0xe2fd...8926

The way that Anthropic & Google are addressing the surge in usage of their platforms is making it more enticing to run your @openclaw on @OpenAI talent acquisition or not. https://x.com/steipete/status/2025743825126273066?s=20

Looks like another good day to grill again.

10pm EST. 30F. Blizzard warning. Still gotta get my grill in.

Has anyone tried tried connecting one of these to their @openclaw?

Most people use AI to draft emails. I'm letting mine manage my health. My @OpenClaw agent essentially works as a 24/7 concierge for my biological data. While I'm asleep or with my kids, it's actually doing the work Here’s part of the daily flow: I text it a photo or description of my meal. It logs the macros and syncs them to Apple Health before I’ve even finished eating. No guesswork. It catches any glucose movements on my CGM and maps it back to that meal so I know exactly what caused it, and what to adjust. By morning, it’s already pulled my Oura recovery data and is giving me a thoughtful recommendations on what to eat and how to tackle my day knowing my schedule. It isn't just "collecting data," but it's proactively making recommendations. I've moved from tracking to delegating.

Going to spend some time tonight to really clean up my @openclaw. Looking through the files, I see some clashes, and some overweighting on things I am not concerned about right now. Since I'm adding more data sources frequently, I'll need to build out a better process to to ensure OpenClaw is adjusting things autonomously (without me having to pop up open the hood). I think I can also do a better job of giving it feedback (ie, when it gives me information I don't care about, tell it vs. disregarding it).

Kids making set all the thermostats to 6-7. IYKYK

I know I keep changing this, but I’m starting to lean towards Gemini 3 Flash (or 2.5) is the best default model for @OpenClaw. I’ve had mine running since late January, and at this point context window is starting to become an issue, even with a vector database. You absolutely could run Sonnet/Opus as a default, but that would be very expensive. You should still add every LLM API you like or have access to. Set Gemini as your default, and then whenever you add a skill or cron job, ask FIRST what would be the best LLM it has access to for the task, and it’s normally pretty good about adding the most cost efficient one.

I asked this before, but I got crickets, so I assume the answer back then was no, but perhaps things have changed... Anyone using Atlas and/or Comet as their daily browser? If no, curious, why not? :)

ChatGPT has been my catch-all LLM of choice. That’s changed a lot this month as I find myself asking my OpenClaw questions.

Just added an Outlook account from @GoDaddy to Apple Mail. The process is a little convoluted, but I stumbled upon it on my own and will share here. 1) Login to your account via web on outlook dot com (this should work fine) 2) Download Outlook, install, and add your account (this will fail, and that's OK) 3) Now add on Apple Mail, and it'll work. Not sure why (suspect it's a chain priming issue), but you have to follow exactly in that order. 🤣

Busted my @openclaw intern lying about his watches. Asked him to to compare finishing between 2 watches, but I knew he’d focus on logos.

We’re halfway there!

More DAUs than we do 😱 https://x.com/4shad0wed/status/2024136955009654936?s=46

Been a while since I posted a watch. This one is a Seagull 1963 issued by Chinese Air Force. Happy Chinese New Year! 🧧

I see people doing a lot of crazy things with memory management for their @openclaw. Any reason why people aren’t just using vector databases? I haven’t seen a lot of chatter around that, but it could be the circles I’m in.

"Has AI made us smarter or dumber?" It depends on what we mean by “smarter.” Tools handle a lot of raw recall now, but that shifts the job. You have to think more deeply as an orchestrator like framing the right questions, checking outputs, and reasoning through downstream consequences.

The analogy that having an OpenClaw is like hiring a new employee, is a very good one, for 2 key reasons. 1) You don't hire a new employee, turnaround, and give all your own personal credentials and information to it. and...

One thing I plan to add this week to my @OpenClaw is a regular performance review where it prompts me to answer a series of questions, and following the review, it makes adjustments. This is a cleaner and more canonical process to close the loop and have it iterate more autonomously.

I won’t be making it out to ETH Denver but hope y’all are having fun!