Issue #24

Summer’s just started and… wait… summer’s almost over! Eeep!

Japan. Vegas. Surfing dogs. The kids start getting shipped to school everyday again, and the NFL’s ramped up again. What wonders will this coming fall bring us? What will the coming winter conceal and reveal to us in its darkness? The anxiety of the unknown in uncertain times is also the wonder that life gives us.

Or, simply put: What’s next?

Japandemonium

Although I wish we could have stayed longer and visited more shrines and temples (especially in Kyoto), we made the best of our 10 days there. We were in Kyoto, Osaka, and Tokyo: 3 days each. I gotta say, Kyoto is my favorite, and everyone agreed we should have had one more day there.

We visited a shrine shaped like a big lion’s head in Osaka. We rented Yukata in Kyoto and walked around town during the pre-amble to the Gion Festival. We soaked in an onsen (public bath house). We met old friends, did some bar hopping, and even went clubbing in Osaka like the old days.

And the yen. OMG, the yen is so cheap right now. Coming from SF Bay Area prices, everything felt like it was half priced. We had a nice dinner that would have cost us $300-400 that ended up being closer to $150.

There’s plenty more I haven’t scribbled out yet. More to review on the trip when time permits.

Hackitty hack hack

Shortly after Japan, I packed up my things again and headed off to the desert to join my crew for DEFCON 34. I took the test and got my amateur radio license at DEFCON! I played in a ham radio fox hunt, got a portable Meshtastic node, checked out First Friday in the Arts District. I saw Bruce Schneier give a talk. What a blast.

After checking out of the hotel, getting to the airport, and waiting… 4-5 hours… my flight got canceled and Alaska Air housed me for one more night, doh! I was pretty pooped by then, so I decided against another night of hanging with the silly hacker kids, but I did manage to find a hack for my meal vouchers (grin).

Typically, after leaving DEFCON, I feel a sense of dread: the world is falling apart, everything is hacked! But after this year, I felt a camaraderie with my fellow hackers: the world is falling apart, but it’s still important for us to be humans with each other. I don’t know how to point at why I feel that way. Maybe just talking with people, feeling the creative energy, getting the sense of community, and just being kinded and being a kinder to folks.

Private Agent Man

Using AI regularly, it’s good to refine how we’re interacting with our agents. I think that, much like how we communicate with humans, it’s useful to refine how we communicate with AI. The better we get at it, the more useful the AI becomes, and the less annoying it gets in trying to get it to behave the way we want it to.

Speaking of annoying, the latest edition of iOS Dev Weekly was kind of irksome to me with its use of LLM writing. It’s not that I’m against using AI to help you write, but I am against using it blindly and producing bad writing (as opposed to just using your own human mind to make the bad writing, like me?).

Most folks are running GPT or Claude from the big AI companies. I prefer not to be sending so much to them, especially for privacy reasons. It’s so easy to just start giving more and more info to AI for the sake of convenience. We need to own our sessions in our harnesses. The memory should be under our control. We should not have to worry that we’re going to be pushed advertisements and given recommendations that may not be in our best interests when using AI agents.

I’ve been super interested in having local models run on my own machine. And it seems more folks are doing so. Friends from DEFCON noted how they’ve downgraded their Claude MAX plan to using their own local models for coding. I’ve been toying around more with it myself. This is the way.

Lilmem 2.0 is still…almost ready

Although I use Little Memory as my daily journaling app, I still keep a handy moleskine on me, and I write in it with those silly things at the ends of my arms: writing by hand is good for you.

Looks like Day One is pushing AI features on its users. I’ve been cautious about this, and also I’ve been keeping my eye on what’s been going on with local and on-device models. A huge part of the Lilmem 2.0 rewrite is to have your memories on-device, and sending your private memories to the cloud to get them sucked into the AI machine isn’t really what I want.

Speaking of machines: here are some nerdy debugging codes I learned about when debugging the 2.0 app. Beep beep boop.

Afterthoughts

Oooh, Hank’s making a new light with a built-in USB-C charger!

I’ve now got Musings for the site. So I can spit out more
silly things at you faster.

I’m available for iOS app consulting. Native apps, SwiftUI, App Store. If you know someone who needs a hand, I’d appreciate an intro.