00:03:39,360 --> 00:03:44,159 368|players were using the Vaxi XE and one 369| 370|93 371|00:03:41,840 --> 00:03:46,400 372|was using the MP01SV3 373| 374|94 375|00:03:44,159 --> 00:03:48,000 376|and another was using an E1. There is 377| 378|95 379|00:03:46,400 --> 00:03:50,080 380|likely a reason for this drop which 381| 382|96 383|00:03:48,000 --> 00:03:52,080 384|we'll explore in a moment. The biggest 385| 386|97 387|00:03:50,080 --> 00:03:54,560 388|surprise is that Razer has finished 389| 390|98 391|00:03:52,080 --> 00:03:58,080 392|third with an 11.3% 393| 394|99 395|00:03:54,560 --> 00:04:00,560 396|share, a decline from the 16.9% share 397| 398|100 399|00:03:58,080 --> 00:04:02,959 400|they held during the Austin Major. 401| 402|101 403|00:04:00,560 --> 00:04:05,040 404|Despite the release of the Viper V4 Pro, 405| 406|102 407|00:04:02,959 --> 00:04:06,959 408|they still dropped users. When you 409| 410|103 411|00:04:05,040 --> 00:04:09,439 412|consider the Valerant Pro scene adopted 413| 414|104 415|00:04:06,959 --> 00:04:13,439 416|it rather quickly, helping Razer to a 417| 418|105 419|00:04:09,439 --> 00:04:16,400 420|38.3% share at VCT Santiago, tying with 421| 422|106 423|00:04:13,439 --> 00:04:18,880 424|Logitech, this was a surprising outcome. 425| 426|107 427|00:04:16,400 --> 00:04:21,519 428|In Cologne, there were just four Viper 429| 430|108 431|00:04:18,880 --> 00:04:24,560 432|V4 Pro users. There were three players 433| 434|109 435|00:04:21,519 --> 00:04:27,840 436|still using the Viper V3 Pro. The Death 437| 438|110 439|00:04:24,560 --> 00:04:30,320 440|V4 Pro had five users, leaving the most 441| 442|111 443|00:04:27,840 --> 00:04:33,600 444|popular Razer mouse used at the Cologne 445| 446|112 447|00:04:30,320 --> 00:04:35,919 448|Major being the Death Adder V3 Pro. This 449| 450|113 451|00:04:33,600 --> 00:04:38,320 452|includes the Hypers Speed. A rather 453| 454|114 455|00:04:35,919 --> 00:04:40,720 456|disappointing turnout by Razer standards 457| 458|115 459|00:04:38,320 --> 00:04:43,120 460|this time around. The second most 461| 462|116 463|00:04:40,720 --> 00:04:47,759 464|popular brand is Zawi, who went from 465| 466|117 467|00:04:43,120 --> 00:04:50,639 468|18.1% last year to 24.4 4%. Their most 469| 470|118 471|00:04:47,759 --> 00:04:54,479 472|popular mouse by a massive margin is the 473| 474|119 475|00:04:50,639 --> 00:04:57,199 476|EC in any size with a 63.4% 477| 478|120 479|00:04:54,479 --> 00:05:00,880 480|share, followed by the FK with six, 481| 482|121 483|00:04:57,199 --> 00:05:03,440 484|three S2s, two U2s, two ZAS, an 485| 486|122 487|00:05:00,880 --> 00:05:07,199 488|unreleased Donk mouse, and an unreleased 489| 490|123 491|00:05:03,440 --> 00:05:09,280 492|Erggo mouse used by Blame F. Tabson was 493| 494|124 495|00:05:07,199 --> 00:05:11,360 496|also possibly using this as well at some 497| 498|125 499|00:05:09,280 --> 00:05:14,320 500|point, but I saw him using a Super 501| 502|126 503|00:05:11,360 --> 00:05:16,240 504|Strike first. The model split might be a 505| 506|127 507|00:05:14,320 --> 00:05:18,720 508|bit off, but I did notice something 509| 510|128 511|00:05:16,240 --> 00:05:20,880 512|else. Previously, there's often been all 513| 514|129 515|00:05:18,720 --> 00:05:22,400 516|Logitech or Razer teams. There's an all 517| 518|130 519|00:05:20,880 --> 00:05:24,880 520|Logitech team, which we'll take a look 521| 522|131 523|00:05:22,400 --> 00:05:27,520 524|at in a moment. This year, there was an 525| 526|132 527|00:05:24,880 --> 00:05:30,479 528|all