floppyd

As the time goes on it only becomes harder to differentiate between model capabilities with just one or two numbers. I would love to see some kind of multi-axis placement of all the models on less objective attributes, like wordiness, willingness to give up, an ability to "think ahead" and pre-solve possible problems in code, for example, that I didn't think of or didn't think of talking about, etc etc etc.

For example I've been really enjoying Deepseek v4.1 Flash, it's very "straightforward" to the point of being almost dumb sometimes, but it's absolutely relentless and would solve almost any problem no matter how inefficient the solution is.

No idea how to measure all that, just average CoT length per task is probably a good approximation for some things, but not others.

At this point I already have "ad blindness" but towards "traditional news publication declaring some technology revolutionary" — probability of thw mentioned tech being a nothing burger is just getting closer and closer to 100% with every publication.

Full table with multipliers from previous prices:

DeepSeek-V4-Flash (off-peak, x2 for peak)

* Cache Hit $0.007 (x2.5)

* Cache Miss $0.22 (x1.5)

* Output $0.66 (x2.25)

DeepSeek-V4-Pro (off-peak, x2 for peak)

* Cache Hit $0.022 (x6)

* Cache Miss $0.66 (x1.5)

* Output $1.98 (x2.25)

Peak Hours: 01:00–04:00 and 06:00–10:00 UTC

Effective from: 16:00, August 16, 2026 (UTC)

About 2x-2.5x off-peak for Flash, 2x-4x I'd say for Pro (x6 on cache in, the biggest increase throughout the board). And twice as much in peak hours.

I personally think that it's just fun and “in the spirit” of writing to be pedantic about it and typographically expressive in general. In many ways, writing is a very creative process, because it involves “fixating thoughts”, making something very analogue into something almost (but not completely) discrete; it's akin to graphic design in general in my eyes. And when I can bring some extra flavour to it – by sticking to pedantic rules, using rare typographic marks that have a meaning that's more historical now, or anything else – I tend to do it, even if just for my own amusement.

Not having a bot API is what stops me from moving to Signal. I have 10-ish telegram bots running for myself, I just use telegram as UI for a script running on backend, and results are easily shareable with my other telegram contacts. I have no clue why not having this is amazing in any way.

Noticing a pattern and just extending it without proving why it works is not really a solution. You can prove it without really "understanding" it using induction, but that still would be proof, same as just counting on a computer.

> This is the key insight: we’re just telling the LLM “here are your tools, here’s the format to call them.” The LLM figures out when and how to use them.

This really blew my mind back then in the ancient times of 2024-ish. I remember the idea of agents just reached me and I started reading various "here I built an agent that does this" articles, and I was really frustrated at not understanding how the hell LLM "knows" how to call a tool, it's a program, but LLMs just produce text! Yes I see you are telling LLM about tools, but what's next? And then when I finally understood that there's no next, no need to do anything other than explaining — it felt pretty magical, not gonna lie.

> What do you think about Ukraine? Should it be a free country? What about its territorial integrity, to whom does Crimea belong?

Just wanted to let you know that asking those questions as if there can be any variation in my answers really shows you are judging russians from the outside. The Crimea question has been akin to a meme for a decade at this point. I'm obviously not saying "everyone" is thinking one way or another, because there's no "everyone", and there's also no "almost everyone". There are huge chunks of the population that are very much on the opposite ends of the spectrum, and judging one chunk by opinions and action of another one is highly disingenuous, imo.

> an entry point for Russians in Europe.

> we have opened an office to easier collaborate with sanctioned individuals

As a Russian living in Belgrade I love how you see "being open to people from Russia" as a negative, and how you just instantly equate being russian with being under sanctions.

I, and everyone I know of, left russia not because they're a weasel trying to help putin move his money in and out of Europe, I moved because I was afraid I'd be forced to go kill and be killed, and because I just didn't want to take part in any of it. And I chose Serbia because it's one of the rare few options that would welcome me without too much restrictions and that's not prohibitively expensive for me to live.

While Serbia obviously has a lot of problems, comparing a country that protests almost weekly to North Korea is extremely entitled.

The non-linear code structure (including visually) is something I've been thinking about for a long time and arrived at very naturally. I'm the "spread all the papers on the table to take in every interaction all at once" type of person, and so often I imagined a code editor that would allow me to just "cut" a piece of code and move it to the side. Separating stuff into files is kinda this, but it's not visual and just creates a lot of mess when I try to separate out small functions that are not reusable somewhere else. I don't even need the underlying non-linearity — just let me move the papers around on my code desk!

The year is 2076. An independent panel of experts has finally confirmed Sam Altman achieved AGI, for real this time. Quantum computers are factorizing numbers left and right. Cold nuclear fusion got so cold that we have to warm it up a little. Americans are still trying to communicate over something called "SMS", a text message protocol from 1993, but nobody knows why.

It's probably just under a huge load right now. Set it up pretty easily earlier on an M1 Air, but agent chats fail quickly with "model under load"

Gemma is an open-weight version of Gemini and obviously much less capable probably even than 2.5 Flash. Also the story you are linking to is a complete nothing burger, models are still very much hallucinating, especially on some extremely niche topics, I don't see how another politician trying to capitalize on that is attention-worthy at all.