runesleo

Now when I want to post, I just tell Claude on my phone, “Save it as a scheduled draft,” and I’m done.
It handles the rest:
1. First checks which posts are already scheduled for today, leaving at least 1 hour between them. I ran the stats on my own original posts over the past two months: posts less than 15 minutes after the previous one had a median of only 35 impressions; those spaced 1–3 hours apart had 2,395
2. Automatically posts at the scheduled time, including quote posts
3. Before posting, verifies that the text matches exactly and that the quote card is present; afterward, checks the

When you have a bunch of Claude conversations open, the most annoying thing is the titles: you start with “Work Priorities,” then somehow end up fixing the radar and clearing the disk, and later you can’t find it at all.
Here’s a small trick: let Claude rename the conversation itself.
The Claude desktop app (Code mode) already has a tool for renaming conversations; it just doesn’t use it proactively by default. Add one line to your global rules (the original text is shown in the image), and it will rename the conversation whenever the main thread shifts.
Result: “Work Priorities” → “Disk Clean

When you're out with only your phone and want to start a new local Claude Code session on your Mac at home:
Method: keep this running on the Mac
claude remote-control --spawn=same-dir
Then configure it to launch automatically at startup (launchd). In the Claude App on your phone → Code, you can directly select this computer to start a new session. The session runs locally, so it can use local files and MCP.
Two pitfalls:
1. Without the parameter, it gets stuck on an interactive prompt, and when running in the background, it will exit because it cannot read input.
2. Sessions from the desktop a

Small bug report @claudeai: when using Remote Control on mobile for a Claude Code session running on my Mac, the effort level shows "Medium" on mobile, while the desktop shows (and actually runs at) "Extra high". Trying to change it on mobile fails with "Unable to change effort. Please try again." Latest app version. Expected: show the real level, or let me change it remotely.

iPhone → Mac remote control is finally working.
I use MCPX / AI Agent to run tasks, and what I fear most is getting stuck on GUI authorization. So I used NetEase UU Remote to provide a manual fallback.
After running into three pitfalls—Clash TUN, API timeouts, and same-account control permissions—the phone was tested to successfully control the Mac.
The Agent executes, monitoring detects anomalies, and the phone takes over. Automation doesn’t have to be error-free, but it must be able to continue after getting stuck. (Screenshot redacted)

A few days ago, I said that what personal AI really needs is not a Memory document, but a long-term memory system.
Over the past two days, I actually rebuilt it from the ground up.
In the end, I discovered:
Memory still isn’t the hardest part.
The truly difficult part is Continuity.
I already have 2764 ChatGPT history entries, along with local full-text and vector search.
In theory, AI can now find at any time:
“What did I say before?”
“What did I research?”
“What did I do?”
But in reality, an absurd problem still occurs:
AI finishes researching a project.
It reaches a conclusion.
It even know

I had my Dot design an exclusive look for itself and set the avatar as well.
I only gave it one prompt: You are my long-term AI command center, coordinating multiple Agents and workflows and spotting important signals early. The look should be calm, sharp, and reliable, recognizable even at a small size; no text, no generic robot, and no cuteness for the sake of being cute.
It chose a dark blue owl with amber eyes and a track pattern on its chest representing multiple workflows, and even gave the look a name: Sentinel.
From generating the look and creating the animated pet to setting it as the