Sessions

YouTube Video Discussion

adaptivebypass~Jul 1, 2026, 12:08 AM UTC
In 7Out 1,287Cache 89,819Time 30.8s
6 system messages
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>
  1. You
  2. Devin

    I'll use yt-dlp to get the auto-generated captions for this YouTube video and summarize it.

    exec(command: yt-dlp --list-subs "https://www.youtube.…)
    {
      "command": "yt-dlp --list-subs \"https://www.youtube.com/watch?v=mpSYWXRYJdE\" 2>&1 | head -50"
    }
    In 3Out 119Cache 8,988TTFT 1,332mstool_calls
  3. OutputOutput from command in shell ab98bc:
    Output from command in shell ab98bc:
    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=mpSYWXRYJdE
    [youtube] mpSYWXRYJdE: Downloading webpage
    [youtube] mpSYWXRYJdE: Downloading tv client config
    [youtube] mpSYWXRYJdE: Downloading player 4918c89a-main
    [youtube] mpSYWXRYJdE: Downloading tv player API JSON
    [youtube] mpSYWXRYJdE: 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] mpSYWXRYJdE: 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 mpSYWXRYJdE:
    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/f4ce9463/content.txt
    </truncation_notice>
  4. Devin

    Good, English auto-generated captions are available. Let me download them now.

    exec(command: cd /tmp && yt-dlp --write-auto-sub --sub…)
    {
      "command": "cd /tmp && yt-dlp --write-auto-sub --sub-lang en --skip-download --sub-format vtt -o \"yt_caption\" \"https://www.youtube.com/watch?v=mpSYWXRYJdE\" 2>&1"
    }
    In 1Out 134Cache 17,945TTFT 1,453mstool_calls
  5. OutputOutput from command in shell 0b99f0:
    Output from command in shell 0b99f0:
    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=mpSYWXRYJdE
    [youtube] mpSYWXRYJdE: Downloading webpage
    [youtube] mpSYWXRYJdE: Downloading tv client config
    [youtube] mpSYWXRYJdE: Downloading player 4918c89a-main
    [youtube] mpSYWXRYJdE: Downloading tv player API JSON
    [youtube] mpSYWXRYJdE: 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] mpSYWXRYJdE: 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] mpSYWXRYJdE: Downloading subtitles: en
    [info] mpSYWXRYJdE: Downloading 1 format(s): 139-20
    [info] Writing video subtitles to: yt_caption.en.vtt
    [download] Destination: yt_caption.en.vtt
    
    [download]    1.00KiB at  Unknown B/s (00:00:00)
    [download]    3.00KiB at  Unknown B/s (00:00:00)
    [download]    7.00KiB at  Unknown B/s (00:00:00)
    [download]   15.00KiB at    5.11MiB/s (00:00:00)
    [download]   31.00KiB at    1.75MiB/s (00:00:00)
    [download]   63.00KiB at    2.42MiB/s (00:00:00)
    [download]   84.75KiB at    2.80MiB/s (00:00:00)
    [download] 100% of   84.75KiB in 00:00:00 at 690.63KiB/s
    
    
    Exit code: 0
  6. Devin
    exec(command: cat /tmp/yt_caption.en.vtt | sed '/^[0-9…)
    {
      "command": "cat /tmp/yt_caption.en.vtt | sed '/^[0-9][0-9]:[0-9][0-9]/d' | sed '/^$/d' | sed 's/<[^>]*>//g' | grep -v '^WEBVTT' | grep -v '^Kind:' | grep -v '^Language:' | sort -u | head -200"
    }
    In 1Out 143Cache 19,603tool_calls
  7. OutputOutput from command in shell 2299b5:
    Output from command in shell 2299b5:
     
