memhole

From my perspective the issue is analytics support. You’ll need a step that turns it into something supported by BI tools. Obviously if something like Trino picks up the format it’s not an issue

On a backpacking trip, a guy and I were hiking together for a moment. I was snapping photos of the landscape and he started to chastise me for it. He made the same point. It’s about the people you’re with. Eventually, all the photos blur together.

He’s not wrong. I’ve got tons of pictures of the outdoors. Not that they aren’t beautiful. Pressed, I mostly couldn’t tell one from the other.

This is part of what I feel keeps driving AI/LLMs. The hope that there’s a new frontier to cash in on.

I’ll typically use the defaults initially and then use a Modelfile if it’s something I plan on using. I think you can dump the modelfile ollama uses to have a template to work with.

Solid advice. Too many people think build it and sell it. When if you did a few interviews and market research you’d realize it might be a flop.

Start ups that choose things like Azure are a red flag for me. Churn is really problematic. Learning a code base by having to poke and grok takes a lot longer than being able to fire off a few questions.

Very accurate description. I think this gets missed sometimes. Sometimes you’re criticizing because you know a subject well and want to see it improved.

Maybe it’s because I started with hydroponics. I don’t get the fascination with soil or animosity about hydroponics being unnatural. People do vastly underestimate what it takes to create a good soil mixture, though. In the end, you’re suspending nutrients in a substrate for the plants to uptake regardless of how you go about providing them.

I want to say it was a buoy. I might be wrong on that. I distinctly remember one of those plastic flamingos. If you’re asking about my event idea. Unfortunately, we didn’t end up doing it.

USPS will mail all sorts of things. WIRED would let you mail them tons of interesting things. Working remotely I thought it would be hilarious to have everyone try and mail each other weird stuff as a company event.

I’ve seen papers that show they can do math. There was one recent-ish one HN that showed an understanding of addition. I’m not convinced myself. At least the open weight models don’t seem to grasp integers. Doing conversions are typically a flop as well. My doubt really comes from how abstract mathematics really is. It’s entirely its own kind of terse language and symbolism. Maybe if there was the kind of focus on it like coding there would be better results?

Likewise been going to synagogue. I would hardly call myself a religious person. Far more a nihilist who thinks we should try and do good if it is truly all pointless. I think the meaningfulness and the sense of community is nice. It’s typically better than being glued to the kinetoscope

I agree. I’m sure there’s limitations, but svg feels more like a wysiwyg for web design than css

For anyone vibe checking the results:

https://novelchallenge.github.io/

At best you’re looking at 60% or so accuracy

I personally like cooperatives too. I’ve always wondered if instead of a tech union a cooperative might be a better fit. No solution is perfect, though. Lots of people want to make an income and not deal with the now governance part of their job. You also have people that are attracted to those kinds of organizations who also desire the governance part. Which goes back to the trope about those that should have authority don’t often want it.

This is called the hamster wheel and destroys trust rather than builds it, ime. Maybe it communicates trust to the org? The people within the team will question the motives. One piece of advice I read early on, was don’t bullshit your engineers. They know.

I’m a big fan of value as prioritization. Work on the things that deliver value. However value is defined. Revenue, nps, etc. Ime, small companies don’t care about deadlines or they shouldn’t. They care about what’s delivering value or the next outcome. It’s only as you grow the company suddenly people want deadlines. Or you have small companies that misunderstand what their focus should be. There are obviously some time constraints that need deadlines. You can’t work on something for the holidays and deliver in Feb.

Nice to see some experiments with languages. I really like the way it looks.

With all the AI focus on coding, it feels like we're going to end up sticking with a few languages because that's what it's good at or trained on.

Any examples you mind sharing? I’m largely convinced it’s due to the founder’s networks and access to capital when you see broken ideas succeed.

Survivorship bias at its finest. For as many people have been lived before ourselves there’s only a handful that get remembered on any scale.

I think the unique perspective part is something I’m understanding. It’s easy to read or see other people’s work and think you don’t have anything to add

It looks like a great overview of recommendation systems. I think my main takeaways are:

1. Latency is a major issue.

2. Fine tuning can lead to major improvements and I think reduce latency. If I didn’t misread.

3. There’s some threshold or problems where prompting or fine tuning should be used.

Nice work! I guess my only thought would be if you’re trying to turn this into a product, is this basically a client server rdbms. What advantages does using embedded dbs in this way provide compared to something like Postgres or MySQL? Or the growing popularity of duckdb on s3?

Nice they’re expanding it to include more people. When I was looking at applying to colleges one of the major themes I saw about going to ivy schools is culture differences. You’re mixing low income students with middle and higher. Their life experiences and how they relate to each other can be very different. That can definitely affect academic success and degree completion

I like this about quarto. You can create a standalone html file and share with people.

Thanks! I had no idea. I have an issue with a large gap on mobile I should dig into. Maybe this will help me figure out how to fix it. Also just a neat fact to know regardless

So things can be styled based on the user agent? Besides the obvious difference between mobile and desktop.

I believe NYC has had one running for quite some time.

Sound proofing is really challenging. I've gone through similar dilemmas as the author. Just short of blocking all ability for air or vibrations to transfer, I don't think there's much you can do. Soft textures absorb sounds, which can kind of help. If it's via air ducts you can try to add baffles. I've been curious if there's studies on noise pollution and mental health.

Thanks for clarifying that. I can understand their take. But, I also think of the models without abliteration as lobotomized

Yeah, that’s why I like the smaller models too. Big context windows and intelligent enough most of the time. They don’t follow instructions as well as the larger models ime. But then on the flip side the reasoning models struggle to deviate. I gave deepseek an existential crisis by accident the other day lol.

Agreed on personalities. Phi, I think because of the curated training data comes across as very dry.