claude code + gemini + codex running side by side. tab colors show whats working vs waiting. schedule tasks, loop commands, access from your phone. side project called PATAPIM -- free for solo devs patapim.ai
English
German Heller
31 posts
German Heller
@germanheller
good stories are made by good characters ||| CEO @ 3dar
Buenos Aires, Argentina Katılım Aralık 2010
140 Takip Edilen72 Takipçiler
@Sabrina_Ramonov /init changed everything for me honestly. the claude.md file it generates means you dont have to re-explain your project structure every new session. saves so much time on bigger codebases
English
Most people open Claude Code and start typing random requests. Here's what the top 1% do instead:
Run /init on day 1. It scans your entire codebase and creates a Claude.md with your project's structure, conventions, and patterns.
Now every conversation starts with full context. No more explaining your stack 47 times.
One command. 30 seconds. And your AI assistant goes from intern to senior engineer.
▸ Type /init in your project root
English
@Sabrina_Ramonov "procrastination wearing a backpack" — stealing this.I spent months reading about AI workflows before I just opened a terminal and started building something dumb. The gap between watching tutorials and debugging your own project is where all the real learning hides.
English
Learning about AI feels productive. You're reading threads, watching demos, comparing tools. But consumption without creation is procrastination wearing a backpack. The moment you stop watching and start building, everything changes. You learn 10x faster by doing than by studying.
English
@raroque respect for sharing this publicly. solo dev means you eat every failure alone and the transparency helps others avoid the same traps. supabase outages hit different when theres no team to fall back on
English
So my app just had its WORST week since launch 😅
2 separate outages (supabase/perplexity) AND we had a little security incident 😮💨
i know ill get some heat from experienced devs but i wanted to share this anyway
learn from my mistakes please 😅
English
@MatthewBerman multi-day autonomous agents are where this is all heading. the interesting problem becomes monitoring -- how do you keep track of what 5 agents did overnight without reading thousands of lines of logs
English
@trq212 this is huge for multi-project workflows. constantly re-explaining project context to claude was my biggest friction. if it remembers architecture decisions and debugging patterns across sessions thats easily 10-15 min saved per session
English
@Sabrina_Ramonov the leverage is insane. I run multiple AI coding sessions in parallel now and ship in hours what used to take a week with a small team. the bottleneck moved from execution to decision-making -- knowing what to build matters more than ever
English
10 years ago, building a real business alone meant burning out. Today, 1 person with AI can automate operations, create content, write code, and serve customers at a level that used to require a 10-person team. We're living in the best era for solopreneurs in history and most people don't see it yet.
English
@onecharteroak @lawrencecchen exactly this. and the state detection has to be fast enough to be useful. a 2-second delay between the agent finishing and you knowing about it means you already switched context to something else. sub-second visual feedback is the bar
English
@lawrencecchen The blue ring concept is smart. Biggest friction with running multiple agents in parallel is knowing which one needs you. Visual attention signals solve that without constant terminal switching.
English
Introducing cmux: the open-source terminal built for coding agents.
- Vertical tabs
- Blue rings around panes that need attention
- Built-in browser
- Based on Ghostty
When Claude Code needs you, the pane glows blue and the sidebar tells you why.
No Electron/Tauri. Just Swift/Appkit.
English
@nyneurotech the biggest gap rn is knowing which agent needs attention. you end up alt-tabbing between terminals checking if theyre done or stuck. visual cues would help -- something that surfaces which session is idle vs thinking vs waiting for input
English
we need agent orchestration built around minimizing context switching
running multiple agents is the new opening multiple tabs
English
@fuzzylogic003 @kylebrussell @steipete i keep a second terminal with git status open. after each prompt cycle i check what changed. some people split their screen -- terminal on one side, vscode on the other just for the file tree and git diff view
English
@kylebrussell @steipete How do ppl see file structure and changes being done to their files if they only use claude code in terminal. Genuine question.
English
Today we announced we’re removing >90 Cursor seats because they haven’t had any use in two weeks
Kyle Russell@kylebrussell
This morning @wangandrewd requested that his Cursor seat be removed since he's so deep into Claude Code and it kicked off an internal cascade of requests within Valon 😬
English
@ThePrimeagen at this point I just assume every video of a public figure is AI until proven otherwise. the uncanny valley is almost gone
English
these new ai videos are getting really good.
if you see me in a music video, its probably ai
English
@noahzweben remote control from phone was the feature I didnt know I needed. being able to check on a long running task from the couch instead of going back to the desk is a game changer for multi-agent workflows
English
In the spirit of building in public:
📈 Wow this was a popular feature. We've scaled up some capacity on our end to accommodate usage. Sorry if you ran into some connectivity issues in the past few hours
🐞 Some text may be covered on Android. We're on it.
