olejorgenb

6 ms·

61.
by olejorgenb 6mo ago
I've had good experiences with the Mistral Voxtral models (I've used the API, but some of the model-variants are open weight)
62.
by olejorgenb 6mo ago
Idk, it almost seem a workaround for slow/broken go-back? If go-back is fast and state preserving, it's basically a fullscreen modal. All(?) browser open links in a new tab when middle-clicked?
63.
by olejorgenb 6mo ago
Can you please fix https://github.com/anthropics/claude-code/issues/43713 ?
64.
by olejorgenb 6mo ago
Wish reword took a commit range though
65.
by olejorgenb 6mo ago
I don't want "standby", I want suspend where the only power usage is keeping the ram alive...
66.
by olejorgenb 6mo ago
Maybe they should fix bugs like this then https://github.com/anthropics/claude-code/issues/17979#issue... ...
67.
by olejorgenb 6mo ago
I see. If that's just hashes and not encrypted content I can't see how they can resume old sessions properly. IIRC they have a 30 days retention policy and surely the thinking traces must be considered data. Wonder how this works
68.
by olejorgenb 6mo ago
git bisect --first-parent
69.
by olejorgenb 6mo ago
So this means I can not resume a session older than 30 days properly?
70.
by olejorgenb 6mo ago
> I wonder if there is a more general solution that can make models spend more compute on making important choices, while making generation of the "obvious" tokens cheaper and faster. I think speculative decoding count as a (pe
71.
by olejorgenb 6mo ago
Source of that image though.. ?
72.
by olejorgenb 6mo ago
The tools was mostly already known, no? (I wish they had a "present" tool which allowed to model to copy-paste from files/context/etc. showing the user some content without forcing it through the model)
73.
by olejorgenb 6mo ago
> This is a production-grade agentic system that happens to live in your terminal. You read the code? > The ink/ directory — roughly 50 files — is not the popular npm ink package. Anthropic built their own React-based terminal
74.
by olejorgenb 7mo ago
To be fair, they display it reasonable prominently in GitHub when you are logged in. Given that, I feel the post title fall under the click bait category. I was fully aware of the Co-pilot opt-out change, but still clicked due the phrasing
75.
by olejorgenb 7mo ago
Do yourself a favor and upgrade your history search with fzf shell integration (or similar): https://youtu.be/u-qLj4YBry0?t=223 / https://junegunn.github.io/fzf/shell-integration/
76.
by olejorgenb 7mo ago
Opera was by far the best browser for a while for sure. Sad they couldn't keep up :/
77.
by olejorgenb 7mo ago
Well, the UI as a whole is ok to me (except the parts which is way too volatile). I was talking about the UX of the autocomplete model. The model are very often spot on and fast, but it's impossible to properly configure it to be less
78.
by olejorgenb 7mo ago
My impression is that they are a compression expert, not a color expert. Make sense they chose uniform flat colors :D
79.
by olejorgenb 7mo ago
The model is great. The UX is ~~horrible~~ annoying...
80.
by olejorgenb 7mo ago
To be fair, is "with RL", "just"? They should have disclosed it though. If they didn't it's a bad look for sure.
81.
by olejorgenb 7mo ago
> I was granted a patent for the Slug algorithm in 2019, and I legally have exclusive rights to it until the year 2038. But I think that’s too long. The patent has already served its purpose well, and I believe that holding on to it any
82.
by olejorgenb 7mo ago
As I'm not familiar with how Kagi is "pushing ai tools" this is mostly a comment on the framing of your question. Are you really saying that a company specializing in search - natural language oriented at its core - should
83.
by olejorgenb 7mo ago
I would trust Firefox if they made a version which did the indexing locally. I think I'd trust chrome as well as long as it was implemented locally.
84.
by olejorgenb 7mo ago
Ah :wooosh:, :D
85.
by olejorgenb 7mo ago
I'm 80%+ sure the claim is BS
86.
by olejorgenb 7mo ago
> Both Google Chrome and Microsoft Edge actually implement this. Behind the scenes, both will upload your browser history to the cloud. You can see it in network packet captures. It's implemented in the browser for the vendor, but n
87.
by olejorgenb 7mo ago
I believe there some projects trying this now using OCR IIRC. Did even Microsoft try something like this? It's of course something you'd only want running locally
88.
by olejorgenb 7mo ago
Off topic, but ref "I remember when you could half-remember a comment from a website, type that into Google, and get taken to the article you were looking for" It's funny to me that (to my knowledge) no browser (mainstream?)
89.
by olejorgenb 7mo ago
I find working more asynchronous with the agents help. I've disabled the in-your-face agent-is-done/need-input notifications [1]. I work across a few different tasks at my own pace. It works quite well, and when/if I find a r
90.
by olejorgenb 7mo ago
> So it seems to me like the problem isn't Bluetooth, it's shitty implementations of it. Perhaps, but it kinda seems to me that there must be some inherent limitation of the standard which make it particular hard to create a

More ›