Zawi team. Gaming Gladiators might 529| 530|133 531|00:05:27,520 --> 00:05:32,160 532|have all been using a Zawi EC. It's 533| 534|134 535|00:05:30,479 --> 00:05:33,919 536|rather hard to confirm this because they 537| 538|135 539|00:05:32,160 --> 00:05:36,000 540|got shot on, so there isn't much footage 541| 542|136 543|00:05:33,919 --> 00:05:38,560 544|or photos of them, but they are all 545| 546|137 547|00:05:36,000 --> 00:05:41,360 548|using a Zoey mouse of some sort. That 549| 550|138 551|00:05:38,560 --> 00:05:44,960 552|leaves, unsurprisingly, Logitech as the 553| 554|139 555|00:05:41,360 --> 00:05:49,280 556|most popular brand with a 51.3% 557| 558|140 559|00:05:44,960 --> 00:05:52,639 560|share or 82 users, up from 42.5% 561| 562|141 563|00:05:49,280 --> 00:05:54,560 564|or 68 users from the Austin Major. After 565| 566|142 567|00:05:52,639 --> 00:05:56,639 568|what looked to be a downward trend, the 569| 570|143 571|00:05:54,560 --> 00:05:59,360 572|Super Strike seems to have reinvigorated 573| 574|144 575|00:05:56,639 --> 00:06:02,400 576|the protein with 31.3% 577| 578|145 579|00:05:59,360 --> 00:06:04,880 580|or 66 players using a Super Strike. 581| 582|146 583|00:06:02,400 --> 00:06:07,280 584|There were 30 Superite users. 585| 586|147 587|00:06:04,880 --> 00:06:09,759 588|Interestingly, despite Monzy having a 589| 590|148 591|00:06:07,280 --> 00:06:11,680 592|15-second ad for the Super Strike, he 593| 594|149 595|00:06:09,759 --> 00:06:14,240 596|doesn't actually use it. One team, 597| 598|150 599|00:06:11,680 --> 00:06:16,560 600|Na'vi, had all their players using a 601| 602|151 603|00:06:14,240 --> 00:06:19,440 604|Logitech mouse. There were two other 605| 606|152 607|00:06:16,560 --> 00:06:22,880 608|Logitech mice. Ins of Fly Quest was 609| 610|153 611|00:06:19,440 --> 00:06:26,080 612|actually using a Gro 2, and I'm around 613| 614|154 615|00:06:22,880 --> 00:06:28,720 616|60% sure that Biomass from Monte was 617| 618|155 619|00:06:26,080 --> 00:06:30,880 620|using a Superite 2 Dex. So, it's looking 621| 622|156 623|00:06:28,720 --> 00:06:35,120 624|like the Superite Dex might be a 625| 626|157 627|00:06:30,880 --> 00:06:37,199 628|Superite dud for the pro scene. 629| 630|158 631|00:06:35,120 --> 00:06:39,520 632|What's even more interesting is that in 633| 634|159 635|00:06:37,199 --> 00:06:42,639 636|the grand finals, there was just one 637| 638|160 639|00:06:39,520 --> 00:06:45,199 640|player using the Super Strike. In fact, 641| 642|161 643|00:06:42,639 --> 00:06:48,000 644|Razer was the most popular choice in the 645| 646|162 647|00:06:45,199 --> 00:06:51,120 648|grand finals, and three out of the five 649| 650|163 651|00:06:48,000 --> 00:06:53,759 652|players on Falcons were using Razer. So, 653| 654|164 655|00:06:51,120 --> 00:06:55,840 656|by this infallible logic, you're more 657| 658|165 659|00:06:53,759 --> 00:06:57,759 660|likely to win a CS2 major with the Razer 661| 662|166 663|00:06:55,840 --> 00:06:59,919 664|Mouse or if you're backed by a pro 665| 666|167 667|00:06:57,759 --> 00:07:01,680 668|state. I know some little rap scallion 669| 670|168 671|00:06:59,919 --> 00:07:03,599 672|is going to be furiously typing about 673| 674|169 675|00:07:01,680 --> 00:07:05,840 676|how they're sponsored, blah blah blah 677| 678|170 679|00:07:03,599 --> 00:07:08,000 680|blah. Give your fingers a rest, my 681| 682|171 683|00:07:05,840 --> 00:07:09,759 684|child. I've checked this as well. I went 685| 686|172 687|00:07:08,000 --> 00:07:11,759 688|through every team, checked