llm/9ad11e16-7acb-4923-bb7e-5d14cd36cf3f/batch-1-5791817a-fd61-4246-ac6a-1ed7bbbe7908-input.json
The following is content for you to classify. Do not respond to the comments—classify them.
<topics>
1. Novelty vs. Rebranding Old Tech
Related: Commenters frequently note that SSHing into a remote machine is a standard practice that has existed for decades, comparing the article's framing to the infamous 'Dropbox comment' on Hacker News. Users express confusion over why a basic remote administration workflow is being presented as a new concept called 'Doom Coding,' while others argue that the integration of LLM agents like Claude Code provides a fresh layer of utility to the established setup.
2. Mobile Ergonomics and Friction
Related: A major theme is the physical difficulty of coding on a smartphone. Commenters discuss the pain of typing on touchscreens, the inability to view side-by-side diffs effectively on small screens, and the general clumsiness of managing terminal windows without a physical keyboard. Many users argue that while the setup is technically possible, the lack of screen real estate and input precision makes it impractical for serious engineering work.
3. Chat-Based and Async Interfaces
Related: To bypass the limitations of mobile terminal UIs, users suggest and share workflows that use chat interfaces like Telegram, Email, or WhatsApp to interact with coding agents. These setups allow users to send prompts and approve pull requests via natural language messages or buttons, treating the coding process as an asynchronous conversation rather than a real-time terminal session, which resolves many formatting and typing issues.
4. Session Persistence with Tmux
Related: Technical advice heavily features `tmux` (terminal multiplexer) as an essential tool for this workflow. Commenters explain that standard SSH connections drop when a mobile device sleeps or switches networks, whereas `tmux` allows the session to persist on the host machine. This enables users to seamlessly resume their work exactly where they left off, regardless of connectivity interruptions or app switching.
5. Alternative Mobile Environments
Related: Users discuss various apps and environments that serve as alternatives to the article's Termius setup. Android users advocate for Termux, which provides a local Linux environment without needing a remote host, while iOS users recommend Blink Shell or Shellfish for better Mosh and SSH integration. Some also mention using native Pixel terminal features or running local LLMs directly on the device to avoid latency and dependency on internet access.
6. Mental Health and Downtime
Related: The concept of 'Doom Coding' sparks a philosophical debate about work-life balance. Commenters question the healthiness of filling every moment of downtime with productivity, suggesting that time spent waiting or commuting might be better used for rest, observation, or 'micro-exercises.' There is criticism of the compulsion to code constantly, with some arguing that being present in social situations or relaxing is more valuable than 'vibe coding' on a phone.
7. Reliability and Network Latency
Related: The discussion highlights the technical challenges of mobile connectivity, specifically latency and dropped connections. Tools like Mosh (Mobile Shell) are frequently recommended over standard SSH because they are designed to handle intermittent networks and roaming between Wi-Fi and cellular data without killing the session. Tailscale is praised for simplifying the networking layer, though some prefer Wireguard for a lighter-weight alternative.
8. Wake-on-LAN and Power Usage
Related: Critiques of the requirement to leave a computer running 24/7 lead to discussions on energy efficiency and remote wake capabilities. Users share solutions using Wake-on-LAN (WOL) via routers or Raspberry Pis to turn on powerful machines only when needed. Others mention macOS settings like `caffeinate` or specific power configurations to ensure the host machine remains accessible without wasting electricity around the clock.
9. Verification and Code Review
Related: A critical point raised is the danger of 'fire-and-forget' coding with LLMs on mobile. Users note that verifying the code generated by AI is difficult on a phone due to limited visibility and syntax highlighting. The conversation touches on the risks of deploying code or merging pull requests without the ability to properly audit the logic or run tests, suggesting that mobile workflows are better suited for prototyping than production engineering.
10. Security Risks of Remote Access
Related: Several commenters express concern over the security implications of the proposed setup. Issues include leaving a computer unlocked at home, opening SSH ports (even via VPN), and the potential for bad actors to gain access to a local network. Discussions involve best practices such as using key-based authentication, locking the keychain via command line, and the general risks of exposing a development machine to remote connections.
11. Vibe Coding vs. Engineering
Related: There is a distinction drawn between 'vibe coding'—prompting an LLM to generate scripts or apps—and traditional software engineering. Some users view this workflow as 'slop' or merely prompting, lacking the depth of actual problem-solving, while others find it empowering for quick prototypes or hobby projects. This theme reflects broader tensions regarding the changing nature of software development in the age of generative AI.