🐞 GH auth is required to see Code pane on mobile (you can still deeplink to your session via QR/link)
✅ Plan mode and AskUserQuestion should be fixed
🚧 WIP - making permissions more flexible in remote control sessions so you can do things like always allow etc...
🚧 Desktop support
And maybe soon ... letting Pro users touch grass too 🌴😎
Noah Zweben@noahzweben
Announcing a new Claude Code feature: Remote Control. It's rolling out now to Max users in research preview. Try it with /remote-control Start local sessions from the terminal, then continue them from your phone. Take a walk, see the sun, walk your dog without losing your flow.
English
@aicodeking same experience here. tried 3.1 pro for a couple days and went back to flash. the extra thinking time doesnt translate to better code, just slower iteration. for agentic stuff speed + reliability matters way more than raw reasoning
English
TL;DR: 3.1 Pro is worse than Gemini 3.0 Pro in all of my tests and daily usage. It thinks too much, costs more and is not a good experience at all for the price. 3.1 Flash is a better google model IMHO.
AICodeKing@aicodeking
Gemini 3.1 Pro is one of the worst models in recent times for Agentic Coding: youtube.com/watch?v=R3GjTB…
English
@MatthewBerman been using opus fast mode all day and honestly cant tell the difference in output quality. if anything its better because I iterate faster and catch issues sooner. speed matters more than people think when youre in a tight feedback loop
English
@Sabrina_Ramonov painfully accurate. went through the same loop -- testing every new AI tool instead of building. committed to one setup and output 3xd in a week
English
If you've tried more than 10 AI tools this month but haven't shipped a single project, you don't have a tool problem. You have a commitment problem. Pick 2 tools. Delete the rest from your bookmarks. Go deep for 90 days. That's the whole strategy.
English
@morph_thecat clean setup guide. I pair /remote-control with patapim.ai so I get the QR phone access plus actual state detection on desktop — shows me when the agent is thinking vs waiting for input vs done. nice combo
English
Woke up to Claude Code /remote-control on my machine 😃
Quick setup:
1️⃣ /config → turn on remote control for all future sessions
2️⃣ /rc → generates a QR code
3️⃣ Scan with your phone → you're in (pre-req: max user for now, your turn is coming)
#ClaudeCode #AI #BuildInPublic
English
@kkmaway the feature velocity is insane. and third party tools are filling gaps fast too - patapim.ai already has the remote monitoring + state detection + multi-terminal that Claude Code alone is still missing. ecosystem is moving
English
@CodeAkram @felixrieseberg @sytaylor lol no rest for the wired. I stack /remote-control with patapim.ai - it gives you a terminal grid so you can watch multiple agents at once plus voice dictation for when you really wanna code from the couch
English
@felixrieseberg @sytaylor When you’re all taking time off, I’m willing to jump in and Claude Code while y’all chill, but I guess y’all won’t really chill cause /remote-control is here btw
English
Can Anthropic just CHILL for like a day so we can process this stuff?
Shitting hell man
Claude@claudeai
New in Cowork: scheduled tasks. Claude can now complete recurring tasks at specific times automatically: a morning brief, weekly spreadsheet updates, Friday team presentations.
English
@yaelkroy nice breakdown. if you want the monitoring piece without the phone, patapim.ai does this natively - shows agent state in real time, plus multi-terminal grid so you can run several sessions at once. different approach but solves the same visibility problem
English
Claude Code just added remote control — a “monitor + steer the agent from your phone” workflow.
How it works:
You still start the session on your PC (so it’s not full mobile-first like OpenClaw).
Then run in terminal: claude remote-control
Connect via a link or QR code from your phone in the Claude app or browser.
From there you can:
add/adjust prompts
monitor status and diffs
interrupt or redirect the run
basically manage it like a normal chat, but tied to the live coding session
Availability: research preview for Max users now, with Pro support promised soon.
Docs: code.claude.com/docs/en/remote…
#AI #Claude #Anthropic #ClaudeCode #AIAgents #DevTools #RemoteControl #LLM #Coding #Automation #SoftwareEngineering
English
@jonnokim this is the way. I use patapim.ai so I can check agent state at a glance between sets - no phone needed, just look at the screen across the room
English
