Mher

🤯 I turned my apartment into a video game/simulator using Claude Code. 🎮 You can play it here: mher.ai Here are the steps I followed: 1. Ask Claude Code to spin up a new ec2 instance for me 2. Install Claude Code on the ec2 instance and run it with `--dangerously-skip-permissions` 3. Took images of my apartment and put it in a folder for Claude Code to view 4. Ask Claude Code to build a threejs simulator of my apartment 5. Ask Claude Code to deploy on an nginx server Inspired by @levelsio and his post about VibeOps

@hxxwhite @1_omar_ramo Can I run it in my own infra? That’s a hard requirement for a lot of companies

@1_omar_ramo With revyl, you can spin up as many simulators or emulators as needed for you or your agents, from anywhere. Can use for development, testing, etc! revyl.com

Serious question for other mobile devs: Why waste compute running Xcode sims locally, when you can just stream them from the cloud?

@trq212 For those of us who use Claude Code in the terminal, how do you suggest we leverage HTML? Just had it produce files and open in browser?

@Silentolls @rob_otor @Dimillian Wow, one-shot?

@rob_otor @Dimillian I just tried that lol. Its not terrible. A bright future is near

I've asked Codex for simple polygons but it seems like it cooked way beyond that goal

@ruben_kostard @ForgeCAD ForgeCAD is awesome, been using it to produce STL files for my 3D printer. I’m in Austin till Thursday, let’s get a coffee or drink!

In Austin this week, who should I connect with for @ForgeCAD?

The way I use plan mode has completely changed. The plans are all implementation plans now to help the AI write the code, not for human eyes. My workflow recently is doing a big context dump (as well as setting up any skills/mcps/etc) then ask it to interview me with AskUserQuestion tool to generate a spec that I’ll read. Then plan mode and implement. Works really well. Interview is crucial to refine spec so it has less room to make its own decisions/hallucinate.

slowly starting to use plan mode a LOT less nowadays i realised whenever i use plan mode, it generates a gigantic plan and then i dont read it and hit build out of laziness having a meaningful conversation with the AI agent to discuss implementation feels a lot easier 🤔

@alliekmiller @claudeai Just use the Claude Code Guide subagent that comes built-in!

@claudeai I need to set up an hourly cron job just to check the Claude X account so it can teach my Claude how to better Claude 😎

New in Claude Code: auto mode. Instead of approving every file write and bash command, or skipping permissions entirely, auto mode lets Claude make permission decisions on your behalf. Safeguards check each action before it runs.

@LogicalThesis Have you tried getting your agent to use the Databricks CLI instead of building an MCP? docs.databricks.com/aws/en/dev-too… You can make an Agent Skill that instructs your Claude how to use the CLI to perform common tasks.

After 10 iterations of a task on Databricks MCP via Claude/Cursor lol

@avarant42 Looking forward to trying this out, have been pretty hesitant on running OpenClaw due to security concerns…this might just be the perfect replacement

Built my own OpenClaw to use claude code via Telegram without running it in dangerously skip permissions mode. I approve requests via Telegram bot UI buttons. + it supports streaming github.com/avarant/telegr…

@addyosmani Would love an official agent skill that works with this CLI!!

Introducing the Google Workspace CLI: github.com/googleworkspac… - built for humans and agents. Google Drive, Gmail, Calendar, and every Workspace API. 40+ agent skills included.

@arslandevs @steipete Impossible to detect, but user submitted downvotes would help a lot

@steipete true but sadly this is inevitable😢 need some tool that detect ai and flag it here

Folks, please don't build bots that automatically reply to stuff on X, or use AI to reply. It makes this site annoying to use, I can barely keep up with the blocking. If you use AI to "tweak" your real replies, it will still smell like AI. Embrace typos and imperfect grammar.

@kr0der Not slow anymore, the script has flipped. Claude Code with Opus 4.6 is significantly slower for me now compared to Codex with 5.3.

does anyone else like that Codex 5.3 is slow? it feels relaxing running multiple chats/projects in parallel since you know you can go do something else while waiting and you know you're gonna come back to a 1 shotted task (assuming your prompt was good)

@VictorTaelin Done this multiple times and it works with no issue. It can one-shot simple to medium complexity apps. You definitely need to use a good ios development skill though

so I just attempted to vibe-code an iOS app with Swift and Codex 5.3. I never used Swift before. I just asked GPT to add features without telling it how it should do it, and with no idea how things work it failed miserably I guess I get how most people experience AI now

@rtwlz VPS. Use self-hosted OpenSearch for hybrid search. Setup firewalls for security. You can ask Codex or Claude code to set it up for you and it should handle everything.

Jmail has crossed 450M pageviews, and our Vercel bill has exploded, even after tons of cache mitigation. Many kind people chipped in to cover the bill, but it’s not sustainable. What are solid, cheaper alternatives?

@filippkowalski Eh it takes like an hour or two its not that bad for a one time thing.

@mher_ai waay to slow, i just type "release" and forget, 10 minutes later both ios and android are submited

This is super handy With this Claude can manage a lot of the app store related stuff on it's own

@hasufl Integration with Apple ecosystem, major plus.

are most of these clawd bot users running local models, or simply giving a server-based model access to all of their files + passwords? if not local model, what's the point of buying clawd its own dedicated machine? is it the optics of a dedicated piece of hardware in your own room that bamboozles users into thinking its "local"?

@steipete @mysticaltech How many developers are working on your projects? I find that if i’m working solo using branches isn’t worth it, different story with a team of devs

@mysticaltech I don’t use branches. just main

me: i don’t use worktrees, I just have multiple checkouts because less mental load people: 500 replies with over-engineered worktree management apps ☠️🙃

@jasonbosco @jshchnz Same here

@jshchnz Yeah I see it too

Anyone else seeing this behavior in Claude Code? Been driving me crazy recently

@chongdashu This is really awesome. Crazy how far vibe coded games have come (especially with the skills you've created).

Some updates today - Archers shoot arrows now 🏹 - Added foliage, trees, etc. - Better level design - Castle now has health and takes damage Opus 4.5 + Phaser agent skill

@deepfates SuperWhispr works with local models.

Can everyone please stop using wispr flow superwhisper and whatever subscription app for your Mac please! Offline local models are now better than the ones they're charging you for. Use something like "spokenly" with Nvidia parakeet models and thank me later