12. Web-Based and Cloud Alternatives
Related: Users suggest that browser-based solutions like GitHub Codespaces, Replit, or self-hosted web IDEs (like `opencode`) offer a superior experience to terminal tunneling. These tools often provide better UI elements for mobile browsers and abstract away the need to manage hardware or VPNs, allowing users to code via a web interface that handles state management and environments automatically.
13. Social Acceptability
Related: The humorous notion of 'coding at the club' mentioned in the article draws specific reactions. Commenters joke about or criticize the anti-social nature of pulling out a phone to code in social settings like parties or bars. This overlaps with the 'doom scrolling' comparison, with some users suggesting that using a phone for work in social spaces is just as rude or 'gross' as using it for social media.
14. Hardware Workarounds
Related: To mitigate the interface limitations of smartphones, users discuss hardware additions such as folding phones, external Bluetooth keyboards, and 'thin client' setups using old laptops or tablets. Some mention specific devices like the 'Clicks' keyboard case or using AR glasses, highlighting that while the phone provides the compute or connection, better peripherals are often needed for actual productivity.
15. Agentic Workflows and Automation
Related: Discussions extend beyond simple coding to fully automated agentic workflows. Users describe setups where agents running on home servers are triggered via mobile to perform tasks, run scans, or manage infrastructure autonomously. This shifts the mobile interaction from writing code to orchestrating agents that perform the heavy lifting, reporting back status via push notifications or chat messages.
16. Subscription Fatigue vs. Open Source
Related: Comments reflect a wariness of paid subscriptions for tools like Claude Pro, Tailscale, or premium terminal apps. Users advocate for open-source alternatives such as Ollama for local LLMs, Wireguard for VPNs, and various free terminal emulators. There is a sentiment that basic remote coding shouldn't require a stack of monthly fees when free tools can achieve similar results with slightly more configuration.
17. Educational Value of LLMs
Related: Some users highlight the learning aspect of this workflow, noting that interacting with Claude Code allows them to understand new concepts (like API behaviors or network scanning) through the generative act. This counters the 'slop' narrative, suggesting that 'doom coding' can be a valid educational tool for hobbyists or those looking to understand how their devices and networks function.
18. Thin Client Philosophy
Related: The thread revisits the concept of thin clients, where the mobile device acts merely as a window into a powerful remote machine. Users compare this to historical mainframe/terminal workflows, arguing that the phone doesn't need to be powerful if it just renders text from a desktop. This philosophy underpins the preference for SSH/Mosh over running heavy local IDEs on the phone.
19. Code Quality and Maintenance
Related: Skepticism arises regarding the quality of code produced during short mobile bursts. Commenters worry about creating 'spaghetti code' or unmaintainable projects when working in fragmented sessions on a phone. However, others argue that for small utilities, prototypes, or personal tools, the quality is sufficient, and the ability to iterate quickly from anywhere outweighs the lack of rigorous structure.
20. Gatekeeping vs. Encouragement
Related: The comment section illustrates a divide between experienced users who gatekeep the term 'coding' or mock the setup, and those who encourage the experimentation. While some dismiss the article as trivial, others defend the author's enthusiasm, noting that lowering the barrier to entry and finding new ways to engage with technology is positive, even if the underlying methods are not strictly new.
0. Does not fit well in any category
</topics>
<comments_to_classify>
[
{
"id": "46522217",
"text": "> Tmux is helpful, too.\n\nYes. tmux is essential. It's great to be able to monitor a session from desktop, or to pick up an existing conversation i'm having on the computer on my phone. In my shell, I have gemini flash wrapper that I use to manage my sessions so I can quickly connect to an existing one, or create a new one with natural language.\n\n> He doesn't seem to be credited on this page, but I believe Pieter Levels (@levelsio) actually popularized this scheme. The author documents a nearly identical scheme.\n\nI've been doing this (tailscale + termius + tmux + ssh) for at least a year and a half. First with Aider in this exact setup, and now with Claude Code and Codex."
}
,
{
"id": "46526981",
"text": "So what about setting up a discord server for you and your LLM? Gets the notification benefit of E-mail but retains the immediate-resposne, no? That's how all my UptimeKuma notifs are setup atleast..."