    "Yeah, this just isn't it's not a great
    $70 allin. Um, scroll wheel, it's fine.
    a keyboard. I got to talk about
    a Logitech hater. I just think the A9
    a rhetorical question. Actually, the
    a thing where I don't think like many 2C
    accidental scrolls. So, this is a pretty
    actually a lot of like post travel on
    actually be able to provide some type of
    advertise as 39 plus - 3 g. So, they are
    all around like the mouse is very
    all. Considering the fact it's a $60
    and this is going to be a video
    And when I compare it to Logitech's
    any features I didn't cover, let me know
    as it is on the standard GPX. I will say
    as well. They say 800 hours of battery
    at a premium price, I'd be like, okay,
    ATK's lying about like or just
    back to mice. The ATK A9 Mini Plus is a
    be polling rate, click latency, but if
    best budget mouse? Um, no. There are
    best of the year so far, at least has
    bored of clones and seeing the same
    bouncier, more spammable feeling, ATK
    build quality, like when I squeeze
    build quality, weight, sensor
    bulky, too filling, too high profile.
    but actually I'm just putting pressure
    But before you invest in a mouse, ask
    but that was a very different mouse
    buttons. Um, but everything else like
    by default. So if you were to throw on
    can provide a more stable connection.
    can say about this that would cause
    catalog of their products that I haven't
    cherry on top. And speaking of the web
    come across my desk. I'm not testing
    companies are, but some people take
    compared to 47 on the 2C. um with these
    compared to like the $80 um G305 Super
    comparing them up close. It's like if
    completely passable when you consider
    connection. Are you missing the forest
    cool, but this implementation it just
    could say like, "Yeah, the 2C at least
    Day, it was like 105. It's just a very
    decent amount of vertical room on the
    discussing the ATK A9 Mini Plus, one of
    does a messy job of it. Like for me,
    does come with very large stock skates
    doesn't make as much sense when you can
    dot skates and the power play removed
    dots, you could probably get it um even
    effectively going to be the same shape.
    either. 3955 Nordic 54 L-15 chip. And
    especially if you're somebody who has
    especially low quality or really
    every cheap Chinese mouse out there, but
    exactly the same shape shell experience.
    example where ATK kind of brought the 2C
    experienced high ping, packet loss,
    extraordinary. Um ATK, it's just
    feel the plastic flex inwards a little,
    feeling raw input. Yeah, it's just it
    feeling, I honest I don't have a strong
    feels like a very average Omron optical
    for anybody who's still watching, I'm
    for like future implementations or a
    for the monthly subscription. You can
    for the trees? Possibly. And this is not
    Fortnite. I actually have paid for exit
    fractions of a millisecond, whether it
    from exit lag of course. Mice,
    from like the stiffer clicks and like a
    frontheavy um but it's not as noticeable
    get free Discord Nitro if you use the
    get something like this dual 8K web
    get the annual plan for 50% off and you
    get you there. So, yeah, while you can
    getting one of these mice, but just
    going to be suspicious of my ATK
    good, not great feel along with the side
    GPX normal shape to just be like too
    GPX than pretty much anything else on
    great issue with it. Like their customer
    hand, it's really not something to like
    hard mouse to recommend. ATK does
    hard to imagine somebody who wants a
    have an issue with. There is kind of
    having this in hand using it, it's not
    hold anything against the mouse because
    honestly, if there was a 2C Super Strike
    I'm just testing these back to back,
    I'm not going to say it will bring your
    if you guys have any other questions,
    implementation, but it's nothing like
    implementation, the battery is very good
    in the comments below. But yeah, the A9
    into play. They're a connection