their 689| 690|173 691|00:07:09,759 --> 00:07:15,280 692|sponsors, and I can confirm that 693| 694|174 695|00:07:11,759 --> 00:07:17,280 696|approximately 27 out of 160 players 697| 698|175 699|00:07:15,280 --> 00:07:19,599 700|could have been contractually obligated 701| 702|176 703|00:07:17,280 --> 00:07:22,639 704|to use a certain mouse. I went one 705| 706|177 707|00:07:19,599 --> 00:07:24,639 708|further. Nico, Zyu, and Monzy are 709| 710|178 711|00:07:22,639 --> 00:07:26,479 712|possibly contractually obligated to use 713| 714|179 715|00:07:24,639 --> 00:07:29,120 716|certain mice because they appear 717| 718|180 719|00:07:26,479 --> 00:07:32,400 720|directly in advertisements. So, that's 721| 722|181 723|00:07:29,120 --> 00:07:34,080 724|around 30 out of 160 players. I even 725| 726|182 727|00:07:32,400 --> 00:07:35,840 728|checked keyboards and that didn't seem 729| 730|183 731|00:07:34,080 --> 00:07:38,240 732|consistent either. If a team was 733| 734|184 735|00:07:35,840 --> 00:07:40,160 736|sponsored by Logitech or Razer, the 737| 738|185 739|00:07:38,240 --> 00:07:42,880 740|players would likely be using one of 741| 742|186 743|00:07:40,160 --> 00:07:44,960 744|their keyboards. Same with Asus ROG. 745| 746|187 747|00:07:42,880 --> 00:07:46,960 748|Mouse pads seem to be the same as mice, 749| 750|188 751|00:07:44,960 --> 00:07:49,199 752|but some players had their logos covered 753| 754|189 755|00:07:46,960 --> 00:07:51,280 756|up and others on the same team didn't. 757| 758|190 759|00:07:49,199 --> 00:07:54,240 760|Even then, across the 32 competing 761| 762|191 763|00:07:51,280 --> 00:07:56,400 764|teams, only 18 had partnerships with 765| 766|192 767|00:07:54,240 --> 00:07:58,240 768|companies that make peripherals. So, 769| 770|193 771|00:07:56,400 --> 00:08:00,000 772|pros are pros because they're very good 773| 774|194 775|00:07:58,240 --> 00:08:01,599 776|at the game. They don't need excuses 777| 778|195 779|00:08:00,000 --> 00:08:04,840 780|like it's the mouse's fault why they 781| 782|196 783|00:08:01,599 --> 00:08:04,840 784|didn't win. 785| 786| 787| 788|Exit code: 0 </file-view>
Here's a summary of the video:
The video is a data-driven breakdown of which gaming mice CS2 pro players used at the IEM Cologne Major, compared to the previous Austin Major.
| Rank | Brand | Share / Notes |
|---|---|---|
| 1st | Logitech | 51.3% (82 players) — up from 42.5% at Austin. The new G Pro X2 Super Strike drove adoption, with 66 players on it. |
| 2nd | Zowie | 24.4% — up from 18.1%. The EC series dominated with 63.4% share within the brand. Zowie even had one all-Zowie team (Gaming Gladiators). An unreleased "Donk mouse" and an unreleased ergo model were spotted. |
| 3rd | Razer | 11.3% — down from 16.9%, a surprise given the new Viper V4 Pro launch. The DeathAdder V3 Pro was their most-used mouse. Notably, Razer was the top brand in the grand finals (Falcons had 3/5 players on Razer). |
| 4th | Vaxee | Dropped from 14 to 7 users. |
| 5th | Pulsar | Dropped from 15 to 6 users. Notably, ZywOo uses their mouse. |
| 6th | Finalmouse | Gained 2 users vs. last year (Slax, Symmetry, Demon on the Ultralite X). |
| 7th–8th | Endgame Gear, ATK, Fallen (brand), Lambzu, WL Mouse | 1 user each |
Logitech has a dominant and growing grip on the CS2 pro scene, with the G Pro X2 Super Strike being the new go-to. Zowie is a strong #2. The Razer Viper V4 Pro underperformed expectations at this event despite doing well in Valorant pro play. Sponsor pressure is real but not the dominant factor in mouse choice.