}
,
{
"id": "46522282",
"text": "I can no longer edit this comment but it wasn’t meant to criticize the author. This is a great post. They are sharing their experiences and more importantly, teaching others.\n\nSorry if anyone, especially the author, took it this way."
}
,
{
"id": "46520431",
"text": "It wouldn't shock me if multiple people came up with this idea independently. I've certainly experimented with it over the last couple years."
}
,
{
"id": "46519056",
"text": "I've been using Claude Code in their iOS app (on a Pro account). I just point it at my GitHub repo, and tell it to work on one of the issues I created. It required very little setup beyond what I did for Claude Code CLI."
}
,
{
"id": "46521989",
"text": "i've seen that work well on existing codebases, but bootstrapping a codebase that way is like pulling teeth in my past experiences.\n\nso it really sort of falls back to what you're doing with the llm. code maintenance isn't novel development, which isn't polishing."
}
,
{
"id": "46523721",
"text": "> So this is the 4th+ article I've seen on using a VPN to vibe code on your phone.\n\nand all of them mentions Tailscale. I would not be surprised if we hear in a few days it got next big fund and all of this is just a preparation for it"
}
,
{
"id": "46524302",
"text": "I did this on my own without reading any of these articles - I already had a terminal program on my Android phone and was already using Tailscale for shared projects in Ghidra so... Maybe it's just a path of least resistance."
}
,
{
"id": "46524755",
"text": "a fully open source alternative would be netbird, it's based on wireguard as well, has 0 closed components but lacks some features (like IPv6 or internal CA).\nhttps://github.com/netbirdio/netbird"
}
,
{
"id": "46524226",
"text": "Right. For a simple setup I think using plain boring Wireguard is the better option. Boring is good."
}
,
{
"id": "46523968",
"text": "Tailscale has been a HM darling for a long time, this isn’t very surprising!"
}
,
{
"id": "46518281",
"text": "Why not use a browser?\n\nOpenCode has a webUI, you can simply host that on your machine at home and VPN to it.\n\nhttps://opencode.ai/docs/server/ (sadly no screenshots, but its a pretty good GUI, looks like their desktop app)"
}
,
{
"id": "46518572",
"text": "You need tmux to be able to resume the same session from anywhere, mosh-server to make ssh resilient to sketchy mobile connections, and blink shell https://blink.sh/ to have a high quality iOS shell with a mosh and ssh client built right in to resume at any time.\n\nFar more resilient and performant than a web client."
}
,
{
"id": "46518630",
"text": "Well the beauty is the logic lives on the server. The client is just a client.\n\nIf it disconnects you just reload the page. It can work just fine in the background because it’s not running on your phone.\n\nJust like you can refresh the ChatGPT website, but OpenCode lives on your pc at home, not OpenAI servers."
}
,
{
"id": "46518722",
"text": "yes but I've never seen a terminal interface embedded in a browser that is as good as a native terminal app interface, and blink shell has been well worth the upfront cost to me (way better than Termius, which was suggested in the writeup)"
}
,
{
"id": "46520498",
"text": "Is it easy to move the text cursor around in the text input in blink shell?"
}
,
{
"id": "46520885",
"text": "> and blink shell https://blink.sh/ to have a high quality iOS shell with a mosh and ssh client built right in to resume at any time\n\nI really like Termius, have you tried it? I think I tested out Blink when I was trying various SSH/shell apps and\nchose Termius over it, but it’s been so long now that I completely forget why.\n\nEDIT: does Blink give you a local shell as well like vs only SSH/mosh?"
}
,
{
"id": "46519431",
"text": "tried tmux but realized claude/gemini/codex's --resume works great and have since started using a single chat for all small work projects"
}
,
{
"id": "46520246",
"text": "From that page:\n\n> The opencode serve command runs a headless HTTP server that exposes an OpenAPI endpoint…\n\nUnless I missed it, there’s no mention of a web UI?"
}
,
{
"id": "46522956",
"text": "The docs are very behind, there is indeed a full blown webui and with opencode serve you can access it"
}
,
{
"id": "46520557",
"text": "`opencode web` runs the web ui. It's very good."
}
,
{
"id": "46519524",
"text": "Email might work, however if you're a Telegram user you could write a bot that runs on your home system that runs the cli commands on your behalf and then sends the output as a response to you. No need to open up any ports on your router."
