haolez

7 ms·

91.
by haolez 1y ago
No obscene fishes observed. Well done! :)
92.
by haolez 1y ago
Anecdote on my side, but Grok 4 was the first model that didn't feel like that to me. Maybe my conversations were just not long enough for it to fallback to sycophantic behavior.
93.
by haolez 1y ago
Maybe also the fact that they can't learn small pieces of new information without "formatting" its whole brain again, from scratch. And fine tuning is like having a stroke, where you get specialization by losing cognitive cap
94.
by haolez 1y ago
My definition of local is the same as yours. It's just that my mind got focused on the open models part and I forgot that you specifically mentioned a "local" setup. My bad
95.
by haolez 1y ago
How do you do it? Do you host on your hardware or do you use cloud-based providers for open models?
96.
by haolez 1y ago
I wonder if these extremely terse languages might see a rise in popularity with LLMs, since they save a lot of tokens. The hard part would be training on them, since training data is not as comprehensive as JavaScript, for example. Fun time
97.
by haolez 1y ago
Broken for me as well
98.
by haolez 1y ago
This assumes that the benchmarks are representative of real usage scenarios. I'm not saying that there is bad faith, but that benchmarking is really hard in the context of LLMs.
99.
by haolez 1y ago
This would be super cool if it becomes a smart TV app. Couldn't be too far off, right?
100.
by haolez 1y ago
Astronomy and quantum physics :)
101.
by haolez 1y ago
I wonder what chain of events led to this design in the PM's mind.
102.
by haolez 1y ago
Well, your prompt adds up to the baseline. The logic still applies.
103.
by haolez 1y ago
Sometimes I get the feeling that making super long and intricate prompts reduces the cognitive performance of the model. It might give you a feel of control and proper engineering, but I'm not sure it's a net win. My usage has con
104.
by haolez 1y ago
This looks amazing. What kind of use cases have motivated its creation? I mean, combining Lua and all these runtime goodies.
105.
by haolez 1y ago
Well, the JS in the name is for JavaScript :)
106.
by haolez 1y ago
Anti-spam and offensive content might be a challenge for small initiatives.
107.
by haolez 1y ago
And then that one friend makes 100 posts per day and you can't see anything else in all this noise :)
108.
by haolez 1y ago
My main issue with Android as a daily driver for desktop computing is the input latency. Even in high end devices, it's noticeable when typing and very noticeable for things like Alt-Tab and other desktop shortcuts. I wonder if this is
109.
by haolez 1y ago
There are probably several, but I had contact in the beginning of my career with a company that made industrial printers. They said that, in the first years of the company (80s), adopting FORTH gave them an edge over the competitors and it
110.
by haolez 1y ago
Man, I played so much of Adventure in my Atari 2600 that I had dreams of this game. I dreamt that I was playing levels that didn't exist. It was fascinating at the time. It was a unique experience, almost psychedelic :D
111.
by haolez 1y ago
Makefiles are bad with state that is not represented as files, right? I remember that I had to create "stamp" files with hashes and things like that to overcome this limitation. There is also .PHONY, but that would make the rule t
112.
by haolez 1y ago
This is WSL2, right? I had the impression that WSL1 provided a better experience, but it was too complex to maintain and Microsoft decided to take the easy road.
113.
Available Free Ollama Servers(freeollama.com)
3 points by haolez 2y ago|1 comments
114.
by haolez 2y ago
Vibe deployments of ollama :)
115.
by haolez 2y ago
This made me remember Fleet[0], which I used in the past on CoreOS. It was a very good abstraction for my simple cluster. It felt as robust as a managed cloud service. It's a shame it didn't catch on (and also what happened to Cor
116.
by haolez 2y ago
This reminded me of this old quote from Hal Abelson: "Underlying our approach to this subject is our conviction that "computer science" is not a science and that its significance has little to do with computers. The computer
117.
by haolez 2y ago
Not exactly related, but I wonder sometimes if the fact that the weights in current models are very expansive to change is a feature and not a "bug". Somehow, it feels harder to trust a model that could evolve over time. It's
118.
by haolez 2y ago
I thought like you in the past. Today, for some reason, I value defaults and reducing my cognitive load so that I can think more and do less. Even Eclipse would work for me nowadays :P
119.
by haolez 2y ago
Cool! The only downside would be keeping your code in sync with upstream, then.
120.
by haolez 2y ago
Maybe it's easier to just set up a proxy mirroring Anthropic's API, but pointing to whatever model you want? Genuine doubt.

More ›