Increasingly thinking that we’re too worried about AI hallucination. AI is never going to be perfect, and humans aren’t either! We should focus on using AI in the right context. Running without supervision is the problem. If a 5-year-old kid shouldn’t be in charge of something, AI shouldn’t either.
We’ve gotten some good feedback about the latest Core Intuition, so this is a rare “in case you missed it” post… Episode 579 is a good place to start if you want to pick up the podcast. AI, Beeper, and related fallout.
Not trying to overhype this but I’m so excited about what we’re going to launch in Micro.blog next month. It’s coming along really well and expanding beyond what I first thought it would look like. Meanwhile the everyday stuff continues too, working on fixes and servers this week.
Finished reading: Age of Legend by Michael J. Sullivan. Didn’t expect this one to end so suddenly because the last 15% left on Kindle was actually an excerpt from another book. 📚
Speaking of self-promotion… 🙂 There’s a new Core Intuition out. On episode 579 we talked AI, Google Gemini, Apple’s MLX, Beeper, iMessage, and all the tech and ethical implications of the above. Recorded before Beeper Mini broke yesterday, so we’ll see how that shakes out.
Link previews or ads https://www.manton.org/2023/12/09/link-previews-or.html
Thinking about what Mosseri said introducing Threads search:
"…having a comprehensive list of
every post with a specific word in chronological order inevitably means spammers and other bad actors pummel the view with content by simply adding the relevant words or tags. And before you ask why we don’t take down that bad content, understand there’s a lot more content that people don’t want to see than we can or should take down."
This is what we’ve always thought at Micro.blog too. Meta solves it with algorithmic timelines. We don’t do that, so we’ve been admittedly slow to improve here.
I’m using ChatGPT more and more for coding help. Sometimes I don’t fully trust the answer and double check with Google or Stack Overflow. Today the AI produced some code that I thought it must have completely hallucinated… Does this even work technically? But it ran perfectly.
Love developing for the web and learning how to do something new for the first time. Processing a ZIP file fully within JavaScript in the browser instead of server-side? Never really considered doing it but it works great.
Sometimes the choices we don’t mean to make end up perfect. When I was traveling this year my shoes fell apart and I stopped at a random shoe store in Boulder. They had one pair of shoes in my size, a brand I’d never heard of, bought them, and they are now some of my favorites. Wear them everyday.
Threads launching in the EU next week. My earlier theory was they were waiting for ActivityPub (and so better account portability) before going live. Guess I was wrong.
https://www.theverge.com/2023/12/8/23993595/threads-meta-eu-europe-launch-date
Noticed load averages briefly jumping up to 10.0 so it’s going to be that kind of day. There’s always something to optimize! Looking okay now, though.
It’s been a couple weeks since we got our first Roomba, so here’s my review: it’s great. Took a day for us to settle into how best to use it, what to pick up from the floor or block, and when to run it… Now it just does its thing every day. Surprisingly good with dog hair too.
Testing new in-progress features and everything is kind of working. Maybe at this point in my career I shouldn’t be surprised when my code actually works, but it’s still a nice feeling.
Interesting juxtaposition of releases today… The new Ivory with Mastodon hashtags support and the Threads update with hashtags for the first time, limited to one tag per post. Also major Mammoth release, with open source on the way.
Watching a few clips from last night’s GOP debate. Chris Christie still fighting the good fight against Trump when no one else will. From Christie’s closing statement:
"…picture in your mind election day. You’ll all be heading to the polls to vote. And that is something Donald Trump will not be able to do. Because he will be convicted of felonies before then and his right to vote will be taken away."
🇺🇸
I love these robots. We’ve seen more of them but it’s not clear if they are actually going anywhere or just practicing for deliveries.
Congrats to @cheesemaker and the Silverpine team for announcing the beta of Evergreen.key, an interactive fiction authoring tool. Think: “choose your own adventure” stories. I’ve been helping with this behind the scenes too, and the upcoming iOS app… Really cool to see it come together.
https://micro.blog/cheesemaker https://jonathanhays.me/2023/12/06/evergreen-ink/
Beeper Mini looks like an impressive feat of reverse-engineering. Apple will surely try to break this in the future, but if doing so requires an iOS update for iMessage API tweaks, it may take years before they can realistically cut off Beeper along with very old iPhones.
https://www.theverge.com/2023/12/5/23987817/beeper-mini-imessage-android-reverse-engineer
Made a little architecture diagram for Micro.blog that reflects the recent server upgrades. A couple simplifications but it’s pretty close to how things work.