    is it's a solid lightweight Police.
    is something you struggle with playing
    is substantiating. It's not going to be
    it has this going for it." Um, aside
    it like hoping there was something I
    it. Um, even if it did, maybe I would
    it's a solid budget release. one of the
    It's like I don't know what to say. It's
    It's like it I kind of see it as like,
    It's very easy to recommend. I don't
    it's Yeah, there's really not anything I
    joint. So yeah, this is kind of just the
    just don't let me like build up a
    keyboards in 2026. I've just been using
    keyboards, doesn't matter. Exit lag all
    Kovac, there's just such a minimal
    lag before, but I was a rube and I paid
    level of ATK experience that goes into
    life. I don't know like how that claim
    Light, $80 W mouse Beast G, I mean, this
    lighter by a gram or two. Um, as far as
    like 1v oneing in Fortnite or playing
    like a bought and paid for review aides
    like a hollow feel to them but um not
    like cheaped out on the sensor or MCU
    like Chinese mouse click implementation
    like small to medium hands. You find the
    like the 2C had the seal of approval,
    like the design. There is a kind of a
    link in the description QR code on the
    looking for something like lighter,
    make a pretty big difference. So, I
    market. Um, yeah, this is definitely
    me that I haven't reviewed the Hex 80.
    Mini can definitely get the budget seal
    mini GPX getting this and feeling like
    Mini Plus is a much more sensible
    money. Um, I will say like the clicks,
    mouse is a bit too small for me compared
    mouse with two actual sensors that will
    mouse, it really makes the 2C at 150
    my wooting all year. Um, and yeah,
    not going to lie. Sometimes I'm like,
    not send me stuff. So yeah, this is not
    notice it feeling slightly larger in
    noticed that though with the way I press
    obsess over or worry about. It is
    of actual experience, it's not like they
    of approval. I really probably I mean
    of side to side wobble. Like this just
    okay, it's just kind of how Chinese
    on central living in New York on
    opticals, which definitely have like a
    optimizer for gaming. If you've
    overexaggerating battery life,
    performance difference. And I went into
    ping from 60 to zero. It really is
    plenty of options, even like other
    position which on paper it sounds very
    preference either way. But if you are
    pretty much all dimensions, but it's not
    price they do provide pretty much
    purchase if you're looking for a mini
    quality like there definitely is a bit
    Realistically, I probably wouldn't use
    really just phoning it in." But on that
    really misses the boat. So, I'm waiting
    really will offer that. And this mouse
    recommend checking them out. Yeah. Now,
    reviewed. Um, they're still upset with
    reviewer Dastrazac. And yeah, ATK, they
    routing issues, disconnections, uh it
    screen. I mean, so many of us care about
    sent this unit. Um, this is from fellow
    servers, something like exit lag can
    service isn't good. There's definitely a
    shape um to more people because really
    shapes from ATK, MHose. Definitely a
    sharper, much more firm, like tighter
    shell clone of the GPro X Superite 2C.
    shells over and over, but this is a good
    shilling. I will say I didn't even get
    side buttons. I like the positioning.
    simple release that I rate highly. It's
    slightly larger than the 2C um, in
    slightly more narrow um, grip width,
    software um just solid functionality
    software, there is virtual sensor
    solid clone release. I know I'm very
    somebody to buy it over the A9 Mini. So,
    somebody who prefers like a raw um input
    something I can recommend if connection
    something I really complain much about.
    specifically at this center point, I can
    spending well over $100 on this. It just
    sponsor of this video, Exit Lag, comes
    state of mice in 2026. This is a really
    style. Really nothing wrong with this at
    switches. It has a good sensor
    … (45 lines truncated)
    