}
,
{
"id": "46519877",
"text": "I didn’t know until I read this comment, but this is exactly what I want. A telegram bot with Claude on the other side and GitHub app to check out the code"
}
,
{
"id": "46521083",
"text": "I've replied with this in another comment, but this seems more pertinent ;)\n\nThats exactly the approach I took with https://github.com/cloud-atlas-ai/miranda , Telegram bot, PR is the human review point, tests + CodeRabbit catch most issues.\n\nBot intercepts Claude's AskUserQuestion calls via a hook, sends me an inline keyboard, injects my answer back into the session. Claude keeps working, PR still happens—but I can unblock it from my phone in 5 seconds instead of rejecting a PR based on a wrong guess."
}
,
{
"id": "46532830",
"text": "That looks really cool I'll have to check it out."
}
,
{
"id": "46521007",
"text": "I have custom scripts I use at home to keep track of various personal data, assisted by an LLM. The idea of using Telegram as a way to have a global, quick, and personal interface from my phone or tablet, is perfect and easy to set up.\n\nClaude is making it easier to have bespoke data and dashboards for anything. We're going to make a lot of them, for all reasons. I've also made apps with Django interfaces, but quick, global interfaces are going to become in demand."
}
,
{
"id": "46525112",
"text": "I concur, but I also think that Home Assistant could be used as a rock bed to build many of those dashboards easily. They just need to revert the \"go all in on UI first configuration\" and keep YAML declarations as first-class citizen to let LLMs easily compose dashboards based on user's desires."
}
,
{
"id": "46521922",
"text": "I've been using Telegram bot to talk to a Claude SDK agent who talks to my Claudes via tmux commands (all running on a DigitalOcean VPS)"
}
,
{
"id": "46524773",
"text": "Can you do it with signal?"
}
,
{
"id": "46520387",
"text": "Cloudflare worker would work, too"
}
,
{
"id": "46518331",
"text": "I have read of people doing remote coding with clause but through having Claude create pull request. The user then looks through the requests, and either approves or sends it back with edits. Seems like a good way to interact with Claude code, especially once one sets up a test suite and those proposed pull requests have proven not to regress."
}
,
{
"id": "46520430",
"text": "Same approach here. PR is the human review point, tests + CodeRabbit catch most issues -> https://github.com/cloud-atlas-ai/miranda .\n\nThe gap I wanted to fill: when Claude is genuinely uncertain (\"JWT or sessions?\" \"Breaking change or not?\"), it either guesses wrong or punts to the PR description where you can't easily respond.\n\nBuilt a Telegram bot that intercepts Claude's AskUserQuestion calls via a hook, sends me an inline keyboard, injects my answer back into the session. Claude keeps working, PR still happens—but I can unblock it from my phone in 5 seconds instead of rejecting a PR based on a wrong guess.\n\nWorks in tandem with a bunch of other LLM enhancers I've built, they're linked in the README or that repo"
}
,
{
"id": "46518960",
"text": "This pretty much sounds like my dream vibe coding dashboard - basically a personal Github populated by AI agents I can assign tasks to. Does this exist yet? Or can something like gitea be setup to behave this way?"
}
,
{
"id": "46519511",
"text": "In terms of issue tracking and agentic \"developers\", with a mobile focus -\n\nYou can connect Linear to Cursor's web agent, which makes Linear issues assignable to the agent directly and kicks off Cursor's take on remote coding agent. You can then guide it further via Cursor's web chat.\n\nIf Claude Code on iOS supported Linear MCP (as it does on desktop), you can run a similar issue handoff to agent to issue update workflow, albeit without direct issue assignment to the agent \"user\". Easy to use labels aka tags for agent assignment tracking, as well.\n\nFor my hobby projects, I've been using Linear + agentFlavorOfTheMonth quite happily this way. I imagine Github issues, Asana, whatever could be wired up in place of Linear."
}
,
{
"id": "46519625",
"text": "Gastown, by Steve Yeggs is that, via tmux. It's rather opinionated and still in development, but it's worth a look if that's what you're looking for."
}
,
{
"id": "46521066",
"text": "Steve Yegge is building awesome things in this space, but I've found them too heavy, started using bd when it was small, but now its trying to do too much IMO, so made a clone, tailored to my use case -> https://github.com/cloud-atlas-ai/ba"
}
,
{
"id": "46532478",
"text": "durch - just starred this repo! Looking forward to testing it out as I learn how to build with multiple agents.\n\nI'm just starting out with building with Claude - after a friend made this post he sent me a Steve Yegge interview ( https://m.youtube.com/watch?v=zuJyJP517Uw ). Absolutely loved it. I come from an electrical/nuclear engineering background - Yegge reminds me of the cool senior engineer who's young at heart and open to change."
