nvahalik

How deep are you going? What gas mixtures are you using? Which Garmin are you using? If you're using a Mk3i or a G2 (or related) you are using a diving computer in a smart watch form factor. It is custom built for this... unlike the Apple Watch Ultra.

130ft/40m is standard limit for non-tec diving.

Past that depth, you will be using diving computers and not a smartwatch. And the training you receive will help you to understand why.

Excited for jj converge - heuristics to (hopefully) automatically resolve divergence.

But that's the thing—there isn't any real automation for this. An adjuster has to physically go out to someone's property. They have to take pictures. They have to inspect things. They have to coordinate with 3rd parties if necessary (like engineers).

There isn't a whole lot of room for automation in much of this.

Underwriting on the other hand, that is a different story.

AI isn’t replacing adjusters.

Anecdotally, many of them are aging out and the demands of adjusters are increasing while the pay hasn’t necessarily gone up. While technically you don’t have to know the ins and outs of construction, it helps a lot and there are just not a lot of people who have that knowledge. And if they do, they often go into construction themselves or sell roofs. Both of which can be far more lucrative.

On top of that, claims software has a lot of automation but very little real AI surface area. A lot of the things AI is good at (text generation) must have certain verbiage or use legally approved templates.

Granted, this may not be the case for all verticals, but this is just my observation from 20 years. (Mostly in smaller IA firms and state-based institutions).

Also, my wife has pointed out that there haven’t been any major cats in the last few years. Katrina kept people working for years. A lack of big storms has also contributed to this.

A little south from you. Car said 102. Some of that depends on where you are. Parking lot, vs say, at the lake or rural areas.

How do you figure? I imagine for _most_ MCP implementations there is some level of technical overhead for having to keep a connection open. I can't imagine this would make huge differences in the token utilization.

Good iterative release. Biggest thing I'm wanting to really try out are the URI-based elicitations, which should allow us to use MFA for interactions happening in MCP.

File management is still a little wonky. I hope we see some additional movement with file uploads soon: https://modelcontextprotocol.io/community/working-groups/fil...

Interesting! I will look into this.

I really want to use this but am stuck (right now) having to use Caddy's docker tags integration for name resolution.

Man. I built an AI food system at my previous company and it was tough: we ended up just using it as a way to look up foods in a real DB and allowed guestimation but ultimately the win was "I don't have to search for everything on this place" we surface _what_ and then allowed the user to enter the real weights.

And this... really was and (will be) the only way for this to ever work.

If you have a dependency tree like this:

       / A \
      -- B - Megamerge 
       \ C /

There is nothing stopping you from doing this:

       / A - D \
      -- B ----- - Megamerge 
       \ C - E /

(Edit, or even this:)

       / A - D \
      -- B ----- - Megamerge 
       \ C ----/
          \ E /

Where E stacks on C and D stacks atop A. In the case above, A-E are revsets of either 1 or more commits. JJ doesn't care if they are or not. You'd generally bookmark the revset on the final "commit" as the pointer.

           / features/add-widgets
          /  / features/add-widget-integration
       / A - D \
      -- B ----- - Megamerge 
       \ C - E /
          \   \ feature/add-new-page
           \ feature/rework-navigation

In the example above, let's say you rework the navigation. You could have it exist alongside the navigation rework, but changes are you don't want to do the work twice. You just say "hey, this depends on the nav rework" and so it's there inside of the repo.

The thing is there is another way to do this where you end up with 4 different parents in a megamerge and your nav rework touches the megamerge and your new page is yet another revset is just a fork off of it. But yeah... JJ gives you a lot of flexibility in this manner.

The mega merge wouldn't handle that based on the way the article shows. You COULD have a revset that includes stacked changes, though. That does work and is what I currently do.

Great article, Isaac!

If anyone is JJ-curious, I also can't recommend the Discord[1] enough. The community is very helpful and welcoming.

[1]: https://discord.com/invite/dkmfj3aGQN

I initially used this one when I started playing around with Zed on a personal project, but I kept it and it has grown on me considerably.

I’ve been using Ollama for this on their $20 plan and I’m amazed that I’ve barely gotten to 5% utilization

This is New York. Not the US as a whole.

Love your other tools, btw!

You're right. It's _one_ of the prime places.

I don't know a single homeschooler that sits at home all day long. They work in family businesses, participate in bands, sports, and co-ops. Many belong to churches where families come from all different strata: our church has surgeons, line cooks, programmers, self-employed handymen, disabled vets. They interact with everyone—including kids. They do things like "kid markets" where they have a business. They watch their parents learn how the house works and how to manage finances.

There is no forced engagement—in fact the peer pressure is often completely gone. They are in an environment (their family) where they are much freer to be themselves.

He's still learning. Driven by what he loves. And this is on top of the "standard" stuff.

I've had to re-learn math skills long forgotten to help my kids with their school work. It's been an interesting experience.

The expectations for home schooling are different and are, in some ways, aimed more towards reality. My son finishes the bulk of his work in an hour most days and then has time for 2 instruments, learning C++, Rust, and Python, community/church participation and more.

All kids are indoctrinated. As parents do you want to have control of that or not?

With that attitude you might as well just tell parents that they shouldn't participate in society!

So interestingly... I went back to LinkedIn just a little while ago and there was a different message "mark as working" which I didn't see before.

For another provider, I was told I had to contact support as I had been "blacklisted" by Sendgrid and they had to issue a request to unblacklist me.

I watched the show but if I remember correctly they actually didn’t die. They survived.

> The reality is much more positive than the myth, with all three men escaping such a grisly fate. Indeed, Alexei Ananenko and Valeri Bespalov are believed to be both still alive as of 2024, while Boris Baranov lived until 2005 when he passed away from heart disease.

Source: https://www.history.co.uk/article/the-real-story-of-the-cher...

Same. What are the odds?

We live in an age where the commercialization/cheapening of sex is celebrated by society but the natural result of that commercialization/cheapening isn't wanted.

You can't have it both ways.

Our anthropology is confused and it shows.

Made it to 33… what fun!

This looks interesting. I'll check it out, thank you!

When doing web development I will occasionally connect my local code base to a remote SQL server via SSH.

This adds enough latency to be noticeable and I’ve found pages that were “OK” in prod that were unbearable in my local environment. Most of the time it was N+1 queries. Sometimes it was a cache that wasn’t working as intended. Sometimes it simply was a feature that “looked cool” but offered no value.

I’m not sure if there is a proxy that would do this locally but I’ve found it invaluable.

Sounds like he shot for the moon and missed.

I've been allowing LLMs to do more "background" work for me. Giving me some room to experiment with stuff so that I can come back in 10-15 minutes and see what it's done.

The key things I've come to are that it HAS to be fairly limited. Giving it a big task like refactoring a code base won't work. Giving it an example can help dramatically. If you haven't "trained" it by giving it context or adding your CLAUDE.md file, you'll end up finding it doing things you don't want it to do.

Another great task I've been giving it while I'm working on other things is generating docs for existing features and modules. It is surprisingly good at looking at events and following those events to see where they go and generating diagrams and he like.