    Exit code: 0
    
    `| head -200` was parsed out (200 of 245 total lines shown).
    <truncation_notice>
    Full output written to: /var/folders/d9/mzhyr7lx6xbf236j5ll9v_300000gn/T/devin-overflows-501/27f085c3/content.txt
    </truncation_notice>
  8. Devin
    exec(command: # Get a cleaner sequential transcript - …)
    {
      "command": "# Get a cleaner sequential transcript - strip VTT formatting, deduplicate adjacent lines\npython3 -c \"\nimport re\n\nwith open('/tmp/yt_caption.en.vtt', 'r') as f:\n    content = f.read()\n\n# Remove VTT header and timestamp lines\nlines = content.split('\\n')\ntext_lines = []\nfor line in lines:\n    line = line.strip()\n    if not line:\n        continue\n    if line.startswith('WEBVTT') or line.startswith('Kind:') or line.startswith('Language:'):\n        continue\n    if re.match(r'^\\d{2}:\\d{2}', line):  # timestamp line\n        continue\n    # Remove HTML tags\n    line = re.sub(r'<[^>]+>', '', line)\n    if line:\n        text_lines.append(line)\n\n# Deduplicate consecutive identical lines\ndeduped = []\nprev = None\nfor line in text_lines:\n    if line != prev:\n        deduped.append(line)\n        prev = line\n\nprint('\\n'.join(deduped))\n\" 2>&1"
    }
    In 1Out 327Cache 20,479tool_calls
  9. OutputOutput from command in shell 1966c7:
    Output from command in shell 1966c7:
    Yo, what's good YouTube? It's Sportsy
    and this is going to be a video
    discussing the ATK A9 Mini Plus, one of
    the best budget deals in gaming mice.
    But before you invest in a mouse, ask
    yourself if you have invested in your
    connection. Are you missing the forest
    for the trees? Possibly. And this is not
    a rhetorical question. Actually, the
    sponsor of this video, Exit Lag, comes
    into play. They're a connection
    optimizer for gaming. If you've
    experienced high ping, packet loss,
    routing issues, disconnections, uh it
    can provide a more stable connection.
    I'm not going to say it will bring your
    ping from 60 to zero. It really is
    something I can recommend if connection
    is something you struggle with playing
    on central living in New York on
    Fortnite. I actually have paid for exit
    lag before, but I was a rube and I paid
    for the monthly subscription. You can
    get the annual plan for 50% off and you
    get free Discord Nitro if you use the
    link in the description QR code on the
    screen. I mean, so many of us care about
    fractions of a millisecond, whether it
    be polling rate, click latency, but if
    you're playing on high ping different
    servers, something like exit lag can
    make a pretty big difference. So, I
    recommend checking them out. Yeah. Now,
    back to mice. The ATK A9 Mini Plus is a
    shell clone of the GPro X Superite 2C.
    Um, it comes in on my skill at 45 g
    compared to 47 on the 2C. um with these
    dot skates and the power play removed
    which actually does make it a bit more
    frontheavy um but it's not as noticeable
    as it is on the standard GPX. I will say
    though ATK for a significantly lower
    price they do provide pretty much
    exactly the same shape shell experience.
    You get a solid coating. Um the click
    quality like there definitely is a bit
    of side to side wobble. Like this just
    feels like a very average Omron optical
    like Chinese mouse click implementation
    which is fine. Like it performs fine.
    And when I compare it to Logitech's
    opticals, which definitely have like a
    sharper, much more firm, like tighter
    feeling, I honest I don't have a strong
    preference either way. But if you are
    looking for something like lighter,
    bouncier, more spammable feeling, ATK
    really will offer that. And this mouse
    does come with very large stock skates
    by default. So if you were to throw on
    dots, you could probably get it um even
    lighter by a gram or two. Um, as far as
    build quality, like when I squeeze
    specifically at this center point, I can
    feel the plastic flex inwards a little,
    but actually I'm just putting pressure
    with like a normal fingertip claw grip
    style. Really nothing wrong with this at
    all. Considering the fact it's a $60
    mouse, it really makes the 2C at 150
    well over double the cost. Even on Prime
    Day, it was like 105. It's just a very
    hard mouse to recommend. ATK does
    advertise as 39 plus - 3 g. So, they are
    very off on the weight. Um, but in terms
    of actual experience, it's not like they
    like cheaped out on the sensor or MCU
    either. 3955 Nordic 54 L-15 chip. And
    all around like the mouse is very
    slightly larger than the 2C um, in
    pretty much all dimensions, but it's not
    a thing where I don't think like many 2C
    users are going to be buying this and
    comparing them up close. It's like if
    you want the mini GPX experience, um,
    then the ATK A9 Mini Plus is going to
    get you there. So, yeah, while you can
    notice it feeling slightly larger in
    hand, it's really not something to like
    obsess over or worry about. It is
    effectively going to be the same shape.
    Um, side button-wise, they also copied
    like the design. There is a kind of a
    decent amount of vertical room on the
    side buttons. I like the positioning.
    The feeling is really fine. and the
    mouse is a bit too small for me compared
    to the Super Strike um to comfortably
    use in Fortnite. But the side buttons
    then the positioning I really didn't
    have an issue with. There is kind of
    like a hollow feel to them but um not
    especially low quality or really
    unusable in any way. I guess there is
    actually a lot of like post travel on
    this bottom button. I never really