}
,
{
"id": "46519927",
"text": "This is how I do mobile device coding. Android terminal w. git and gh installed and authenticated. Claude manages the feature branching and PR process; I review the PR in the GitHub mobile app."
}
,
{
"id": "46524321",
"text": "Claude Code does a very decent UI, somehow the text mode is much more attractive. As if there is once in a lifetime opportunity to make the console great again."
}
,
{
"id": "46521866",
"text": "> What is the downside to using email?\n\nIf true to the post, it lacks \"real time\". Doom scrolling by nature is while chat is async. Refreshing Gmail constantly is not fun."
}
,
{
"id": "46518765",
"text": "I’ve been doing some of this through a term on my phone, but it honestly sucks. Other interfaces (telegram, web ui, email) are gonna be much better experiences on your phone."
}
,
{
"id": "46518367",
"text": "ive tried slack before, but a challenge is how well you can get results returned back in a way where you can actually see what it did and give proper next steps\n\ngetting a PR back and being able to put comments on it is fine, but ive had middling success getting qcli at least to actually match the comments with the code that was commented on. i get the sense that there isnt any training with the comments inlined well on a diff:/\n\nit doesnt have to be a vpn though, i was on an oauth webbrowser terminal, and things like coder[0] let you run vscode on the browser, including on your phone browser. there's also happy coder[1] which i tried using to connect between the new builtin android linux vm, and skip all the remote stuff entirely, but the phone would inevitably kill the terminal runbing claude, killing the whole thing. you can currently just run claude from your phone in that, which only has the problem that when the vm crashes, all you can do is wipe the partition.\n\n[0] https://coder.com/\n[1] https://happy.engineering/"
}
,
{
"id": "46526163",
"text": "No syntax highlighting, I do like to review snippets of code. Also the interactive questions / answers during planning would be a pain over email. And what about text wrapping? Headache.\n\nEdit: also setting up an email interface API to Claude Code seems like a lot more work than just setting up a VPN."
}
,
{
"id": "46527501",
"text": "it looks like some kind marketing push or 'growth hack', just to get some viral thing around which justify why do you extra reason to pay for Claude or Tailscale subscription.\n\nI personally not even convinced that Claude Code any better on average than something like Aider+Gemini 3 or other good model. May be in some specific cases it actually better but in those Aider+'Antropic Model via API' most likely will work too."
}
,
{
"id": "46522395",
"text": "If setting up a VPN is that difficult for you you may have bigger problems my friend. (I joke). But really I am surprised that a VPN is the part you take issue with."
}
,
{
"id": "46522082",
"text": "You don't need a VPN to vibe code on your phone. I've been happily doing thumb-driven development for the last 4 months now using GitHub Copilot on github.com from my phone. It even has real-time chat with copilot as it works. Having your PRs deploy to an environment allows you to check it. I also have playwright tests that record screenshots and traces that get uploaded as artifacts I can check too."
}
,
{
"id": "46525255",
"text": "So basically back to the chat-interface. You could also replace e-mail with WhatsApp, Telegram, Slack, Mattermost or whichever you prefer, it would be all the same."
}
,
{
"id": "46518153",
"text": "I see no downsides. Seems like an actually useful udea."
}
,
{
"id": "46518172",
"text": "that would be perfect ."
}
,
{
"id": "46518215",
"text": "This is genius! The tailscale vpn was stupid easy to setup (I'm a near novice and figured it out). An email interface with progress updates would be even better than doom coding."
}
]
</comments_to_classify>
Based on the comments above, assign each to up to 3 relevant topics.
Return ONLY a JSON array with this exact structure (no other text):
[
{
"id": "comment_id_1",
"topics": [
1,
3,
5
]
}
,
{
"id": "comment_id_2",
"topics": [
2
]
}
,
{
"id": "comment_id_3",
"topics": [
0
]
}
,
...
]
Rules:
- Each comment can have 0 to 3 topics
- Use 1-based topic indices for matches
- Use index 0 if the comment does not fit well in any category
- Only assign topics that are genuinely relevant to the comment
Remember: Output ONLY the JSON array, no other text.
50