linolevan

Encoder I’m working on as a side project, not publicly released (but feel free to pull the included files to prove it to yourself)

I'll post it to bsky/HN at some point in the future!

It's just a side project I'm working on with some friends. May release it if it's interesting to folks.

I think people are really forgetting that JXL's encoders are bad right now. For instance, in TFA Jake presents

- The Firefox screenshot w/ lossless JXL at 66.9kB, when with a better encoder you get get it down to like 40.7kB (39% smaller) [1] (or 43kB if you need it to work with jxl-rs [2])

- The cat illustration w/ lossless JXL at 255.3kB, when with a better encoder you can get it down to 178kB (30% smaller) [3]

[1] https://www.linolevan.com/05-screenshot.jxl

[2] https://www.linolevan.com/05-screenshot-web.jxl

[3] https://www.linolevan.com/06-illustration.jxl

39.2% (!!!) of people get diagnosed with cancer at some point in their lives

What is hard about migrating between password managers? At least on MacOS you can just open passwords and export all to another app?

While it is definitely true that original power drops were due to lower lumens, few things:

- reflective housing is super inefficient (~70% losses on the bounced light off typical material, at the top end expensive stuff still ~23%), still better than just losing the light upwards but not as many lumens on the road as a lab test would show

- Today LEDs are more efficient than sodium

Some math: 100W HPS cobra head lamp prdouces 9.5k lumens, including ballast it draws ~133W from the wall. Lab test shows about 70% of the lamp's light leaves the fixture (with the rest being lost in the housing) leading to ~7k lumens exiting. Roughly 53 lm/W.

LED cobra head lamps get like ~125 lm/W on the low end + fade much more slowly than sodium

Looks like 100% AI (https://www.alphaxiv.org/abs/2609.01877) except for citations?

Not quite a million USD, but ~$120k

https://abcnews.com/ABCNews/finnish-businessman-handed-12100...

I’m very skeptical of these results. I run a pretty large group paper website on top of arXiv and we run pangram on papers. The numbers are not nearly this high.

One thing I see a lot is papers flagged as AI because they include llm rollouts in the paper as examples.

Neat optimization, I've never seen that before. Thanks for sharing.

Can we note that this is a 2024 paper in the title?

Pangram is legit. I don't work at pangram, we integrated it in our paper website and one of the cool emergent behaviors I've seen is that on AI papers with example rollouts, it will accurately mark the paper's main text as human generated and the rollouts as AI generated.

My understanding is that they strongly believe in no false positives, so it's definitely possible to slip something by them but if it marks something as AI, it very likely is.

What I really don't get is why more people don't talk about LMStudio, I switched to it months ago and it seems like a straight upgrade.

Hosted on Parasail + Google (both for free, as of now) themselves, probably would give those a shot

I’m surprised that people are surprised. Qwen has been hosting private plus and max variants for a while now.

Well written piece on an attack vector I'd never thought too hard about before. Thanks for elaborating on why sending an email or two to a random person helps an attacker achieve their goal. A lot of similar articles skip over details like that.

How much is OpenAI paying you for this

https://tangled.org/ <--- GitHub on ATProto

that's all I'm aware of

(edit) Oops, just saw that you mentioned it, confused by your first line then. Tangled is awesome!

John too:

https://bsky.app/profile/johndonmoyer.com/post/3mhbkepa7gs2v

Where on earth are you living with that kind of price point? Unreal.