    noticed that though with the way I press
    them with like this part of my thumb
    joint. So yeah, this is kind of just the
    state of mice in 2026. This is a really
    solid clone release. I know I'm very
    bored of clones and seeing the same
    shells over and over, but this is a good
    example where ATK kind of brought the 2C
    shape um to more people because really
    spending well over $100 on this. It just
    doesn't make as much sense when you can
    get something like this dual 8K web
    software um just solid functionality
    weight for so much cheaper. And
    honestly, if there was a 2C Super Strike
    at a premium price, I'd be like, okay,
    you know, there's an argument to be made
    there at least. But this has the optical
    switches. It has a good sensor
    implementation, but it's nothing like
    extraordinary. Um ATK, it's just
    completely passable when you consider
    the price. It just I mean that's the
    cherry on top. And speaking of the web
    software, there is virtual sensor
    position which on paper it sounds very
    cool, but this implementation it just
    does a messy job of it. Like for me,
    somebody who prefers like a raw um input
    feeling raw input. Yeah, it's just it
    really misses the boat. So, I'm waiting
    for like future implementations or a
    mouse with two actual sensors that will
    actually be able to provide some type of
    virtual sensor position similar to what
    the Gwolves V did. Yeah, I don't really
    hold anything against the mouse because
    that feature doesn't work properly.
    Realistically, I probably wouldn't use
    it. Um, even if it did, maybe I would
    though. And yeah, I know some people are
    going to be suspicious of my ATK
    shilling. I will say I didn't even get
    sent this unit. Um, this is from fellow
    reviewer Dastrazac. And yeah, ATK, they
    just don't let me like build up a
    catalog of their products that I haven't
    reviewed. Um, they're still upset with
    me that I haven't reviewed the Hex 80.
    It's like I don't know what to say. It's
    a keyboard. I got to talk about
    keyboards in 2026. I've just been using
    my wooting all year. Um, and yeah,
    there's nothing like nothing really that
    this provides and it's causing ATK to
    not send me stuff. So yeah, this is not
    like a bought and paid for review aides
    from exit lag of course. Mice,
    keyboards, doesn't matter. Exit lag all
    that matters. Um, and yeah, clearly not
    a Logitech hater. I just think the A9
    Mini Plus is a much more sensible
    purchase if you're looking for a mini
    GPX than pretty much anything else on
    the market. Is it bo like bar none the
    best budget mouse? Um, no. There are
    plenty of options, even like other
    shapes from ATK, MHose. Definitely a
    very easy product to recommend,
    especially if you're somebody who has
    like small to medium hands. You find the
    GPX normal shape to just be like too
    bulky, too filling, too high profile.
    It's very easy to recommend. I don't
    think you can go wrong for like $60 to
    $70 allin. Um, scroll wheel, it's fine.
    They have a b hop mode equivalent as
    well. Um, so there won't be any like
    accidental scrolls. So, this is a pretty
    simple release that I rate highly. It's
    hard to imagine somebody who wants a
    mini GPX getting this and feeling like
    they've really been cheated out of their
    money. Um, I will say like the clicks,
    they do just have like this kind of
    good, not great feel along with the side
    buttons. Um, but everything else like
    build quality, weight, sensor
    implementation, the battery is very good
    as well. They say 800 hours of battery
    life. I don't know like how that claim
    is substantiating. It's not going to be
    that long. Yeah, it's a good segue. Um,
    for anybody who's still watching, I'm
    not going to lie. Sometimes I'm like,
    "Yeah, this just isn't it's not a great
    video. This is not my best work. I'm
    really just phoning it in." But on that
    topic, what do you guys think about
    ATK's lying about like or just
    overexaggerating battery life,
    underelling the weight on their mice?
    It's like it I kind of see it as like,
    okay, it's just kind of how Chinese
    companies are, but some people take
    great issue with it. Like their customer
    service isn't good. There's definitely a
    level of ATK experience that goes into
    getting one of these mice, but just
    having this in hand using it, it's not
    something I really complain much about.
    I'm just testing these back to back,
    like 1v oneing in Fortnite or playing
    Kovac, there's just such a minimal
    performance difference. And I went into
    it like hoping there was something I
    could say like, "Yeah, the 2C at least
    it has this going for it." Um, aside
    from like the stiffer clicks and like a
    slightly more narrow um, grip width,
    it's Yeah, there's really not anything I
    can say about this that would cause
    somebody to buy it over the A9 Mini. So,
    if you guys have any other questions,
    any features I didn't cover, let me know
    in the comments below. But yeah, the A9
    Mini can definitely get the budget seal
    of approval. I really probably I mean
    like the 2C had the seal of approval,
    but that was a very different mouse
    market. Um, yeah, this is definitely
    it's a solid budget release. one of the
    best of the year so far, at least has
    come across my desk. I'm not testing
    every cheap Chinese mouse out there, but
    compared to like the $80 um G305 Super
    Light, $80 W mouse Beast G, I mean, this
    is it's a solid lightweight Police.
    
    
    Exit code: 0
  10. Devin

    Here's a summary of the video:


    ATK A9 Mini Plus — Gaming Mouse Review

    Creator: Sportsy (YouTube)

    What it is

    The ATK A9 Mini Plus is a shell clone of the Logitech GPro X Superlight 2C, priced at around $60–$70 all-in. It weighs 45g (vs. 47g on the 2C), uses a 3955 sensor and Nordic 54 / L-15 chip, and comes with optical switches and large stock skates.

    The Good

    • Shape: Near-identical to the GPro X Superlight 2C — great for small-to-medium hands who find the standard GPX too bulky/high-profile
    • Value: At half the price (or less) of the 2C (~$150, even ~$105 on Prime Day), it's a much easier recommendation
    • Build quality: Acceptable for the price — minor plastic flex under pressure but nothing to worry about
    • Battery: Claimed 800 hours (reviewer is skeptical of the exact figure, but it's good)
    • Software: Solid, with virtual sensor positioning (though the implementation is currently rough)
    • Scroll wheel: Has a "b-hop mode" equivalent to prevent accidental scrolls

    The Meh

    • Clicks: Average Omron optical feel — a bit looser/bouncier vs. Logitech's tighter, firmer feel
    • Side buttons: Some post-travel on the bottom button; slightly hollow feel, but functional
    • Virtual sensor position: Sounds cool on paper but doesn't work well in practice
    • Weight claim: ATK advertises 39g but it actually weighs ~45g — a common complaint about ATK overpromising

    The Verdict

    Budget seal of approval. It's one of the best budget mice of 2026 so far. If you want the mini GPX experience without paying GPX prices, this is the obvious pick. It's hard to find a reason to buy the actual 2C over this at $60–70. Not the absolute best budget mouse (there are other shapes from ATK and Mhose to consider), but very easy to recommend for mini-GPX fans.

    Sponsor note: ExitLag (connection optimizer for gaming) is the video's sponsor — 50% off annual plan with Discord Nitro included via description link.

    In 1Out 564Cache 22,804TTFT 1,321ms